International Journal of
AL
RNAL OF
RE
JOU
M
TI
O
198
0 – 2009
incorporating
Remote Sensing Reviews
Volume 30
Numbers 13–14
ISSN 0143–1161
July 2009
SING
th
Anniversary
SEN
INTERNA
TE
REMOTE 30
SENSING
O
International Journal of REMOTE SENSING
N
Volume 30
Numbers 13–14
July 2009
This article was downloaded by: [US Geological Survey Library]
On: 24 July 2009
Access details: Access Details: [subscription number 907834750]
Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
International Journal of Remote Sensing
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713722504
Global irrigated area map (GIAM), derived from remote sensing, for the end of
the last millennium
Prasad S. Thenkabail a; Chandrashekhar M. Biradar b; Praveen Noojipady c; Venkateswarlu Dheeravath d;
Yuanjie Li c; Manohar Velpuri e; Muralikrishna Gumma f; Obi Reddy P. Gangalakunta g; Hugh Turral h;
Xueliang Cai f; Jagath Vithanage f; Mitchell A. Schull i; Rishiraj Dutta j
a
Southwest Geographic Science Center, U.S. Geological Survey (USGS), Flagstaff, AZ 86001, USA b
University of Oklahoma, Norman, OK 73019, USA c Department of Geography, University of Maryland
College Park, MD 20742, USA d United Nations Joint Logistics Center, Juba, Sudan e Geographic Information
Science Center of Excellence, South Dakota State University, Brookings SD 57007, USA f International Water
Management Institute (IWMI), Colombo, Sri Lanka g National Bureau of Soil Survey and Land Use Planning,
Indian Council of Agricultural Research (ICAR), Nagpur, India h On the Street Productions, North Carlton,
Australia i Department of Geography, Boston University, USA j Indian Institute of Remote Sensing, Dehra
Dun, India
Online Publication Date: 01 January 2009
To cite this Article Thenkabail, Prasad S., Biradar, Chandrashekhar M., Noojipady, Praveen, Dheeravath, Venkateswarlu, Li, Yuanjie,
Velpuri, Manohar, Gumma, Muralikrishna, Gangalakunta, Obi Reddy P., Turral, Hugh, Cai, Xueliang, Vithanage, Jagath, Schull,
Mitchell A. and Dutta, Rishiraj(2009)'Global irrigated area map (GIAM), derived from remote sensing, for the end of the last
millennium',International Journal of Remote Sensing,30:14,3679 — 3733
To link to this Article: DOI: 10.1080/01431160802698919
URL: http://dx.doi.org/10.1080/01431160802698919
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
International Journal of Remote Sensing
Vol. 30, No. 14, 20 July 2009, 3679–3733
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map (GIAM), derived from remote sensing, for the
end of the last millennium
PRASAD S. THENKABAIL*{, CHANDRASHEKHAR M. BIRADAR{,
PRAVEEN NOOJIPADY§, VENKATESWARLU DHEERAVATH",
YUANJIE LI§, MANOHAR VELPURI**, MURALIKRISHNA GUMMA{{,
OBI REDDY P. GANGALAKUNTA{{, HUGH TURRAL§§,
XUELIANG CAI{{, JAGATH VITHANAGE{{, MITCHELL A. SCHULL""
and RISHIRAJ DUTTA***
{Southwest Geographic Science Center, U.S. Geological Survey (USGS), 2255 N.
Gemini Drive, Flagstaff, AZ 86001, USA
{University of Oklahoma, 101 David L. Boren Blvd, Norman, OK 73019, USA
§Department of Geography, University of Maryland College Park, MD 20742, USA
"United Nations Joint Logistics Center, Juba, Sudan
**Geographic Information Science Center of Excellence, South Dakota State University,
Brookings SD 57007, USA
{{International Water Management Institute (IWMI), 127 Sunil Mawatha, Colombo,
Sri Lanka
{{National Bureau of Soil Survey and Land Use Planning, Indian Council of
Agricultural Research (ICAR), Nagpur, India
§§On the Street Productions, 28 Newry Street, North Carlton, Melbourne 3054, Australia
""Department of Geography, Boston University, USA
***Indian Institute of Remote Sensing, Dehra Dun, India
(Received 26 December 2007; in final form 12 June 2008 )
A Global Irrigated Area Map (GIAM) has been produced for the end of the last
millennium using multiple satellite sensor, secondary, Google Earth and
groundtruth data. The data included: (a) Advanced Very High Resolution
Radiometer (AVHRR) 3-band and Normalized Difference Vegetation Index
(NDVI) 10 km monthly time-series for 1997–1999, (b) Système pour
l’Observation de la Terre Vegetation (SPOT VGT) NDVI 1 km monthly time
series for 1999, (c) East Anglia University Climate Research Unit (CRU) rainfall
50 km monthly time series for 1961–2000, (d) Global 30 Arc-Second Elevation
Data Set (GTOPO30) 1 km digital elevation data of the World, (e) Japanese Earth
Resources Satellite-1 Synthetic Aperture Radar (JERS-1 SAR) data for the rain
forests during two seasons in 1996 and (f) University of Maryland Global Tree
Cover 1 km data for 1992–1993. A single mega-file data-cube (MFDC) of the
World with 159 layers, akin to hyperspectral data, was composed by re-sampling
different data types into a common 1 km resolution. The MFDC was segmented
based on elevation, temperature and precipitation zones. Classification was
performed on the segments.
Quantitative spectral matching techniques (SMTs) used in hyperspectral data
analysis were adopted to group class spectra derived from unsupervised
*Corresponding author. Email: pthenkabail@usgs.gov. Formerly of International
Water Management Institute (IWMI), 127 Sunil Mawatha, Colombo, Sri Lanka.
International Journal of Remote Sensing
ISSN 0143-1161 print/ISSN 1366-5901 online # 2009 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/01431160802698919
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3680
P. S. Thenkabail et al.
classification and match them with ideal or target spectra. A rigorous class
identification and labelling process involved the use of: (a) space–time spiral curve
(ST-SC) plots, (b) brightness–greenness–wetness (BGW) plots, (c) time series
NDVI plots, (d) Google Earth very-high-resolution imagery (VHRI) ‘zoom-in
views’ in over 11 000 locations, (e) groundtruth data broadly sourced from the
degree confluence project (3 864 sample locations) and from the GIAM project
(1 790 sample locations), (f) high-resolution Landsat-ETM + Geocover 150 m
mosaic of the World and (g) secondary data (e.g. national and global land use and
land cover data). Mixed classes were resolved based on decision tree algorithms and
spatial modelling, and when that did not work, the problem class was used to mask
and re-classify the MDFC, and the class identification and labelling protocol
repeated. The sub-pixel area (SPA) calculations were performed by multiplying
full-pixel areas (FPAs) with irrigated area fractions (IAFs) for every class.
A 28 class GIAM was produced and the area statistics reported as: (a) annualized
irrigated areas (AIAs), which consider intensity of irrigation (i.e. sum of irrigated
areas from different seasons in a year plus continuous year-round irrigation or gross
irrigated areas), and (b) total area available for irrigation (TAAI), which does not
consider intensity of irrigation (i.e. irrigated areas at any given point of time plus the
areas left fallow but ‘equipped for irrigation’ at the same point of time or net irrigated
areas). The AIA of the World at the end of the last millennium was 467 million
hectares (Mha), which is sum of the non-overlapping areas of: (a) 252 Mha from
season one, (b) 174 Mha from season two and (c) 41 Mha from continuous yearround crops. The TAAI at the end of the last millennium was 399 Mha. The
distribution of irrigated areas is highly skewed amongst continents and countries.
Asia accounts for 79% (370 Mha) of all AIAs, followed by Europe (7%) and North
America (7%). Three continents, South America (4%), Africa (2%) and Australia
(1%), have a very low proportion of the global irrigation. The GIAM had an
accuracy of 79–91%, with errors of omission not exceeding 21%, and the errors of
commission not exceeding 23%. The GIAM statistics were also compared with: (a)
the United Nations Food and Agricultural Organization (FAO) and University of
Frankfurt (UF) derived irrigated areas and (b) national census data for India. The
relationships and causes of differences are discussed in detail. The GIAM products
are made available through a web portal (http://www.iwmigiam.org).
1.
Introduction, background and rationale
The population of the world is now approaching 6 billion and is expected to near
8 billion by 2025. Some estimate that, to meet future food demand, at least another
2000 km3 of water (equivalent to the mean annual flow of 24 additional Nile rivers)
will be needed (Postel 1999). Irrigation is widely thought to provide 40% of the
world’s food from around 17% of the cultivated area. It accounts for 2–4% of
diverted water in Canada, Germany and Poland, but is an impressive 90–95% in
Iraq, Pakistan, Bangladesh, Sudan, Kyrgyzstan and Turkmenistan (Merrett 2002).
However, the actual areas irrigated and their spatial distributions can be further
improved using modern remote sensing data. Given that nearly 80% of all freshwater
used by humans is for irrigation, the importance of irrigated areas cannot be
overemphasized.
Following the end of the Second World War and a period of decolonization, there
was a boom in irrigation development, particularly in Asia, which coincided with
strongly motivated nation building, poverty alleviation and famine eradication. In
this era, a key developmental agenda for many countries was the construction of
large and small dams and river diversions to abstract and store water for agriculture.
Over 40 000 large dams (.15 m in height) irrigate about 30–40% of the world’s
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
3681
irrigated areas (www.dams.org) and are complemented by an estimated 800 000
smaller dams. Irrigated areas increased at about 2.6% per annum from a modest
95 million hectares (Mha) in the early 1940s to between 250 and 280 Mha in the early
1990s (Van Schilfgaarde 1994, Seckler et al. 2000, Siebert et al. 2005a,b, 2006). These
are massive increases when compared with the earlier era when irrigated areas
increased form a meagre 8 Mha in year 1800 to 95 Mha by 1940. Since the 1980s,
there has been a progressive decline in public and international donor funding for
irrigation, which has been replaced in many countries by the private development of
groundwater irrigation based on availability of cheap drilling and pumping
technologies. The number of groundwater wells in India, for example, are now
estimated at 26 million, followed by the USA (16 million), China (3.4 million),
Bangladesh (800 000), Pakistan (700 000), Germany (500 000) and South Africa
(500 000) (see Shah et al. 2003, 2004, Endersbee 2005, www.wellowner.org,). Yet, the
areas irrigated from groundwater are often missing from the statistics and maps
produced (e.g. the Central Board of Irrigation and Power (CBIP) 1994 map). At
present, globally, the irrigated landscape remains very dynamic. Although the
annual rate of increase of irrigated areas has slowed to about 1%, this still represents
an increase of between 3 and 4 Mha each year. There is a smaller corresponding
annual loss of irrigated area to salinity and water logging, as well as abandonment of
uneconomic projects. Countries, such as China and India, continue to build large
multi-purpose dam projects that also supply water for irrigation. In sub-Saharan
Africa, irrigation is perennially seen as having unfulfilled potential. Elsewhere in the
world, there are moratoria on dam building and even on the decommissioning of
dams in western USA.
There remains considerable uncertainty about the exact extent or area,
cropping intensity and the precise spatial distribution of irrigated areas in
different parts of the world due to both the absence of systematic irrigated area
mapping at global level and systematic problems in underreporting and overreporting of irrigation in different contexts (e.g. groundwater). Indeed, often the
irrigated area statistics do not include minor or informal irrigated areas (e.g.
groundwater, small reservoirs and tanks). Yet, in many countries, minor irrigated
areas are very significant and even exceed the major irrigated areas (e.g. major
and medium reservoirs created by building large dams and barriers) (MoWR
2005). The United Nations Food and Agriculture Organization (FAO)/University
of Frankfurt (UF) study on irrigated areas of the world (Döll and Siebert 2000,
Siebert et al. 2005a,b, 2006) is primarily based on FAO AQUASTAT statistics,
which, in turn, is based on census statistics from individual nations. It provides
estimates of area ‘equipped’ for irrigation (but not necessarily irrigated) in the
world as 278.8 Mha around year 2000 (see Siebert et al. 2006), which is about
19% of the total croplands (1.5 billion ha) around year 2000. Irrigated areas are
estimated, rather coarsely, in global land use classifications (DeFries et al. 1995,
1998, 2000a,b, Loveland et al. 2000, Bartholome and Belward 2005) derived from
remote sensing, which are usually focused on other objectives, such as forestry,
rangelands and rain-fed croplands.
There will be other causes for uncertainty in irrigated areas in the near future.
Rain-fed croplands are identified as areas for productivity increases (CA 2007) and
may yet have an impact on limiting expansion of irrigated areas in the coming
decades. If serious advances are made in using less water to produce more food
(better water productivity), irrigated areas may drastically change. Spatial
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3682
P. S. Thenkabail et al.
distribution of irrigated areas may also change if the concept of ‘virtual water trade’
(where countries with surplus water grow food and export to water-deficit countries
for other trade benefits) takes hold. Irrigated croplands are significantly being
converted to bio-fuel farms in certain parts of the world. Genetic engineering may
help increase yields, but is increasingly questioned by environmental activists and
more ecologically sensitive governments. The irrigated landscape of the world will be
shaped increasingly by the effects of competition for water from other sectors,
notably urban and rural domestic water supply and industrial needs. Groundwater
overdraft may ultimately exhaust and/or substantially reduce irrigated areas in the
Ogallala aquifer in the mid-western USA, northeast China and most of India.
Reserving and reallocation of flows for environmental and health purposes will, in
the end, place even greater competing demands in terms of water volume. River
basins are becoming over-allocated, as in the case of the Krishna basin in India
leading to reallocation of water and change in spatial distribution of irrigated areas
(Biggs et al. 2006). Climatic change will impose additional challenges that will
reshape the irrigated landscape through changes in snowmelt runoff and rainfall.
The greater the certainty in area estimation and geographic precision, the greater
the certainty in water-use calculations and food-production planning. For example,
in India, in order to produce 1 kg of rice, 3700 l of water are evaporated, whereas 1 kg
of wheat evaporates 2560 l and 1 kg of maize evaporates 4350 l (http://en.wikipedia.
org/wiki/Virtual_water, http://www.clw.csiro.au/issues/water/water_for_food.html).
In China, 1 kg of rice, wheat and maize require much lesser amounts of water at
1370, 1280 and 1190 l, respectively. In the USA, this comes to 1920, 1390 and 670 l
for rice, wheat and maize, respectively. In comparison, 1 kg of beef requires 14 379 l
in India, 12 600 l in China, and 10 060 l in the USA.
Based on the above needs and possibilities, the International Water Management
Institute (IWMI) initiated a Global Irrigated Area Mapping (GIAM) project. We
used the availability of a wide range of increasingly sophisticated remotely sensed
images and techniques to reveal vegetation dynamics that:
N
N
N
define more precisely the actual area and spatial distribution of irrigation in the
world;
elaborate the extent of multiple cropping over a year, particularly in Asia,
where two or three crops may be planted in one year, but where cropping
intensities are not accurately known or recorded in secondary statistics; and
develop methods and techniques for consistent and unbiased estimates of
irrigation over space and time for the entire world.
Thereby, the overarching goal of this research was to create a GIAM by
developing repeatable and robust methods and techniques of analysis using remote
sensing data. Two types of irrigated areas will be reported: (1) TAAI, which does
not consider the intensity of irrigation and (2) annualized irrigated areas (AIA),
which consider the intensity of irrigation by summing areas from different seasons
and perennial crops such as plantations. Specific emphasis will be placed on
mapping classes of: (i) major irrigation from large and medium surface-water
reservoirs and (ii) minor irrigation from groundwater, small reservoirs and tanks.
Through this effort, it was envisaged to: (a) provide irrigated area statistics and
maps for every country in the world, (b) determine accuracies and uncertainties in
area estimates and (c) compare them with results from FAO/UF (Siebert et al.
2006) and national statistics. The study is expected to provide baseline remote
Global irrigated area map
3683
sensing based data and products on global irrigated areas at the end of the last
millennium.
2.
Methods and materials
First, we describe the data sets and the reasons for choosing them. This will be
followed by the methods used.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
2.1
Data used in the creation of the IWMI’s GIAM
The process used in this study starts with a number of publicly available primary and
secondary data sets, which are processed into one single large 159 layer time series
file, known as a mega-file data-cube (MFDC) (see illustrations in figure 1 and
table 1), similar to a hyperspectral data-cube (Thenkabail et al. 2004a,b). The dropdown menu illustrates how the data layers are composed in the MFDC (e.g. figure 1),
which consisted of 144 Advanced Very High Resolution Radiometer (AVHRR)
10 km layers for the years 1997–1999 (4 bands * 3 years * 12 months; with red, nearinfrared, thermal infrared band number 4 and a scaled Normalized Difference
Vegetation Index (NDVI) band), 12 Système pour l’Observation de la Terre
vegetation (SPOT VGT) 1 km data layers (see table 1), each for every month of the
year 1999, a single layer of Global 30 Arc-Second Elevation Data Set (GTOPO30)
digital elevation model (DEM) 1 km, mean annual rainfall for 40 years at 50 km
resolution and AVHRR-derived forest cover at 1 km. All layers in the MFDC were
resampled to 1 km to analyse data at the SPOT VGT resolution. This increases the
data volume, but makes it possible to view the data characteristics such as band
reflectivity, the NDVI of different sensors, precipitation, elevation and temperature
at a click of a mouse, instantaneously, for any given point in the world. However,
since the overwhelming numbers of data layers were from AVHRR, the final
product is referred to as a nominal 10 km. Co-registration of the MFDC required
very careful synthesis as a result of inherent difficulties associated with varying
resolutions: AVHRR, 10 km; SPOT VGT, 1 km; Japanese Earth Resources Satellite1 (JERS-1) Synthetic Aparture Radar (SAR), 100 m; Precipitation, 50 km; and
GTOPO30, 1 km. Co-registration was achieved using ground control points (GCPs)
matched between the two different types of images (e.g. AVHRR versus SPOT)
resampled to 1 km. Polynomial warping with nearest neighbour resampling was
preferred because of its simplicity. An evaluation was conducted using spectral
values between the original and wrapped images from specific locations. The results
showed that the resampled AVHRR and SPOT images retained spectral integrity
and other data, such as rainfall (mm yr21) and elevation (m), had the same values in
specific geographic locations in comparison to their original resolution data. The
multi-sensor data sets widely vary in their spectral, spatial and radiometric
characteristics, have gone through complex normalization algorithms to correct
for issues such as Sun elevation, Earth–Sun distance, sensor calibration coefficients
and cloud and haze removal. All this will add its own uncertainties in irrigated area
estimates. Recognizing this, the coarser resolution time series used in this study from
the AVHRR pathfinder and the SPOT VGT are supported by: (a) high-quality
secondary spatial data such as GTOPO30, precipitation and temperature, (b)
JERS-1 SAR, (c) high resolution ‘groundtruth’ from Landsat Geocover, Google
Earth and (d) actual groundtruth from degree confluence and GIAM projects. In
addition, sophisticated and rapid access to groundtruth data from Google Earth
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3684
Figure 1. Mega-file data cube (MFDC): (a) A single global file MFDC of 159 data layers,
consisting of time-series primary and secondary satellite sensor data from various sources. (b)
illustrates the mega-file, at any given point, providing characteristics of all the 159 data layers.
Note: SNDVI 5 scaled normalized difference vegetation index, AVHRR 5 Advanced Very
High Resolution Radiometer, SPOT 5 Système pour l’Observation de la Terre Vegetation,
NIR 5 near-infrared. X-axis provides time-series values of MFDC month after month. Y-axis
represents AVHRR or SPOT digital numbers of bands or SNDVI in 8-bits. The rainfall in
mm\month and tree cover class numbers. B50:9802-b2: represents band 50 in the MFDC
which is for year 1998, month 2, and band 2.
Table 1a
Band number Or
primary source (#)
Satellite sensor data
AVHRR 10-km
Band 1 (B1)
Band 2 (B2)
Band 4 (B4)
(top-of-atmosphere)
NDVI
Secondary data
GTOPO30 1-km
one-band
Rainfall 1-km
one-band
Forest cover 1-km
one-band
Wavelength range (mm)
Duration
(years)
Number of bands1 (#)
Data final format radiometry
(percent: for reflectance)
Range Z-scale
(dimensionless)
0.58 – 0.68
0.73–1.1
10.3–11.3
1997–1999
1997–1999
1997–1999
36
36
36
reflectance @ ground, 8-bit
reflectance @ ground, 8-bit
Brightness temperature
0–100
0–100
160–340
(B22B1)/(B2 + B1)
1982–2000
36
unitless, 8-bit scaled NDVI
21 to + 1
DCW, DTM, and others2
1 time
1
meters, 16-bit
21 to + 1
Mean of monthly 40-years
1961–01
1
mm, 16-bit
0–65536
None
1992–93
1
class names, 8-bit
0–256
same as above
1981–2001
239*
(B32B2)/(B3 + B2)
1999
12
unitless, 8-bit scaled NDVI
21 to + 1
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Table 1. Characteristics of the Mega-file datasets used in the study. The characteristics of the primary satellite sensor time-series data and the secondary
datasets (see Table 1a) as well as other datasets (see Table 1b).
Table 1b
1. Band 1, 2, NDVI
2. SPOT 1-km2
NDVI
3. JERS SAR 100-m
one-band
L-band;24.5 cm
3685
Jan. –Mar 1996 1
unitless, 8-bit
0–256
Oct-Nov 1996 1
unitless, 8-bit
0–256
Note: 1 5 for satellite sensor data: 36 bands from 3 years with 1 band per month. 2 5 DCW 5 digital chart of the World, DTM 5 digital terrain model.
* 5 animations of the irrigated area classes were run for the entire AVHRR time-series data to help understand the change history of the class.
There was data for 239 months in 19 years (July 1981– September 2001). September-December 1994 data was not acquired due to failure of the satellite.
3686
P. S. Thenkabail et al.
‘zoom-in views’ of very-high-resolution imagery (VHRI), degree confluence data and
actual groundtruth data. Combinations of these data sets, at global levels, make it
feasible to map and study irrigated areas and help determine their uncertainty.
The following sections provide a brief description of each of the data sets, which
are summarized in detail in table 1. Readers interested in further details of these
data may look into detailed documentation in the web portals of the GIAM
project (http://www.iwmigiam.org), IWMI’s data storehouse pathway or IWMIDSP
(http://www.iwmidsp.org) and in various references (Thenkabail et al. 2005, 2006,
2007a,b, Biggs et al. 2006).
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
2.2
Primary remote sensing datasets
2.2.1 AVHRR data characteristics. The monthly time-composite NOAA AVHRR
0.1u data were obtained from the NASA Goddard DAAC site (www.daac.gsfc.gov/
data/data set/AVHRR). This ‘Pathfinder’ data set has gone through many stages of
calibration and recalibration (Kidwell 1991, Rao 1993a,b, Agbu and James 1994,
Smith et al. 1997) and normalization (Fleig et al. 1984, NGDC 1994, Kogan and Zhu
2001), making it a high-quality science data set and minimizing the known
limitations (see Eidenshink and Faundeen 1994, Loveland et al. 1999, 2000, http://
daac.gsfc.nasa.gov/www/islscp/). The original scaled 16 bit and 8 bit data have been
converted to three primary variables: (a) at-ground reflectance, (b) top of
atmosphere brightness temperature and (c) NDVI. These parameters were derived
using calibration parameters (Abu and James 1994, Smith et al. 1997). In the GIAM
project, the monthly data of AVHRR band 1, band 2, thermal band 4 and NDVI
maximum value composite (MVC; Holben 1986) were used for the years 1997–1999
(figure 1 and table 1).
2.2.2 SPOT data characteristics. The SPOT VGT (Lissens et al. 2000) 1 km NDVI
10 day synthesis for year 1999 was downloaded for the entire world (http://
free.vgt.vito.be/), converted to monthly MVCs (Holben 1986, Thenkabail et al. 2005,
Biggs et al. 2006) and used in this study (figure 1 and table 1).
2.2.3 JERS-1 SAR-derived forest cover. Mapping irrigated areas in rain forests is
more complex than in other parts of the world as a result of forest fragmentation,
significant cloud cover and the presence of natural wetlands. Therefore, we obtained
100 m resolution JERS-1 SAR L-band (24.5 cm wavelength) imaging radar tiles
(http://southport.jpl.nasa.gov/GRFM/, Saatchi et al. 2000) in conjunction with
AVHRR, SPOT and secondary data for South America to assist us in mapping
areas in major rain forest areas. These images were classified separately and the class
backscatter coefficients were determined and linked to groundtruth knowledge to
understand irrigation versus no irrigation. Normalized radar cross section (sigma0)
is measured in decibels (dB) and is used for quantitative characterization of land
cover and land use (Saatchi et al. 2000). Typical values of sigma0 for natural
surfaces range from + 5dB (very bright) to 240dB (very dark) (Saatchi et al. 2000).
Qualitatively, flooded irrigated lands will appear bright, and drier targets will
appear dark. Young vigorous irrigated crops appear very bright. The smooth body
of water will act as a flat surface and reflect incoming pulses away from a target;
these bodies will appear dark. Forests appear medium bright and clear-cut areas
very dark.
Global irrigated area map
2.3
3687
Mask data for stratification
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Secondary data sets (table 1) in the mega-file are used to stratify or segment the
world into characteristic regions based on precipitation, elevation, temperature and
forest cover. The MFDCs are created for each of the seven segments listed below,
classified and investigated for presence or absence of irrigation. The seven global
masks are:
N
N
N
N
N
N
N
precipitation less than 360 mm yr21 (PLT360);
precipitation greater than 2400 mm yr21 (PGT2400);
temperature less than 280 K yr21 (TLT280);
forest cover greater than 75% canopy cover (FGT75);
special forest SAR (FSAR);
elevation higher than 1500 m (EGT1500); and
all other areas of the world (AOAW) that are outside the above six segments.
The above seven segments cover the area of the entire terrestrial world. MFDCs
were composed for each of the above seven segments. The segments were used to
generate class spectra using unsupervised classification. The classes were then further
investigated to identify and label them. Segmentation helps in focusing on particular
precipitation, temperature, forest cover and elevation zones and helps us in analysing
areas within these zones.
2.3.1 Climate Research Unit (CRU) precipitation. The 40 year (1961–2000)
monthly, 0.5u, interpolated precipitation data were obtained from Dr Tim Mitchell of
the CRU, University of East Anglia, UK (Mitchell et al. 2003, http://www.cru.uea.ac.uk/
,timm/index.html). Two precipitation segments, one where precipitation is less than
360 mm yr21 (PLT360) and another where precipitation is greater than 2400 mm yr21
(PGT2400), were used in this study. The segment with less than 360 mm yr21 (PLT360)
identifies areas where any green cropland vegetation has a very high likelihood of being
irrigated, since average evaporation rates of 30 mm month21 will be considerably less
than evaporative demand. This segment will help focus on identifying irrigated and nonirrigated classes in the arid and semiarid areas and deserts. By contrast, the segment with
precipitation over 2400 mm yr21 (PGT2400) mainly identifies the rain forest areas of the
world, although there are considerable areas of irrigation in this segment within the
southeast Asian lands, identified based on protocols discussed later.
2.3.2 Temperature segment. The 20 year mean AVHRR thermal band 4 data were
used to segment the world for areas less than 280 K. Where the mean temperature is
below 280 K on average (TLT280), it is too cold for agriculture, and irrigation is not
likely to be found there. However, some northern hemisphere areas have low average
temperatures but short summer seasons (May–October) in which supplemental irrigation
is actually practiced. Thereby, even in this zone classes are created and identified.
2.3.3 Forest cover data. A forest cover of greater than 75% (FGT75) was used as
one of the segments. Areas of very high forest cover imply that these areas are
unavailable for cultivation and the likelihood of irrigation is even less. Nevertheless,
the MFDC of FGT75 is classified and class identification and labelling process
followed. Forest cover was derived from the 1992 AVHRR 1 km data by the
University of Maryland (DeFries and Townshend 1994, DeFries et al. 2000a,b). If
forest density is greater than 75%, it is also rare that there will be any irrigation, due
to high rainfall and limited infrastructure. There is likely to be slash-and-burn
3688
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
agriculture in small fragments. This mask is complemented by a rain-forest mask
derived from the JERS-1 SAR (FSAR) imagery, in order to better identify other
land use fragments at higher resolution within the rain forest areas, including where
there might be irrigation. The rainforest mask implies that the areas of rainforests
are segmented and analysed separately and includes the use of JERS SAR data apart
from all other data used globally. The JERS SAR data was not used in areas outside
the rainforests.
2.3.4 GTOPO30 1 km DEM. The GTOPO30 is a 1 km global elevation data
derived from eight sources (USGS 1993, Verdin and Greenlee 1996, Verdin and
Jenson 1996, Tucker et al. 2005) and were used to segment the world for elevations
higher than 1500 m (EGT1500). There is a lower likelihood of irrigation above an
elevation of 1500 m, although there are certainly hill irrigation systems in the Andes,
Himalayas and the Philippines at higher elevations. The classes of the EGT1500 are
likely to be dominated by forests as likely land cover, but should be separable from
irrigation and agriculture due to their continuous vegetation signature using the
protocols described below.
Finally, the segment ‘AOAW’ focuses on where there are few biophysical
constraints to irrigation and shows where we are most likely to find it in various
forms. Overall, the segments help us to focus interpretation; but the presence or
absence of irrigation is investigated in detail in every segment. Even in segments with
very low likelihood of irrigation, detailed investigations were carried out to track any
remnants of irrigation.
3.
Overview of methods
An overview summary of the methods and analytical techniques are shown in figure 2.
The basic process begins with segmenting the MFDC (figure 1 and table 1) into
characteristic temperature, elevation and precipitation regions that makes it easier for
analysis, generating class spectra through classification by classifying the 159 layer
MFDC for each of the seven segments, grouping class spectra based on class
similarities and/or by comparing them with target spectra, rigorous protocols for class
identification and labelling that include use of large volumes of groundtruth data and
the use of VHRI, resolving mixed classes through specifying decision trees and spatial
modelling, standardized class naming and class name verification and establishing
innovative methods for irrigated area calculations and accuracy assessments. These
processes are described in the following sections and presented in further detail in a
research report (Thenkabail et al. 2006, http://www.iwmigiam.org).
The MFDC retains the integrity of each data layer and unlike data fusion does not
merge data. In contrast to data fusion, the MFDC retains a series of data layers, akin
to hyperspectral data layers, each with its own characteristics but resampled to 1 km.
The various data layers are geographically precise.
3.1
Class spectra generation through unsupervised classification
The mega-files of each of the seven segments are processed using unsupervised
ISOCLASS k-means classification (Tou and Gonzalez 1975, Leica 2005) to produce
a large number of class spectra. In each segment, we began with 250 classes as a start.
In some smaller segments or more homogeneous segments (e.g. segment TLT280),
the maximum number of classes produced by the k-means algorithm was less than
250 classes, even when we specified 250 classes.
3689
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
Figure 2. Methodology for mapping global irrigated areas (GIAM). The flow-charts provide
an overview of the GIAM methodology.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3690
(Continued.)
P. S. Thenkabail et al.
Figure 2.
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3.2
3691
Class grouping through Spectral Matching Techniques (SMTs)
In more localized applications, it is common to undertake groundtruth to identify
and label the classes generated using the ISOCLASS algorithm. However, at the
global scale, this is not possible due to enormous resources required to cover vast
areas to identify and label classes. So as a first step, SMTs (Farrand and Harsanyi
1997, Bing et al. 1998, Granahan and Sweet 2001, Schwarz and Staenz 2001,
Shippert 2001, Homayouni and Roux 2003, Thenkabail et al. 2007a) were used to
group classes. Time-series of NDVI (e.g. sample illustration in figure 3) or other
metrics are analogous to spectra, where time is substituted for wavelength.
The principle in spectral matching is to match the shape, or the magnitude or
(preferably) both to an ideal or target spectrum (commonly known as a pure class or
‘end-member’) (Thenkabail et al. 2007a). In cases where the class does not have
matching ideal spectra, the class identities are investigated through the methods
described below in order to label them.
Two quantitative SMTs, to group classes, adopted in this study were (Thenkabail
et al. 2007a):
(a) Spectral Correlation Similarity (SCS), which is the shape measure and
(b) Spectral Similarity Value (SSV), which is the shape and magnitude measure.
The range of SCS R2 values (where R2 is the coefficient of determination) lies
between 21 and + 1, but negative values have no meaning in this application. The
higher the positive value, the greater the similarity. The normal range of SSV is from
0 to 1.415. The smaller the SSV value, the greater the similarity of classes. The
process of grouping classes based on SCS R2 values is illustrated in figure 3.
Figure 3(a) shows how the classes group based on similar SCS R2 values. Figure 3(b)
shows the results of grouping similar spectra for double crop irrigation, continuous
forest cover and bare or fallow soils. The SMTs perform two key functions: (a) first,
they group data of similar classes (e.g. figure 3) and (b) second, they help identify
classes by matching the class spectra, with ideal or target spectral data bank, which is
generated based on precise groundtruth knowledge.
In this paper, we use SMTs extensively to: (i) group a large number of classes to a
few distinct groups of classes (e.g. figure 3) and (ii) group and label classes by
comparing the group of similar classes of class spectra with ideal/target spectra from
the precise groundtruth locations. The reader can refer to Thenkabail et al. (2007a)
for a detailed discussion on the application of SMTs in irrigated area mapping.
3.3
Class identification and labelling
A comprehensive set of protocols for identifying and labelling the classes was
adopted (figure 2(b)). Once the classes are grouped by SMTs, each class in a group is
investigated by using multiple data sets and the procedures described below, which
lead to labelling a class or group of classes.
3.4
Groundtruth data application
Precise knowledge of the real situation on the ground is essential to interpret all
remote sensing products for the purposes of training, class identification, naming
and accuracy assessment. The GIAM project relied on two large groundtruth data
sets. These are made available through IWMI data storehouse pathway or
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3692
Figure 3. Spectral Matching Techniques (SMTs) to group classes. The spectral characteristics (e.g. NDVI or spectral reflectivity over time) of any given class is compared with other
classes and/or with ideal spectra: (a) quantitatively or (b) qualitatively.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
3693
Figure 4. Groundtruth (GT) data for class identification and labelling. Nearly 6 000
groundtruth points were used in the class identification and labelling process. These included
4000 + points from the degree confluence project and nearly 2000 points from the groundtruth
missions of the GIAM project.
IWMIDSP (http://www.iwmidsp.org) in standard geographic information system
(GIS) formats and are now briefly described.
3.4.1 Public domain groundtruth from the Degree Confluence Project (DCP). The
DCP (http://www.confluence.org/) is an organized sampling of the entire world at
every 1u latitude and 1u longitude intersection. This is a perfect stratified random
sampling, stratified by latitude and longitude grids. It is a voluntary effort. In all, we
used 3864 confluence points based on the availability during the project period. The
data consisted of precise latitude, longitude, a digital photo of land cover and a
description of the land use/land cover (LULC). These were converted to proprietary
GIS formats (figure 4) and used in the GIAM class identification and labelling, as
well as accuracy assessment along with groundtruth data collected during this
project and various VHRI from Google Earth. One in four points (966 points) was
used for accuracy assessment. While performing accuracy, the 966 DCP points were
added to 1005 points of the GIAM groundtruth data for a total of 1971 points.
3.4.2 Groundtruth data collected by the GIAM team members. Detailed ground
truth data were collected by IWMI specifically for the GIAM project similar to
procedures and approaches described by Thenkabail et al. (2005, 2007a) and Biggs et al.
(2006). The precise locations of the samples were recorded by GPS in the Universal
Transverse Mercator (UTM) and the latitude/longitude coordinate system with a
common datum of WGS84. At each location, land use, land cover, crop dominance,
crop types, crop growth stages, irrigation source and irrigation intensity were recorded
(e.g. figure 4). The statistical design was based on stratified random sampling. They
were stratified by the road network, and randomized by the distance from road
intersections or time. For example, the sample site locations were selected at 5 km from
a road intersection or 5 minute travel from a road intersection. As far as possible, minor
road were used. In addition, sampling was carried out using the diversions in these
minor roads and travelling a set distance or time from an intersection. In all, 1790
3694
P. S. Thenkabail et al.
groundtruth points were sourced from India, Sri Lanka, Syria, west and central Africa,
South Africa and central Asia. One in two points (895 points) were used in class
identification and labelling, and the rest of the 895 points were used in accuracy
assessment. An additional 110 points from our very recent missions to the USA, China
and Uzbekistan were added to accuracy assessment, making the total points 1005.
3.5
Utilizing the Google Earth data set for labelling GIAM classes
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Google Earth (http://earth.google.com/) contains increasingly comprehensive image
coverage of the globe at very high resolution, 0.61–4m, allowing the user to zoom-in
to specific areas in great detail, from a base of 30 m resolution data, based on
Geocover 2000. In GIAM, Google Earth data were used for:
N
N
N
identification and labelling the GIAM classes;
deriving irrigated area fractions (IAFs) that helped in sub-pixel area (SPA)
calculations; and
assessing accuracy of irrigated area classes.
In all, nearly 11 000 + Google Earth locations were used during the class
identification and labelling process (e.g. figure 5). The process starts with zooming in
to a precise location and investigating the areas in and around the location. Often, a
few thousand hectares are viewed at sub-metre to 4 m by zooming in and around the
location, leading to a class name. In order to identify a class, a minimum of 30–60
Figure 5. Google Earth data for class identification and labelling. Over 11 000 Google Earth
data very-high-resolution imagery (VHRI) ‘zoom-in points’ were used in class identification
and labelling.
Global irrigated area map
3695
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
spatially well spread out Google Earth data locations were used. The same process is
repeated for identifying another class. At the end of the project, in order to identify
thousands of classes, 11 000 + Google Earth locations had accumulated. The veryhigh-resolution data had some real advantage over groundtruth in that they
provided information on a much larger area and are, therefore, more representative
of the area than is normally sampled directly on the ground. The interpretation of a
class is based on visual indicators such as shape (e.g. central pivot circles), size (e.g.
large- and small-scale reservoir size), pattern (e.g. contiguous farms) and texture (e.g.
the smooth texture of a farm compared to the rough texture of a forest). The date of
the 0.61–4 m imagery varies from place to place in Google Earth. Therefore, we may
see irrigated area as cropped or fallow depending on the season. Further, Google
Earth does not have a wall-to-wall coverage of the 0.61–4 m imagery of the world. In
contrast, Geocover has a wall-to-wall coverage at 30 m.
3.6 Using the Environmental System Research Institute (ESRI) Landsat 150 m
Geocover in classification
The ESRI resampled the 8500 ortho-rectified Landsat ETM + 30 m ‘Geocover’ tiles
(University of Maryland, http://glcf.umiacs.umd.edu/index.shtml, Tucker et al.
2005), and made them available as a single mosaic of the world. These data are
used to provide contextual information and pseudo ‘groundtruth’ by geo-linking to
the class maps in order to identify and label classes. The resampled ‘Geocover’
images have a pixel resolution of 150 m compared with the original pan-sharpened
size of 15 m, but provide rapid assessment for checking a class for any part of the
world and are positionally the most accurate image set covering the entire globe. The
images are optimized to provide maximum greenness for the nominal year 2000,
offer a detailed ‘zoom-in’ view of any part of the world and are ideal for geo-linking
to identify and label a class.
The ‘zoom-in views’ of high-resolution imagery of Geocover 150 m and Google
Earth help in class identification in many ways. First, no irrigated classes, such as
forests, deserts, water and rangelands, are quickly separated from agricultural lands.
Second, irrigation sources, such as central pivot systems and canals, are easily
detected (e.g. figure 5). Third, a large number of water bodies in the area implies, but
not necessarily confirms, irrigation. In such cases, we will use other data such as
groundtruth and knowledge bases of data gathered from the national system (CBIP
1994) to supplement/complement inferences drawn from higher resolution imagery.
True cover type is determined based on majority view within the GIAM team on
what the class could be in higher resolution imagery similar to interpretative
techniques described for the International Geosphere Biosphere Programme (IGBP)
DISCover data (see Belward et al. 1999, Loveland et al. 2000).
3.7 Techniques for class identification: Space–time spiral curves (ST-SCs) and
brightness–greenness–wetness (BGW) plots
A two-dimensional (2D) near-infrared versus red band spectral reflectivity plot of
unsupervised classes is referred to as a BGW plot (Kauth and Thomas 1976,
Thenkabail et al. 2005). The BGW plots help determine whether a class is: (a) green,
(b) bright, (c) wet or (d) somewhere in between these classes. Classes that occupy the
green area have high near-infrared reflectivity and low red reflectivity. Typically,
these areas are forests, agricultural lands and natural vegetation. Classes that occupy
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3696
P. S. Thenkabail et al.
Figure 6. Space–time spiral curves (ST-SCs) in class identification and labelling. The ST-SCs
track changes of time series over time and across space. The numbers seen in each class
represent Julian date and each class moves around a ‘territory’ in 2D feature space over time.
bright areas have high near-infrared and high red reflectivity. The LULC categories
of these classes are likely to be open/barren areas, sparse vegetation, dry vegetation,
clouds and built-up areas. Classes that occupy wet areas have low near-infrared and
low red reflectivity. These classes are likely to be wetlands, moist lands, water bodies,
cloud shadows and swamp forests. The classes that are in between have different
combinations of these broad LULC classes. The BGW plots provide clear and useful
information on class dynamics over time and are a very helpful tool in identifying
and labelling a class.
The 2D ST-SCs (e.g. figure 6) provide very useful information on class behaviour.
Each class has its own ‘territory’ and moves around in its territory year after year.
For example, irrigated areas, forests and rain-fed areas have the largest territories
(figure 6). In contrast, barren lands, wetlands, scattered vegetation and grasslands
have smaller territories. This approach is used to match and group classes that: (a)
fall within similar 2D feature space of a ST-SC plot and (b) have characteristic
territory that leads to more precise interpretation of the nature of the class (based on
sound field knowledge of at least one or more classes in a group). In figure 6,
irrigated areas have the largest ‘territory’. It is important to know the spatial
distribution of the class and groundtruth knowledge to be definitive of the class
name. However, the 2D ST-SCs provide very good indications of the classes based
on where they occur and their ‘territorial’ characteristics.
Global irrigated area map
3697
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3.8 Employing NDVI time series and brightness temperature in identification of
categories
The NDVI time series can categorize and identify irrigated area classes into
categories such as double crop (e.g. class 31 in figure 3(b)), continuous crop and
single crop. For example, time series NDVI are plotted, compared and contrasted,
resulting in distinct categories. This is illustrated for five distinct classes in figure 3(b):
(a) forests, (b) barren/desert lands, (c) savannah croplands, (d) irrigated mix and (d)
irrigated double crop.
During the class identification process, the AVHRR time series earth ‘skin’
temperatures (plots not illustrated) were also plotted along with time series NDVI
(figure 3). In the tropics, the greater the biomass levels of a crop, the lower the skin
temperature and vice versa. The skin temperatures of irrigated crops are low due to
crop transpiration and background moisture/wetness. In the northern hemisphere,
crops grown in May–October (summer months) exhibit high NDVI and high skin
temperatures. In contrast, during November–April (winter) snow and leaf-off
conditions, there are low NDVI and low skin temperatures. Thus, the skin temperature
time series helps identify LULC classes in different climatic zones of the world and is
often complementary and/or supplementary to NDVI time series plots (see a detailed
discussion on skin temperature to LULC classes in Thenkabail et al. 2007a).
3.9
Resolving mixed classes and class verification
In spite of the rigorous class identification process described above, there are often
‘mixed’ classes. Typically, the unresolved classes were split up into 10 to 50 subclasses (depending on extent of area and complexity) before applying the decision
tree, GIS spatial modelling and contextual groundtruthing process. Decision tree
algorithms (DeFries et al. 1998) involving factors such as NDVI, band reflectivity
and thermal temperatures in resolving the mixed classes based on the rule base, are
followed by class identification and labelling process discussed above. When classes
continue to be mixed, in spite of the various methods and techniques discussed in
previous sections, we adopted the GIS spatial modelling approaches to resolve
classes. This involved taking a mixed class and applying GIS spatial modelling
techniques, such as overlay, matrix, recode and sieve and proximity analysis (Leica
2007), based on the theory of map algebra and Boolean logic (Peuquet and Marble
1990, Tomlin 1990, Tomlinson 2003). The GIS spatial data layers used include
precipitation zone, elevation zones, Koppen ecological zone, temperature zone and
tree cover categories (see figure 2(b)). Any one, or a combination of these data layers,
often helped separate the mixed classes. Other global LULC products, such as the
USGS LULC (Loveland et al. 2000, Agrawal et al. 2004), USGS seasonal LULC
(Loveland et al. 2000), GLC2000 (Bartholome and Belward 2005), IGBP (IGBP
1990) and Olson eco-regions of the world (Olson 1994a,b) were also used in verifying
final class names assigned in GIAM as a cross check and were specifically useful for
verifying no irrigated classes.
3.10
Class naming convention and the generic map
The classification of the various segments, identifying and labelling these classes,
reclassification of numerous mixed classes and identifying these reclassified classes lead
to thousands of classes. Synthesizing these classes becomes extremely complex, unless a
standardized system is adopted. In order to make the process seamless and logical for
3698
P. S. Thenkabail et al.
analysts working in the project, we adopted a standard class-naming convention
(figure 7), which was supported by groundtruth data, Google Earth data, the SMT and
other techniques. This ensured a consistent class-naming pattern irrespective of the
analyst. Therefore, every analyst named the classes in a set pattern (see figure 7): watering
method, type of irrigation, crop type, scale, intensity, location and type of signature. The
irrigation intensity is determined directly using the class signature (e.g. figure 3).
The classes can then be grouped and aggregated as follows:
1.1 Major and medium
irrigated areas
1.11 Surface water
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
2.1 Minor irrigated areas 2.11 Groundwater
1.111 reservoirs with .2000 ha
water spread area
2.111 groundwater
2.112 small reservoirs
(,2000 ha water spread area)
2.113 Tanks
2.21 Conjunctive use
(surface + groundwater)
2.31 Supplemental
(predominantly rain-fed
with significant irrigation)
Note. Irrigation by drip, sprinkler, etc. can be from surface water and/or groundwater.
3.11
Irrigated area estimation through SPA calculations
The precise irrigated areas are calculated as:
SPAn ~ðFPAn Þ|ðIAFn Þ,
ð1Þ
where SPAn is the SPA of class n; FPAn is the full-pixel area (FPA) of class n; IAFn is
the irrigated area fraction (IAF) for class n. The FPA is calculated for the 28 GIAM
Figure 7. Class labelling protocol. The class labelling protocol ‘forces’ an analyst to label a
class exactly in a standardized pattern so that every analyst labels the class exactly the same
way. Shown here are different levels of class labelling.
Global irrigated area map
3699
classes based on Lambert azimuthal equal area (LAEA) projection. The IAFs are
determined using the three methods (Thenkabail et al. 2007b): (a) Google Earth
Estimate (IAF–GEE); (b) high-resolution imagery (IAF–HRI); and (c) sub-pixel
decomposition technique (IAF–SPDT).
The reader is referred to Thenkabail et al. (2007b) for a detailed understanding of
the SPA calculation methods and the IAFs used in this paper.
3.12
Definition of irrigated areas
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Then two types of irrigated areas were defined and calculated:
(1) TAAI. The TAAI does not consider seasonality or intensity of irrigation. This
is the area actually irrigated at any given point of time plus area ‘equipped for
irrigation’, but left fallow at the same point of time. The TAAI is similar to
FAO/UF’s ‘equipped’ area and ‘net irrigated areas (NIA)’ in national
statistics. The TAAI is determined by multiplying the FPA by the IAF of the
TAAI (IAFTAAI). The IAFTAAI were taken as average of the IAF–GEE and
IAF–HRI (of the June–October season) (see IAFTAAI in table 2).
(2) AIA. The AIA considers the seasonality or intensity of irrigation. This is the
sum of the areas irrigated during season one and season two, and those that
were irrigated continuously throughout the year. The equivalent of AIA in the
national statistics is ‘gross irrigated area (GIA)’. The AIA is determined by
multiplying the FPA by the IAF of the AIA for each season, as well as year
round, and then summing up these areas. The IAFAIA were taken as an average
of the IAF–HRI and IAF–SPDT of the particular seasons (sea seasonal IAFs
in table 2). The seasonality is derived from the AVHRR NDVI profiles of the
classes. Each class is observed for single crop (single NDVI peak), double crop
(double NDVI peak) and continuous perennial crops, such as plantations
(NDVI threshold of 0.4 or more throughout the year). For further details of the
cropping calendars, seasons and NDVI profiles, refer to Thenkabail et al.
(2006) and the GIAM web portal (http://www.iwmigiam.org).
The IAFs of each class for TAAI and AIA are provided in table 2. The TAAI and
AIAs are computed by multiplying the respective IAFs with the FPAs.
3.13
Accuracy assessment
Accuracies were determined based on two independent data sets. These were:
(1) Accuracy based on groundtruth data from the GIAM project and the DCP.
Altogether, 1971 points (966 DCP points plus 1005 GIAM project collected
points) were pooled to determine accuracy. Of the 1971 points, 1005 points
were irrigated: 463 by surface water and 542 by groundwater. The rest were
other LULC and were not used in the final accuracy assessment.
(2) Accuracy based on Google Earth VHRI (sub-metre to 4 m). The groundtruth
from Google Earth (GT–GE) generated 670 points that were randomly
distributed around the world, with a higher density of distribution of points
where the irrigated area is denser. Of the 670 sample points, 323 were
irrigated (220 by surface water and 103 by groundwater). The surface-water
irrigation was fairly easy to detect with canals, reservoirs and tanks. When
there is no evidence of the existence of the surface water, but if the area is
Class Names
Full
Pixel area
(FPA)
hectares
1
2
3
4
5
6
7
10,639,378
Irrigated, surface
water, single crop,
wheat-corn-cotton
Irrigated, surface
6,896,128
water, single crop,
cotton-rice-wheat
14,135,930
Irrigated, surface
water, single crop,
mixed-crops
69,830,220
Irrigated, surface
water, double crop,
rice-wheat-cotton
72,501,012
Irrigated, surface
water, double crop,
rice-wheat-cotton-corn
51,769,022
Irrigated, surface
water, double crop,
rice-wheat-plantations
2,569,367
Irrigated, surface
water, double crop,
sugarcane
Season
1 area
IAFSeason 2
Season
2 area
IAFcontinuous
season
unit less
hectares
unit less
Annualized
irrigated
areas
continuous
(AIAs) or
season
gross areas
area
unit less
hectares
unit less
hectares
0.73
7,766,444
0.61
6,471,843
6,471,843
hectares
0.85
5,880,717
0.55
3,813,841
3,813,841
0.68
9,628,687
0.46
6,511,261
6,511,261
0.71
49,710,095
0.53
36,711,650
0.67
46,745,513
83,457,163
0.63
45,369,799
0.56
40,938,905
0.52
37,483,023
78,421,928
0.72
37,389,472
0.58
29,807,112
0.48
24,769,631
54,576,742
0.74
1,910,007
0.67
1,716,980
0.53
1,372,877
3,089,857
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Class
Number
Total area
Irrigated available
for
area
IAFfraction irrigation
or net area season1
(IAF)
3700
Table 2. Global irrigated areas. The irrigated areas of each of the 28 classes for the world determined by multiplying the full pixel area (FPA) with irrigated
area fractions (IAFs) leading to sub-pixel areas (SPAs). The SPAs are actual areas. The annualized irrigated areas (AIAs) is sum of the irrigated areas during
season one, season two, and continuous year round. The total area available for irrigation (TAAI) is the irrigated area during any given point of time in the
season (taken for the main crop growing season in the world which is during June-October) plus the areas equipped for irrigation but left fallow at the same
point of time.
Class
Number
8
9
10
11
12
13
14
15
Class Names
Full
Pixel area
(FPA)
60,312,587
Irrigated, surface
water, double crop,
mixed-crops
116,418
Irrigated, surface
water, continuous
crop, sugarcane
13,427,918
Irrigated, surface
water, continuous
crop, plantations
12,780,583
Irrigated, ground
water, single crop,
rice-sugarcane
5,997,678
Irrigated, ground
water, single crop,
corn-soybean
1,570,188
Irrigated, ground
water, single crop, rice
and other crops
11,799,752
Irrigated, ground
water, single crop,
mixed-crops
3,554,656
Irrigated, ground
water, double crop,
rice and other crops
Total area
Irrigated available
for
area
IAFfraction irrigation
or net area season1
(IAF)
0.37
Season
1 area
IAFSeason 2
Season
2 area
22,446,718
0.37
22,213,443
IAFcontinuous
season
Annualized
irrigated
areas
continuous
(AIAs) or
season
gross areas
area
0.64
38,779,483
0.49
56,932
0.42
49,302
49,302
0.61
8,184,907
0.44
5,865,373
5,865,373
0.52
6,653,732
0.49
6,255,930
6,255,930
0.70
4,181,556
0.49
2,916,140
2,916,140
0.68
1,063,691
0.15
241,540
241,540
0.47
5,590,581
0.38
4,518,047
4,518,047
0.73
2,583,423
0.55
1,949,455
0.51
1,800,169
44,660,161
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Table 2. (Continued.)
3,749,623
3701
Class
Number
16
17
18
19
20
21
22
23
Class Names
Irrigated, conjunctive
use, single crop,
wheat-corn-soybeanrice
Irrigated, conjunctive
use, single crop,
wheat-corn-orchardsrice
Irrigated, conjunctive
use, single crop, cornsoybeans-other crops
Irrigated, conjunctive
use, single crop,
pastures
Irrigated, conjunctive
use, single crop,
pasture, wheat,
sugarcane
Irrigated, conjunctive
use, single crop,
mixed-crops
Irrigated, conjunctive
use, double crop,
rice-wheat-sugarcane
Irrigated, conjunctive
use, double crop,
sugarcane-other crops
Full
Pixel area
(FPA)
Total area
Irrigated available
for
area
IAFfraction irrigation
or net area season1
(IAF)
Season
1 area
IAFSeason 2
Season
2 area
IAFcontinuous
season
Annualized
irrigated
areas
continuous
(AIAs) or
season
gross areas
area
29,919,283
0.84
25,082,625
0.47
13,994,126
13,994,126
10,479,639
0.68
7,135,193
0.57
5,982,487
5,982,487
17,658,270
0.73
12,810,184
0.51
9,039,700
9,039,700
9,150,534
0.62
5,672,425
0.25
2,287,634
2,287,634
2,521,549
0.77
1,942,683
0.46
1,162,908
1,162,908
17,131,259
0.77
13,120,827
0.57
9,836,226
9,836,226
71,510,203
0.67
48,004,873
0.49
35,361,814
0.43
30,967,596
66,329,410
1,838,672
0.69
1,265,539
0.39
720,494
0.50
916,272
1,636,766
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3702
Table 2. (Continued.)
Class
Number
24
25
26
27
28
Class Names
Irrigated, conjunctive
use, double crop,
mixed-crops
Irrigated, conjunctive
use, continuous crop,
rice-wheat
Irrigated, conjunctive
use, continuous crop,
rice-wheat-corn
Irrigated, conjunctive
use, continuous crop,
sugarcane-orchardsrice
Irrigated, conjunctive
use, continuous crop,
mixed-crops
Full
Pixel area
(FPA)
Total area
Irrigated available
for
area
IAFfraction irrigation
or net area season1
(IAF)
0.48
Season
1 area
IAFSeason 2
Season
2 area
12,463,458
0.34
8,700,640
IAFcontinuous
season
Annualized
irrigated
areas
continuous
(AIAs) or
season
gross areas
area
25,756,897
0.51
13,057,718
21,164,097
13,969,654
0.51
7,186,641
0.47
6,618,040
6,618,040
15,427,976
0.69
10,573,933
0.50
7,672,155
7,672,155
13,018,909
0.76
9,912,989
0.55
7,168,857
7,168,857
22,304,422
0.81
18,011,795
0.56
12,393,114
12,393,114
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Table 2. (Continued.)
3703
3704
P. S. Thenkabail et al.
cultivated, then this area becomes a candidate point for rain-fed or
groundwater irrigated. This is when the evapo-transpiration (ET) 16 km grid
data from the World Water and Climate Atlas (http://www.iwmi.cgiar.org/
WAtlas/atlas.htm) was used to determine whether the ET far exceeds
precipitation (40 year CRU precipitation data available in http://www.
iwmidsp.org). If the answer to this is yes, then irrigation should exist for crops
to grow. If the answer to this is no, then the area ought to be rain-fed.
The accuracies were performed to determine how well the irrigated area was
mapped. Point-based accuracy and error estimates (Congalton 1994, Foody 2002)
were established based on:
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
accuracy of irrigated area class
~
groundtruthed irrigated points classified as irrigated area
|100,
total number of groundtruthed points for irrigated area class
ð2Þ
errors of commission for irrigated area classes
~
non-irrigated groundtruth points falling on irrigated area class
|100,
total number of non-irrigated groundtruthed points
ð3Þ
and
errors of omission for irrigated area class
~
4.
irrigated groundtruth points falling on non-irrigated area class
|100:
total number of irrigated area groundtruth points
ð4Þ
Results
The comprehensive methodology leads to identification and labelling of a final set of
classes from each of the seven segments. By aggregating similar types of irrigated
area classes, a 28 class global irrigated area map (GIAM28) was produced
(figure 8(a)), and the area statistics computed for the classes (table 2), continents
(table 3) and countries (table 4).
In GIAM28, classes 1–10 are surface water (major and medium irrigation from
surface water based on large and medium dams); classes 11–15 are groundwater
(minor irrigation from groundwater, small reservoirs and tanks); and classes 16–28
are conjunctive use (predominately minor irrigation from groundwater, small
reservoirs and tanks, but with some mix of surface water irrigation from major
reservoirs). Within each irrigation type (surface water, groundwater and
conjunctive use), there are classes for single, double and continuous cropping
(figure 8(a) and table 2). Dominant crop types have also been labelled. The GIAM
demonstrates the spatial distribution of irrigated areas in the world, and clearly
establishes its overwhelming concentration in a few countries such as China, India,
the USA and Pakistan (figure 8(a)). The presence of a large number of classes in
GIAM 28 classes ensures varying seasonality of classes by taking more precise
cropping calendars between northern and southern hemispheres, the tropics and
the higher latitudes.
3705
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
Figure 8. Global Irrigated Area Maps (GIAMs). (a) The aggregated 28 class GIAM
product and (b) the irrigated croplands of the world along with the rain-fed croplands of the
world.
The distribution of the irrigated areas (figure 8(a)) are overlaid on rain-fed
cropland areas of a parallel study (Biradar et al. 2009, http://www.iwmigiam.org)
using the same methodology. This shows the spatial distribution of irrigated
croplands relative to rain-fed croplands of the world (figure 8(b)). China and India
with about 2.4 billion people depend on irrigation and often have double cropping to
feed their populations. In contrast, North America and Europe, with a combined
population of about 1.3 billion, depend on rain-fed agriculture (figure 8(b)), with
only one crop per year. They also export large quantities of their food grains to other
countries and continents. Throughout this paper, we discuss the global irrigated
areas (figure 8(a) and table 2). The global rain-fed croplands (Biradar et al. 2009)
distribution overlaid on the global irrigated areas (figure 8(b)) is presented briefly for
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
1
Continent
1
2
3
4
5
6
7
Africa
Asia
Australia
Europe
North America
South America
Oceania
TOTAL
SPATAAI
(TAAI) (ha)
Season 1
(ha)
Season 2
(ha)
Continuous
(ha)
Annualized
sum (ha)
8 687 044
290 641 673
11 865 244
33 937 745
35 426 895
17 842 959
125 390
398 526 951
5 601 273
192 600 664
2 991 344
20 126 797
22 316 537
8 055 356
68 146
251 760 118
3 680 659
152 312 096
0
7 691 113
6 448 147
3 363 798
58 034
173 553 847
1 020 078
24 700 163
2 382 064
4 627 243
3 089 990
5 608 671
15 505
41 443 716
10 302 011
369 612 923
5 373 409
32 445 154
31 854 673
17 027 825
141 686
466 757 680
Percentage FAO/UF V4.03
of world (Area equipped for
Irrigation) (ha)
total (%)
2
79
1
7
7
4
0
100
13 432 285
187 600 089
2 056 580
26 770 001
36 889 071
11 495 806
581 254
278 825 086
Note. (1) SPA from combined coefficients of Google Earth estimate and high-resolution images, (2) SPA from combined coefficients of high-resolution
images and sub-pixel decomposition technique, (3) area irrigated obtained from FAO AQUASTAT and Earth trends (http://www.fao.org/ag/agl/aglw/
aquastat/water_use/croppat.htm/ http://earthtrends.wri.org/country_profiles/index.php?theme58).
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3706
Table 3. Continental irrigated areas based on GIAM. The distribution of irrigated areas and their percentages in different continents of the world.
1
Rank
Country
SPATAAI
(TAAI) (ha)
Season 1 (ha)
Season 2 (ha)
Continuous (ha)
Annualized (ha)
China
India
USA
Pakistan
Russia
Argentina
Thailand
Bangladesh
Kazakhstan
Myanmar(Burma)
Australia
Uzbekistan
Vietnam
Brazil
Mexico
Indonesia
Egypt
Spain
Germany
Canada
France
Italy
Iraq
Iran
Japan
Ukraine
Korea, Dem. Rep.
Romania
Turkmenistan
111 988 772
101 234 893
28 045 478
14 036 151
13 886 856
9 304 258
6 610 586
5 235 050
7 227 718
4 452 997
11 865 244
3 601 487
4 384 022
4 195 118
3 854 673
3 172 879
2 144 099
3 421 724
2 197 697
2 658 297
2 399 518
2 829 523
2 220 024
2 623 336
2 525 096
2 995 578
1 467 262
2 375 239
1 522 372
75 880 320
72 612 189
18 182 104
7 895 566
8 865 013
3 601 505
3 228 550
3 882 847
4 625 716
3 360 330
2 991 344
2 733 397
1 865 074
2 165 151
1 818 168
1 221 384
1 635 323
1 516 815
1 642 692
1 727 915
1 249 368
1 342 442
1 242 694
1 308 727
1 157 850
1 631 677
935 934
1 128 692
994 264
68 233 355
53 685 066
4 006 141
7 302 243
2 113 783
1 605 815
2 209 523
3 076 494
1 760 606
2 798 234
0
2 427 259
1 419 401
869 365
916 083
716 038
1 491 605
683 698
1 318 567
1 124 721
829 980
539 802
1 254 929
679 564
656 470
258 515
923 533
315 485
904 352
7 688 411
5 956 598
2 120 942
761 533
224 734
3 559 092
1 959 295
206 686
83 362
148 108
2 382 064
134 859
1 665 058
1 051 327
874 479
1 385 021
165 798
825 310
40 415
21 616
607 806
761 896
128 942
500 268
654 276
491 607
194 157
605 711
101 368
151 802 086
132 253 854
24 309 188
15 959 342
11 203 530
8 766 412
7 397 368
7 166 028
6 469 685
6 306 671
5 373 409
5 295 515
4 949 533
4 085 844
3 608 730
3 322 443
3 292 726
3 025 823
3 001 674
2 874 252
2 687 153
2 644 140
2 626 564
2 488 558
2 468 596
2 381 799
2 053 625
2 049 888
1 999 984
FAO/UF V4.03
Irrigated (area equipped for
area (%)
irrigation) (ha)
32.523
28.335
5.208
3.419
2.400
1.878
1.585
1.535
1.386
1.351
1.151
1.135
1.060
0.875
0.773
0.712
0.705
0.648
0.643
0.616
0.576
0.566
0.563
0.533
0.529
0.510
0.440
0.439
0.428
53 823 000
57 291 407
27 913 872
14 417 464
4 899 900
1 767 784
4 985 708
3 751 045
1 855 200
1 841 320
2 056 580
4 223 000
3 000 000
3 149 217
6 435 800
4 459 000
3 422 178
3 575 488
496 871
785 046
2 906 081
3 892 202
3 525 000
6 913 800
3 129 000
2 395 500
1 460 000
2 149 903
1 744 100
3707
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Table 4. Country-by-country irrigated areas based on GIAM. The distribution of irrigated areas and their percentages for the 198 countries in the world and
their comparison with the FAO AQUASTAT, which is based on the country statistics.
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
Country
Sudan
Philippines
Turkey
Nepal
Chile
Korea, Rep.
Morocco
United Kingdom
Bulgaria
Netherlands
Denmark
Cambodia
Afghanistan
South Africa
Azerbaijan
Sri Lanka
Venezuela
Kyrgyzstan
Greece
Czech Republic
Taiwan, Province of
China
Cuba
Syria
Colombia
Saudi Arabia
Belgium
Poland
Tajikistan
Somalia
Mongolia
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
Season 1 (ha)
Season 2 (ha)
Continuous (ha)
Annualized (ha)
FAO/UF V4.03
Irrigated (area equipped for
area (%)
irrigation) (ha)
1 737 118
1 542 629
1 753 382
1 251 988
1 514 922
1 192 469
1 045 119
970 733
1 301 804
870 243
1 164 705
736 318
1 008 138
821 040
835 627
948 029
894 880
700 876
907 739
518 036
499 043
1 185 252
1 024 930
882 867
681 267
703 120
546 413
578 582
810 688
579 629
681 847
976 705
480 153
403 083
574 487
441 335
169 255
499 284
447 852
271 632
380 186
282 608
643 655
589 003
332 404
530 989
345 867
432 289
460 512
233 603
62 782
299 991
2 835
329 683
218 706
206 929
218 092
111 161
93 686
247 134
106 151
321 296
314 359
101 685
175 175
362 042
265 047
396 243
335 053
114 723
15 913
369 652
29 502
0
128 606
301 701
47 075
162 553
529 164
214 109
75 288
388 895
245
80 910
1 930 592
1 789 108
1 577 313
1 477 303
1 445 230
1 313 755
1 153 817
1 060 204
1 012 064
1 011 340
979 539
938 441
923 490
828 491
821 980
809 579
807 078
770 274
766 678
701 727
677 877
0.414
0.383
0.338
0.317
0.310
0.281
0.247
0.227
0.217
0.217
0.210
0.201
0.198
0.177
0.176
0.173
0.173
0.165
0.164
0.150
0.145
1 863 000
1 550 000
4 185 910
1 168 349
1 900 000
880 365
1 458 160
228 950
545 160
476 315
476 000
284 172
3 199 070
1 498 000
1 453 318
570 000
570 219
1 075 040
1 544 530
50 590
525 528
486 898
566 990
546 186
678 677
324 796
351 514
383 243
372 476
422 332
342 202
302 293
336 538
143 187
294 221
268 183
277 736
162 324
265 966
269 666
235 219
176 558
89 073
204 916
185 150
156 376
117 817
110 413
25 291
58 751
79 399
318 806
8 293
779
15 040
123 434
0
637 159
596 263
592 495
551 066
507 430
454 111
449 153
403 574
376 378
0.137
0.128
0.127
0.118
0.109
0.097
0.096
0.086
0.081
870 319
1 266 900
900 000
1 730 767
35 170
134 050
719 200
200 000
57 300
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Rank
SPATAAI1
(TAAI) (ha)
3708
Table 4. (Continued.)
1
Rank
Peru
Uruguay
Guinea
Portugal
Senegal
Ecuador
Malaysia
Serbia
Moldova
Albania
Nigeria
Libya
Hungary
Bolivia
Ethiopia
Guinea Bissau
Georgia
New Zealand
Algeria
Macedonia
Armenia
Laos
Israel
Kenya
Guyana
Cote d’Ivoire
Tunisia
Austria
Swaziland
Guatemala
Dominican
Republic
SPATAAI
(TAAI) (ha)
355 956
381 403
302 633
358 865
211 416
288 581
258 766
171 939
294 070
223 777
197 909
230 656
241 714
214 091
184 239
108 042
128 538
125 390
144 349
169 843
106 695
105 585
99 806
85 401
96 276
95 138
109 144
116 456
149 274
69 373
70 876
Season 1 (ha)
189 766
311 863
153 448
133 115
148 318
127 918
123 739
140 266
161 373
117 469
103 154
67 173
166 069
28 854
62 157
84 650
96 950
68 146
90 667
113 105
73 185
78 350
39 883
53 025
61 736
79 392
30 355
69 017
97 004
47 776
45 462
Season 2 (ha)
113 945
25 602
95 459
54 464
129 202
85 157
66 638
92 171
20 311
55 223
61 884
60 076
14 990
9 777
25 604
66 770
46 285
58 034
34 731
9 610
37 092
21 795
37 020
37 354
30 935
20 756
23 663
19 025
0
40 864
25 851
Continuous (ha)
71 243
22 591
71 442
126 330
13 052
68 091
84 189
1 910
47 749
53 172
51 115
82 773
5 162
124 404
75 047
3 969
2 907
15 505
11 548
8 905
8 047
7 589
27 639
14 148
10 259
1 742
46 628
10 509
0
2 673
8 335
Annualized (ha)
374 954
360 055
320 350
313 908
290 572
281 166
274 565
234 348
229 433
225 864
216 154
210 022
186 221
163 036
162 808
155 389
146 141
141 686
136 946
131 620
118 324
107 734
104 542
104 527
102 930
101 890
100 647
98 551
97 004
91 313
79 648
FAO/UF V4.03
Irrigated (area equipped for
area (%)
irrigation) (ha)
0.080
0.077
0.069
0.067
0.062
0.060
0.059
0.050
0.049
0.048
0.046
0.045
0.040
0.035
0.035
0.033
0.031
0.030
0.029
0.028
0.025
0.023
0.022
0.022
0.022
0.022
0.022
0.021
0.021
0.020
0.017
1 729 069
217 593
94 914
792 008
119 680
863 370
362 600
163 311
307 000
340 000
293 117
470 000
292 147
128 240
289 530
22 558
300 000
577 882
569 418
127 800
286 027
295 535
183 408
103 203
150 134
72 750
394 063
97 480
49 860
129 803
269 710
3709
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
Country
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Table 4. (Continued.)
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
Country
Yemen
Honduras
Slovakia
Madagascar
Finland
Ghana
Sweden
United Arab
Emirates
Mali
Rwanda
Thegambia
Belarus
Mozambique
Haiti
Jordan
Cameroon
Tanzania
Panama
Croatia
Lithuania
Switzerland
Angola
Uganda
Oman
Sierra Leone
Chad
Qatar
Kuwait
Lebanon
Paraguay
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
Season 1 (ha)
Season 2 (ha)
Continuous (ha)
Annualized (ha)
FAO/UF V4.03
Irrigated (area equipped for
area (%)
irrigation) (ha)
91 688
70 584
109 904
72 359
125 307
60 647
83 918
93 810
42 912
51 034
71 826
41 627
71 961
28 411
69 968
10 249
16 073
21 071
1 044
19 039
0
24 173
1 140
4 867
20 203
5 623
2 618
14 490
0
19 181
0
55 487
79 188
77 729
75 488
75 156
71 961
71 764
71 108
70 603
0.017
0.017
0.016
0.016
0.015
0.015
0.015
0.015
388 000
73 210
15 643
1 086 291
103 800
30 900
188 470
280 341
56 355
80 067
39 872
84 088
56 415
50 848
72 717
52 694
47 022
49 069
35 202
57 272
29 523
23 316
30 017
17 853
21 807
25 234
38 509
37 333
24 747
28 582
38 220
64 806
34 993
60 731
39 402
29 974
574
35 415
33 678
21 997
28 102
41 591
21 079
16 671
26 957
15 247
16 343
15 932
0
0
11 240
12 913
26 100
0
28 422
195
16 753
15 438
568
5 861
7 852
6 477
15 511
0
15 897
14 371
3 447
14 898
12 481
8 020
0
0
8 170
1 670
1 559
0
0
0
4 587
8 490
51 399
10 852
5 467
16 574
1 018
0
0
3 116
183
0
213
3 747
27 596
26 753
5 859
10 445
65 879
64 806
63 415
60 926
60 742
53 903
52 541
52 128
46 998
45 048
44 630
41 591
36 976
34 158
30 586
30 145
29 037
27 698
27 596
26 753
25 268
25 029
0.014
0.014
0.014
0.013
0.013
0.012
0.011
0.011
0.010
0.010
0.010
0.009
0.008
0.007
0.007
0.006
0.006
0.006
0.006
0.006
0.005
0.005
235 791
8 500
0
115 000
118 120
91 502
76 912
25 654
184 330
34 626
5 790
4 416
40 000
80 000
9 150
72 630
29 360
30 273
12 520
6 968
117 113
67 000
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Rank
SPATAAI1
(TAAI) (ha)
3710
Table 4. (Continued.)
Rank
121
122
123
124
125
126
127
128
129
130
Togo
Nicaragua
Suriname
Congo, Dem. Rep.
Mauritania
Costa Rica
Benin
Burkina Faso
Estonia
Bosnia and
Herzegovina
Montenegro
Eritrea
Puerto Rico
El Salvador
Namibia
Burundi
Latvia
Gaza Strip
Cyprus
Jamaica
Niger
Botswana
East Timor
Mauritius
Lesotho
Zimbabwe
Belize
French Guyana
Malawi
Equatorial guinea
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
Season 1 (ha)
Season 2 (ha)
Continuous (ha)
Annualized (ha)
FAO/UF V4.03
Irrigated (area equipped for
area (%)
irrigation) (ha)
21 727
16 439
19 845
21 833
15 124
12 628
15 173
15 663
24 637
10 766
9 624
12 165
14 491
19 326
9 814
9 730
4 383
4 539
14 476
6 696
7 433
9 941
5 070
191
10 007
5 448
3 797
4 420
0
5 445
6 786
614
1 213
857
214
613
7 235
5 702
0
2 062
23 843
22 720
20 774
20 375
20 036
15 791
15 415
14 660
14 476
14 203
0.005
0.005
0.004
0.004
0.004
0.003
0.003
0.003
0.003
0.003
7 300
61 365
51 180
10 500
45 012
103 084
12 258
25 000
1 363
4 630
10 331
17 017
11 964
11 592
10 526
11 793
12 683
5 909
7 099
4 881
4 129
5 417
3 800
5 312
5 675
4 744
3 887
2 860
3 293
2 812
6 940
11 467
7 082
7 839
7 508
534
7 260
3 192
2 751
3 058
3 121
3 687
3 257
2 381
3 681
3 234
2 919
2 217
2 794
2 644
5 604
2 309
1 582
2 508
1 795
36
65
3 223
129
492
1 196
590
804
0
0
299
306
351
0
0
1 364
0
2 588
54
0
7 921
0
375
1 983
1 006
0
0
0
1 528
0
0
286
254
0
0
13 908
13 776
11 253
10 401
9 303
8 490
7 325
6 790
4 863
4 556
4 317
4 278
4 061
3 910
3 681
3 533
3 510
2 822
2 794
2 644
0.003
0.003
0.002
0.002
0.002
0.002
0.002
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0
21 590
37 079
44 993
7 573
21 430
1 150
0
55 813
25 214
73 663
1 439
14 000
21 222
2 638
173 513
3 000
2 000
56 390
0
3711
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
Country
SPATAAI1
(TAAI) (ha)
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Table 4. (Continued.)
Rank
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
Country
Antigua and
Barbuda
Guadeloupe
Trinidad and
Tobago
West Bank
Norway
St. Kitts and Nevis
Bhutan
Central African
Republic
Virgin Islands
Brunei
Reunion
San Marino
Djibouti
Zambia
Slovenia
Comoros
Anguilla
Liberia
Turks and Caicos
Islands
Montserrat
St. Pierre and
Miquelon
Cayman Islands
Monaco
Seychelles
Andorra
Bahrain
Barbados
SPATAAI
(TAAI) (ha)
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
Season 1 (ha)
Season 2 (ha)
Continuous (ha)
Annualized (ha)
FAO/UF V4.03
Irrigated (area equipped for
area (%)
irrigation) (ha)
2 270
1 378
706
384
2 468
0.001
130
1 894
1 859
1 498
1 672
342
0
183
48
2 022
1 720
0.000
0.000
2 000
3 600
1 612
2 072
1 650
997
1 155
538
1 323
1 314
796
1 086
533
130
84
600
0
471
0
48
0
0
1 542
1 453
1 445
1 396
1 086
0.000
0.000
0.000
0.000
0.000
0
134 396
18
38 734
135
827
799
651
1 102
905
779
439
241
489
237
214
563
481
517
0
587
0
293
218
404
201
117
361
369
329
0
0
0
217
199
0
100
0
91
152
0
797
0
536
0
0
0
0
53
1 015
1 002
846
797
587
536
510
417
404
300
170
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
185
1 000
13 000
0
1 012
155 912
0
130
0
2 100
0
69
70
51
59
65
0
0
0
115
59
0.000
0.000
0
0
66
73
66
0
0
0
55
0
44
0
0
0
0
0
0
0
0
0
0
53
0
0
0
0
55
53
44
0
0
0
0.000
0.000
0.000
0.000
0.000
0.000
0
0
260
150
4 060
1 000
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
1
3712
Table 4. (Continued.)
1
Rank
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
Country
Cape Verde
Congo
Fiji
Gabon
Gambia
Grenada
Guam
Ireland
Liechtenstein
Luxembourg
Malta
Martinique
Northern Marianna
Islands
Palestine
Papua New Guinea
Pitcairn Islands
Sao Tome and
Principe
Singapore
St. Lucia
St. Vincent and the
Grenadines
Vatican city
TOTAL
SPATAAI
(TAAI) (ha)
SPA–HRI/SPDT: IWMI GIAM 10 km V2.0 (actual irrigated area)2
Season 1 (ha)
Season 2 (ha)
Continuous (ha)
Annualized (ha)
FAO/UF V4.03
Irrigated (area equipped for
area (%)
irrigation) (ha)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
3 109
2 000
3 000
4 450
2 149
219
312
1 100
0
27
2 300
3 000
60
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.000
0.000
0.000
0.000
19 466
0
0
9 700
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.000
0.000
0.000
225 310
297
0
0
398 526 951
0
251 760 118
0
173 553 847
0
41 443 716
0
466 757 680
0.000
100.000
0
278 825 086
3713
Note. (1) SPA from combined coefficients of Google Earth estimate and high-resolution images, (2) SPA from combined coefficients of high-resolution
images and sub-pixel decomposition technique, (3) area equipped for irrigation from FAO and UF Global Map of Irrigated Area V3.0 (based on national
statistics), (4) area irrigated obtained from FAO AQUASTAT and Earth trends (http://faostat.fao.org/faostat/ http://earthtrends.wri.org/country_profiles/).
Global irrigated area map
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Table 4. (Continued.)
3714
P. S. Thenkabail et al.
discussion purposes and for showing the spatial distribution of the croplands. The
results of the rain-fed croplands or other LULC is not the focus of this paper, and
further presentation of results and discussions will be overwhelmingly focused on the
GIAM.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
4.1
Irrigated areas of the world
Irrigated areas of the world are calculated with and without considering the intensity
or seasonality. The TAAI does not consider intensity and provides area irrigated
plus area left fallow at any given point of time. This is equivalent to NIAs. The AIA
considers intensity or seasonality. The AIA is the sum of the area irrigated during
season one, season two and continuous year-round crops, such as sugarcane, or
permanent crops, such as plantations (table 2). The nearest equivalent of the AIA in
the national statistics is ‘GIA’.
The total AIAs of the world are 467 Mha, of which 252 Mha are during season
one, 174 Mha during season two and 41 Mha continuous (figure 8(a) and table 2).
The TAAI is 399 Mha. Of the AIA of 467 Mha, a high proportion of 55% (267 Mha)
is surface-water irrigation with double crop (table 2). Only about 5% of the AIA is
surface water, single crop and 1% surface water, continuous crop. The total surfacewater irrigation (classes 1–10 in table 2) is, therefore, 61%. The conjunctive use
classes (classes 16–28) are overwhelmingly groundwater-dominant with very minor
(roughly, less than 15%) surface-water influence. Therefore, if we group all the
groundwater classes from 11 to 15 and conjunctive use classes from 16 to 28 into one
category, the total groundwater irrigation in the world will be 39% of the AIA (or
186 Mha). Of the 39% groundwater irrigation, 19% were double crop, 12% single
crop and 8% continuous cropping. The distribution of irrigated areas in the
continents and countries was summarized in tables 3 and 4, respectively.
The FAO of the United Nations and the UF estimates the ‘equipped’ area for
irrigation (but not necessarily irrigated) in the world to be 279 Mha (http://www.fao.org/
ag/agl/aglw/aquastat/irrigationmap/index.stm; also reported in http://www.iwmigiam.
org, Siebert et al. 2005a,b, 2006). Based on the definition, the FAO/UF values (279 Mha)
should be compared with GIAM TAAI (399 Mha) (table 4). The reasons for these
differences are discussed in §4.6. The GIAM AIA is 467 Mha of which season one (June–
October) has 252 Mha, season two (November–February) has 174 Mha and continuous
year-round has 41 Mha. An overwhelming proportion of the global agriculture takes
place during season one. It is the main cropping season of all major irrigated area
countries, including China, India, the USA, Pakistan and most of the Asian and central
Asian countries. Together, these countries have over 85% of global irrigation.
4.2
Global irrigated area trends over the last two centuries
The development of global irrigated areas over the last two centuries (Framji et al.
1981, http://www.iwmigiam.org, http://www.fao.org/ag/agl/aglw/aquastat/irrigationmap/
index.stm) is summarized in figure 9. In the year 1800, there was a meagre 8 Mha
irrigated area (Framji et al. 1981). As the trends in figure 9 show, the increase in
irrigated areas for the next 140 years was modest, reaching a value of 95 Mha in the
early 1940s (Van Schilfgaarde 1994). Rapid increases in global irrigated areas took place
between 1940 and 2000. The FAO/UF global area ‘equipped for irrigation’ was around
279 Mha by the mid-1990s (see http://www.iwmigiam.org, http://www.fao.org/ag/agl/
aglw/aquastat/irrigationmap/index.stm, Döll and Siebert 2000, Siebert et al. 2005a,b,
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
3715
Figure 9. Global irrigated area trends. The global irrigated areas at the end of the last
millennium (this study) were provided as: (a) annualized irrigated areas (AIAs), which
consider cropping intensity or seasonality (sum of irrigated areas during season one + season
two + continuous year-round), and (b) total area available for irrigation (TAAI), which does
not consider intensity and is area irrigated at any given time plus the area equipped for
irrigation but remains fallow at the same point of time. In the national statistics, AIA is often
referred to as ‘gross irrigated area’ and TAAI as ‘net irrigated area’. The figure also shows the
trends of irrigation from year 1800 gathered from various sources.
2006). In this study (figures 9(a) and 10 and tables 2, 3 and 4), irrigated areas at the end
of the last millennium were reported after: (a) considering intensity (cropped areas from
different seasons are added) and (b) without considering intensity (NIA). Considering
intensity (i.e. AIA), the irrigated area was 467 Mha. Without considering intensity (i.e.
TAAI), it was 399 Mha.
4.3
Irrigated areas of the continents
Of the 467 Mha AIAs in the world, Asia accounts for 79% (370 Mha), followed by
Europe (7%) and North America (7%) (see table 3). Three continents, South America
(4%), Africa (2%) and Australia (1%), have a very low proportion of the global
annualized irrigation (table 3). In Europe and North America, an overwhelming
proportion of irrigation is during the one main cropping season (May–October). In
Asia, 154 Mha are irrigated in season two (November–February) compared with
195 Mha during season one (May–October), showing strong double cropping. In
Asia, the TAAI is 291 Mha, so the intensity of cropping is 127% (370/291), compared
to the global intensity of 117% (467/399).
4.4
Irrigated areas of the countries
Irrigated area statistics are provided for the 198 countries (table 4) and compared
with FAO/UF AQUASTATS. The countries have been ranked based on the global
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3716
Figure 10. Comparison of GIAM country-by-country irrigated areas with FAO/UF countryby-country irrigated areas. The GIAM areas are correlated with the FAO/UF statistics for: (a)
all the 198 countries and (b) 154 countries leaving out the 37 countries with near-zero areas
(either in FAO or GIAM) and seven countries where the two differ by large margins.
AIA. Of the total global AIA of 467 Mha, China has 32.5% and India has 28.3%;
together constituting a staggering total of nearly 63%. The next ranked countries
have comparatively low percentage AIAs: USA (5.2%), Pakistan (3.4%) and Russia
3717
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
(2.4%). There are eight countries (Argentina, Thailand, Bangladesh, Kazakhstan,
Myanmar, Australia, Uzbekistan and Vietnam) with 1 to 2% irrigation. Brazil is
ranked 14th with 0.88%, followed by Mexico, Indonesia and Egypt with around
0.7% each (table 4). All other countries of the world have less than 0.7% each of the
global AIA. The first 40 countries, ranked in table 4, have nearly 96% of all AIAs of
the world. Most studies (e.g. Postel 1999, Droogers 2002) consider India as the
leading irrigated area country, closely followed by China. However, our estimates
show that China has 152 Mha of AIA, while India has 132 Mha. In season one
(June–October) China with 76 Mha and India with 73 Mha are close to one another.
However, in season two (November–February), China has 68 Mha and India
54 Mha (Table 4). In addition, there is continuous (annual or plantation) irrigated
crops of 7 Mha in China and 6 Mha in India. The AIAs are equivalent to the GIAs in
the national statistics. There is no equivalent area in FAO/UF statistics to compare
with the AIA. The TAAI for China is 112 Mha and for India is 101 Mha. The TAAI
is equivalent to NIAs in the national statistics and ‘area equipped for irrigation’ in
the FAO/UF study.
4.5
Accuracies and errors
Accuracies were determined using two independent data sets for the irrigated areas
as a whole (all 28 irrigated areas put together) and for irrigated area sources: (a)
major irrigation (surface water) and (b) minor irrigation (groundwater, small
reservoirs and tanks). The groundtruth data provided an accuracy of 79% in
mapping irrigated areas with errors of omission of 21% and commission of 23%
(table 5). The Google Earth data provided an accuracy of 91%, with very low errors
of omission of 9% and also low errors of commission of 16%. The accuracies for the
irrigation sources (surface and groundwater) varied between 71 and 85% and the
errors of omission and commission were also much higher than those of aggregated
Table 5. Accuracy assessment of IWMI GIAM V2.0. The accuracy was performed for surface
water classes, ground water classes and their combinations.
Total
Correctly
Accuracy
groundtruth classified
of irrigated Errors of Errors of
sample size groundtruth area classes omissions commissions
(number)
(number)
(%)
(%)
(%)
I.
A.
Accuracy based on independent groundtruth data points
1005
793
79
Surface-water and
groundwater irrigated
areas
B. Surface-water irrigated
463
347
75
areas
C. Groundwater irrigated
542
385
71
areas
II.
A.
Accuracy based on independent Google Earth data points
323
295
91
Surface-water and
groundwater irrigated
areas
B. Surface-water irrigated
220
187
85
areas
C. Groundwater irrigated
103
79
77
areas
21
23
25
25
29
26
9
16
15
17
23
36
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3718
P. S. Thenkabail et al.
irrigated area classes, mainly as a result of intermixing between the surface water and
groundwater classes (table 5). The surface-water classes provided significantly higher
accuracies (75–85%) when compared with groundwater classes (71–77%). Accuracies
using Google Earth can be considered even better than the groundtruth data as a
result of their ability to provide: (a) a spatial view of the landscape in determining
irrigation at 1 km and 10 km scale, which can often be unrealistic from the ground, as
discussed in the following paragraph, and (b) spatially well-distributed random
points around the world.
There are a number of fundamental issues related to accuracy assessments at such
large scales as 1 km or 10 km resolution pixel size. First, there are considerable
difficulties in groundtruthing and establishing the exact percentage of area irrigated
in a 161 km (100 ha), and even more so at 10610 km (or 10 000 ha) resolutions.
Take, for example, groundtruth data collected in a portion of a pixel of area of
100 ha (161 km). Certain portions of the 100 ha may have irrigation and certain
other portions not. It is not always possible on the ground to see the entire 161 km
to understand the representativeness of the sample site location within the pixel.
Therefore, there are times that the sample site may be unrepresentative. For
example, in a pixel with 40% area irrigated and the rest ‘LULC,’ we may have a
sample site location in the LULC portion and say the pixel is non-irrigated,
completely ignoring the 40% area that is irrigated. This will lead to the pixel being
labelled ‘other LULC’ in groundtruth data, which, in reality, has 40% irrigation.
Satellite sensors capture the average reflectivity from the pixel and are hence
influenced by both the irrigated, as well as the non-irrigated components within the
pixel, leading to average spectra for the pixel. Whereas satellite data distinctly show
the difference in a pixel with zero irrigation and one with 40% irrigation,
groundtruth data often fail to do so. This will lead to situations such as, for
example: (a) rain-fed groundtruth points or other LULC points falling on a pixel
mapped as irrigated (commission error) and (b) irrigated groundtruth points falling
on a pixel mapped as other LULC (omission error). This can lead to somewhat
higher omission and commission errors. The phenomenon is acute when dealing with
pixels of low percentage (,30%) of irrigation, which have a greater likelihood of
being labelled as classes other than irrigation, resulting in highly exaggerated errors
of commission. This discussion also implies that an area-based accuracy assessment
may be more powerful and robust than a point-based accuracy assessment.
However, quality area-based reference data (e.g. irrigated area maps from national
sources) are nearly non-existent or inconsistent for most parts of the world. Offset
against this spatial advantage of remote sensing is the fact that there are multiple
reasons for an average pixel-scale signal, and it is therefore possible to confound an
interpretation with another reality.
In contrast, the VHRI (sub-metre to 4 m) available as ‘groundtruth’ from Google
Earth facilitates an aerial view of the entire 100 or 10 000 ha, which will be invaluable
in determining irrigation versus non-irrigation, based on a complete view of the pixel
rather than a certain portion of it as in ground-based data collection. Hence, the
‘zoom-in views’ of the very-high-resolution Google Earth imagery are considered
superior for accuracy assessment, compared to ground-based groundtruth.
5.
Discussions
The discussions are divided into three main parts. First, on the evaluation of the
irrigated areas obtained in this study with non-remote sensing based FAO/UF and
Global irrigated area map
3719
national statistics. Second, on the study of the causes of uncertainties in the irrigated
areas obtained in this study. Third, on the GIAM products and their potential
applications for global food security and climate change studies.
5.1 Evaluation of irrigated areas: GIAM versus other sources through four
approaches
The GIAM and areas obtained using remote sensing data and methods reported in
this paper were evaluated using four distinct approaches. These were:
N
N
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
N
N
accuracy and error assessments based on groundtruth and Google Earth data;
comparisons with FAO/UF country-by-country statistics, which, in turn, were
derived from the national statistics;
Evaluation with India’s state-by-state statistics from national sources Ministry
of Water Resources (MoWR) and Central Water Commission (CWC) state-bystate statistics; and
assessment with irrigated areas derived from finer resolution data.
It is obvious from these results and discussions why the remote sensing data and
methods provide a unique perspective on irrigated areas.
5.1.1 Comparison of country-by-country irrigated areas from GIAM versus FAO/
UF. The TAAI and AIA statistics are reported for 198 countries (table 4). Of these,
40 leading irrigated area countries consist of 96% of the global irrigation. The GIAM
areas (table 4) were compared with: (a) an FAO/UF map (figure 10) and their
statistics in FAO AQUASTAT (see a summary in the last column in table 4); and (b)
national statistics (figure 11). Of the 198 countries (table 4), the GIAM areas were
significantly similar (difference ,5000 ha) to FAO/UF in 26% of the countries. In
44% of the countries, GIAM underestimates areas and in 30% of the countries
GIAM overestimates areas. The GIAM TAAI in the world is 399 Mha, the
equivalent of which in FAO/UF is area equipped for irrigation, which was 279 Mha.
However, there is a definitive trend between GIAM and FAO/UF area (see
figure 11). The combined GIAM TAAI for China and India is 102 Mha higher than
the FAO/UF equipped irrigated areas. A comparison of national statistics helps
explain some of these differences. For example, the official statistics of irrigated
areas in India, the second leading country in irrigated areas, released by the
Department of Economics and Statistics (DES) is 57 Mha (FAO AQUASTAT
reports national statistics and hence has the same numbers as those of the DES). The
official statistics from the DES overwhelmingly depend on the 162 major and 221
medium (major and medium commonly referred to simply as major) command areas
and some other surface-water schemes.
5.1.2 Evaluation of state-by-state irrigated areas from GIAM versus national
statistics for India. However, recently released minor irrigation (groundwater, small
reservoirs and tanks) statistics for 2000–2001 from India’s MoWR (http://
mowr.gov.in/micensus/mi3census/index.htm) when combined with major and
medium irrigated area statistics provide a more realistic estimate of irrigated areas.
The MoWR estimates show irrigation potential utilized (IPUutilized2total) as 84 Mha
and irrigation potential created (IPCcreated2total) as 111 Mha. The IPUutilized2total and
IPCcreated2total both have intensity of cropping significantly lower than the AIA
(figure 12). In China, it was demonstrated that Landsat 30 m based remote sensing
estimated arable areas were 6.2 times higher than the areas estimated by the Ministry
3720
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
of Agriculture (Liu 2000). A 1 : 1 plot between the FAO/UF and GIAM TAAI for
the 198 countries showed a slope of 0.54 and a high R2 value of 0.94 (figure 10(a)).
The GIAM TAAI and FAO/UF have a remarkable slope of nearly a perfect 1
(R250.94) for the 154 countries (out of 198), each of which has 10 irrigated areas of
1 Mha or less (figure 10(b)). Of the 198 countries in figure 10(b), 44 were left out
because the GIAM TAAI and FAO/UF had: (a) zero or near-zero irrigated areas in
37 countries and (b) huge differences in seven countries (China, India, Russia,
Australia, Argentina, Kazakhstan and Iraq).
Figure 11. Evaluation of GIAM state-by-state irrigated areas in India with the state-by-state
irrigated areas from the Indian National Statistics. The GIAM AIAs are correlated with the
irrigated potential utilized (IPU) from India’s Ministry of Water Resources (MoWR) and
Central Water Commission (CWC) for: (a) all 33 Indian states and union territories, (b) 31
states and union territories after leaving out the two states with greatest discrepancy. The
GIAM AIAs are also compared with irrigation potential created (IPC) from MoWR and
CWC for: (c) all 33 Indian states and union territories and (d) 31 states and union territories,
leaving out the two states with the greatest discrepancy. (e) The sum of the areas from the
GIAM versus national statistics for the entire country.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
(Continued.)
Global irrigated area map
Figure 11.
3721
P. S. Thenkabail et al.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3722
Figure 12. Validation of GIAM irrigated areas using finer resolution data. The Landsat
30 m data was used for detailed studies in (a) the Krishna basin and (b) Ghana to establish
irrigated areas at 30 m. WSA: Water Spread Area in hac.
Global irrigated area map
5.2
3723
Causes of uncertainties in irrigated areas
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
The factors that influence the varying estimates of irrigated areas reported by the
IWMI GIAM versus FAO/UF versus national statistics are discussed below.
5.2.1 Minor irrigation statistics are inadequately accounted for in national
statistics. There is sufficient evidence that the minor irrigation (groundwater, small
reservoirs and tanks) statistics are inadequately accounted for in many countries. We
illustrate this for India. The DES reports India’s NIAs as 57 Mha, a huge difference
from the 101 Mha of the GIAM TAAI. The DES overwhelmingly depends on the
162 major and 221 medium surface-water schemes. However, recently released minor
irrigation (groundwater, small reservoirs and tanks) statistics for 2000–2001 from
India’s MoWR (http://mowr.gov.in/micensus/mi3census/index.htm), when combined
with major and medium irrigated area statistics, provide a more realistic estimate of
irrigated areas. In India, the AIA estimates of various Indian states are compared
with the MoWR estimates of the Indian states for: (a) IPUutilized2total (figures 11(a)
and (b)) and (b) IPCcreated2total (figures 11(c) and (d)). First, the AIA with
IPUutilized2total showed an R2 value of 0.76 for a 1 : 1 line. The AIA is 1.34 times
the IPU. The biggest differences were in two states: Madhya Pradesh where GIAM
AIA overestimates, and Punjab, where GIAM AIA underestimates. If we leave these
two states, the R2 value goes up to 0.89. Second, the AIA with IPUcreated2total
showed an R2 value of 0.84 for a 1 : 1 line. The AIA is 1.05 times the IPC. If we leave
out the two states where the differences are very high, the R2 value goes up to 0.92. It
is most appropriate to compare the AIA with the IPUutilized2total. However, it is
likely that the AIA may be picking up the IPUcreated2total, given the nominal pixel
size (10km610km). Overall, the AIA is consistently higher than areas reported in
the national statistics (figure 11) as a result of the inadequate accounting of the
minor irrigation (groundwater, small reservoirs and tanks). In order to prove this
point, we carried out two special case studies in India and Africa using Landsat 30m
data (Gumma et al. 2009, Velpuri et al. 2009). First, the Krishna basin in India
showed 6097 small reservoirs (figure 12(a)) that, along with groundwater, irrigated
54% of all the irrigated areas, which is estimated as 9.4 Mha (Velpuri et al. 2009).
Only 46% of the total irrigated areas of 9.4 Mha is irrigated by the 24 major
reservoirs (figure 12(a)). The CBIP (1994) irrigated area map (see grey areas in
figure 12(a)) almost completely ignores these smaller reservoirs, tanks and groundwater irrigation. The study involved extensive field visits. In contrast, the total
known irrigated area reported for the Krishna basin is just 4.16 Mha, as calculated
from the FAO/UF map. However, it is obvious from the detailed high-resolution
study that the statistics are underestimated, and an area of 9.4 Mha is about the same
as that determined using 500m data by Dheeravath et al. (2009).
Sinha (2003) also indicated irrigated areas in India to be 100 Mha. In a recent
independent study, Dheeravath et al. (2009) used 500 m MODIS 8 day time series
data of 2001–2003 for India to show the TAAI for India was 113 Mha and the AIA
was 147 Mha, which are closer to the TAAI and AIA of this study. The MoWR
(2005) data show that, out of 111 Mha, 74 Mha are from minor irrigation sources
and 37 Mha from major irrigation sources. The GIAM TAAI estimates minor
irrigation as 60 Mha and major irrigation as 41 Mha. This is mainly as a result of the
growth in groundwater wells in India, estimated to vary between 19 and 26 million
(Endersbee 2005). An overwhelming number of these are used for irrigation.
However, the irrigated area maps and statistics on groundwater irrigation are
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3724
P. S. Thenkabail et al.
sketchy and/or missing. Indeed, overwhelming evidence (see Shah et al. 2003, 2004,
Endersbee 2005, MoWR 2005) shows massive overexploitation of groundwater in
most of India, and the majority of the potential is already exploited. In addition, the
massive exploitation of surface water from minor reservoirs is a missing link. Given
these facts, it is obvious that the NIA of India far exceeds the officially reported
57 Mha, and the value is closer to 100 Mha of the TAAI reported in this study, or
even slightly higher as reported by Dheeravath et al. (2009). The studies of Liu (2000)
in China also indicate similar trends.
In Africa, the FAO/UF reports equipped irrigated areas of Ghana to be 6374 ha,
whereas GIAM reported the TAAI to be 60 647 ha. The Ghanaian National
statistics reports areas to be 14 699 ha (gathered from Busia, the irrigation
department of Ghana). However, our case study, using Landsat 30m data
(Gumma et al. 2009), backed by field visits showed the TAAI to be as high as
61 826 ha (e.g. figure 12(b)), which is about the same as reported for Ghana by the
GIAM. Our field observations in Ghana also established that, if we include the
supplemental irrigation of rice in the inland valley bottoms, the irrigated areas will
still be higher than the 60 647 ha.
5.2.2 IAFs in GIAM may need local fractions. In this study, IAFs (Thenkabail et al.
2007b) were derived using three distinctly independent methods, and at least two
methods were used to compute the irrigated areas: TAAI and AIA. Nevertheless,
deeper understanding of each of the GIAM28 classes through groundtruths will help
improve IAFs and hence irrigated areas. Computation of local IAFs, for nations and
regions, instead of global IAFs are expected to help improve irrigated area classes.
However, it is not clear whether local IAFs will increase or decrease the areas. The
AIA is calculated based on the seasonality of every class (table 2 and figure 8(a)). The
seasonality of a class is determined based on the NDVI time-series plot of every
class.
5.2.3 Resolution influencing irrigated areas. Within the GIAM project, irrigated
areas have been estimated for certain regions of the world at 500 m (Dheeravath et al.
2009) and 30 m (Velpuri et al. 2009) resolution, apart from the nominal 10 km
resolution reported in this paper. A comparison of areas estimated from these
different resolutions showed that the finer the resolution, the greater the area was.
This is because, at finer resolution, fragmented irrigated areas, such as from
groundwater, can be picked up better. Coarser resolution imagery can miss some of
the fragmented irrigation, especially when the fragmented proportion is less than
40%. Ozdogan and Woodcock (2006) imply that, with coarser resolution, areas are
actually higher. This is as a result of the FPA computed as actual area without using
IAF and/or using a very-high area fraction when using coarser resolution imagery.
Ozdogan and Woodcock also showed that the Landsat 30 m was too coarse to
estimate the real extent of cultivated land in parts of China, while, for the USA, a
resolution of 500 m was sufficient. However, in China, there are vast stretches of
contiguous areas of irrigation, even when the field sizes are small. Therefore, even
using coarse resolution imagery, these areas can be mapped with certainty. Our
recent field visit to China established this fact. We use the filed visit data in an
accuracy assessment. It is clear from these discussions that further studies are
required to draw firm conclusions on relationships between irrigated areas and
resolution of the imagery. Indications are that: (a) ‘the finer the resolution, the
greater is the area when irrigated areas are in fragments’. This is because in coarser
Global irrigated area map
3725
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
resolution pixels, fragmented areas do not get accounted adequately and/or may
miss out completely, resulting in underestimation of irrigated areas and (b) ‘the finer
the resolution, the lesser is the area in contiguous areas’. This is because, in
contiguous areas, finer resolution imagery will separate out small fragments, such as
settlements, roads and abandoned lands, and these areas get deducted out of the
total area of fine-resolution imagery. However, in coarser resolution, smaller
proportional areas, such as roads, settlements and other non-irrigated areas, get
merged into larger proportional irrigated areas in the pixel, thus overestimating
irrigated areas. These discussions clearly imply the need for further study in relating
resolution to areas.
5.2.4 Minimum mapping unit (MMU) in determining areas. When the MMU is
large (e.g. unit area is 10 000 ha), fragments of irrigated areas, such as 100 or 1 000 ha
blocks, can miss out completely from being represented on a map at particular
scales, such as, for example, 1 : 10 million. This will lead to miscalculations of the
areas and their underestimation. We have also observed irrigated area maps that had
far higher areas when the areas were calculated after digitizing, because polygons of
areas are drawn as if they are contiguous units, whereas in reality, the polygons will
have several LULC. In one such map, the irrigated areas of India from India’s CBIP
(1994) were digitized by the authors, and the area summed up to 75 Mha, whereas
the statistics reported the area as 57 Mha.
It is also possible that the AVHRR off-nadir views, missing scan lines and
processing for global area coverage (GAC) can also induce greater MMU than that
which the pixel resolution indicates.
5.2.5 Supplemental classes. In reality, when evapo-transpiration outweighs precipitation, the only way for the crops to be sustained is through irrigation. The
GIAM considers areas with significant supplemental irrigation (more than one
irrigation in the crop-growing season) as an irrigated conjunctive use class (classes
16–28 in figure 8(a)). Often, supplemental irrigated area classes are categorized as
‘rain-fed’ in irrigated area maps, resulting in underestimation of irrigated areas.
5.2.6 Traditional versus remote sensing data. Traditionally, grassroots level
irrigated areas observed in the field are reported next to the higher administrative
unit and so on till the synthesis reaches national level. However, this type of datagathering has many pitfalls, such as misreporting, inconsistencies in reporting as a
result of the involvement of a large number of data gatherers and reporters and
errors in data entry and/or synthesis. In contrast, remote sensing offers a platform of
consistent data across space and time, facilitating application of consistent methods
and techniques to derive irrigated area statistics.
On the other hand, time series remote sensing data potentially allow the often
distinct dynamics of irrigated agriculture to stand out from other land uses, but there
are many confusing situations, for instance, in the tropics, where rice may be mainly
rain-fed in the monsoonal season, but receives some irrigation and is followed by one
or more dry season crops, which may be completely irrigated. In tropical
environments, there is generally a high degree of land cover the whole year-round
and everything is ‘green,’ making precise definition of irrigated crops more difficult,
especially if relatively coarse-scale imagery is used. Such definition issues will cause
uncertainty in irrigated area estimates.
3726
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
5.3
P. S. Thenkabail et al.
Discussions on the methods used
The strengths of the methods used in this paper to analyse and discern irrigated
areas using remote sensing are manyfold. First, is the innovative composition of the
MFDCs that facilitates analysis of multiple sensor time-series data of hundreds, or
even thousands, of layers in one go. Second, the unique concept of the
development of ideal spectra, based on ground knowledge from precise locations,
which are then used to generate ideal spectra using time-series MFDCs. Third,
development and/or adoption of unique methods, such as the SMT to group and
identify classes of similar characteristics and to match class spectra with ideal (or
target) spectra to help identify and label classes. Fourth, comprehensive and
innovative class identification and labelling protocol that encompasses use of
extensive groundtruth data, Google Earth VHRI zoom-in views (GE VHRI ZIW),
BGW 2D feature space plots, ST-SC plots and time-series NDVI plots. Fifth,
adoption of approaches to resolve mixed classes by using decision-tree algorithms
and spatial modelling using myriad GIS data. Sixth, validation of output products
using a number of approaches that include accuracies and error assessments,
comparison with national statistics and use of GE VHRI ZIW. Seventh, adoption
of sub-pixel irrigated area (SPIA) calculation methods that are robust and provide
actual irrigated areas.
There is no single method or technique that can be successfully applied to obtain a
solution to mapping irrigated areas at a global scale (Thenkabail et al. 2005, 2006). A
suite of methods, as discussed in §2, are required. It is possible to apply other
methods, such as decision-tree algorithms (DeFries et al. 1998) and Fourier
transforms (Canisius et al. 2007). However, all of them have their own strengths and
limitations. Initially, in the project, we did explore a number of possibilities that
include decision trees, Fourier transforms and unsupervised classifications.
However, the methods used in this study were innovative and powerful.
6.
GIAM products and their applications
The IWMI’s GIAM nominal 10 km V2.0 products are released through the web
portal http://www.iwmigiam.org. The pixel resolution of the product is actually
1 km, as SPOT VGT data had that resolution. However, since the overwhelming
amount of data is AVHRR 10 km, the product is referred to as 10 km. Irrigated
area statistics are provided for the 198 countries (http://www.iwmigiam.org/stats).
The GIAM web map server makes it possible to zoom-in to the area of interest
and instantly print the dynamic and automated map composition for any
country or region (http://www.iwmigiam.org/mapper.asp). Spatial spread of
irrigated areas of any country in the world can be obtained instantly by calling
the country from a drop-down menu. The maps are made printable with all
map layout features. The GIAM map can also be uploaded onto Google Earth
using the ‘kmz’ file (http://www.iwmigiam.org/info/main/index.asp). The portal is
the place to find GIAM maps, images, method documents, area calculation
procedures, posters, animations, comparisons and a host of other data and
products.
The products are expected to play a key role in a number of applications, which
include:
N
N
global food security studies;
water use/evapo-transpiration studies;
Global irrigated area map
N
N
N
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
7.
3727
water productivity mapping;
inputs leading to improvements of existing global maps; and
climate change studies.
Conclusions
The study developed a suite of methods and protocols for GIAM using remote
sensing data. First, the study demonstrated the value of composing multiple-sensor
and secondary data in a single MFDC of hundreds of data layers, akin to
hyperspectral data. Second, the paper demonstrated the utility of quantitative SMTs,
such as SCSs and SSVs to group, identify and label classes derived by classifying
MFDCs of various segments of the world. Third, extensive standard protocols for
class identification and labelling were developed to establish different types of
irrigated area classes and to differentiate irrigated areas from non-irrigated areas
using a hierarchical classification system. These protocols consisted of: (a) ST-SCs,
(b) BGW plots, (c) Google Earth VHRI, (d) two large sources of well-distributed
global groundtruth data from 5 654 locations, (e) high-resolution Landsat ETM +
mosaics, (f) time-series NDVI plots and (g) secondary data. Fourth, SPA calculation
methods were introduced where SPAs were determined by multiplying IAFs by
FPAs.
The study produced the first satellite-sensor-based GIAM at a nominal resolution
of 10 km. This 28 class map (GIAM28) provided classes labelled based on irrigation
source (e.g. surface water, groundwater or conjunctive use), intensity (e.g. single,
double or continuous crop) and crop dominance. The global irrigated areas are
reported for the end of the last millennium in terms of: (1) AIA and (2) TAAI. The
AIA considers intensity or seasonality of irrigation and hence sums up areas
irrigated during season one, season two and continuous (e.g. perennial crops or
plantations or year-round crops). This is also referred to as gross area. The TAAI is
the area irrigated at any given point of time plus the area equipped for irrigation but
left fallow during that same period of time. This is equivalent to net area. The AIA of
the world at the end of the last millennium was 467 Mha and the TAAI was 399 Mha.
Of the 467 Mha AIAs, there were: (a) 252 Mha during season one, (b) 174 Mha
during season two and (c) 41 Mha continuous year-round crops, such as sugarcane
and plantations. Globally, irrigation by surface water was 61% and the rest (39%) by
groundwater. Asia accounts for 79% (370 Mha) of all AIAs, followed by Europe
(7%) and North America (7%). Three continents, South America (4%), Africa (2%)
and Australia (1%), have a very low proportion of global irrigation. China and India
are the two leading irrigated area countries, with a combined total of nearly 62% of
the global AIAs, mainly as a result of cropping intensity (cropping during multiple
seasons in a given year). Of this, China has 32.5% (152 Mha) of the global AIAs and
India has 28.3% (132 Mha) of the AIAs. This is followed by the USA (5.2%),
Pakistan (3.4%) and Russia (2.4%). Eight other countries (Argentina, Thailand,
Bangladesh, Kazakhstan, Myanmar, Australia, Uzbekistan and Vietnam) have areas
between 1 and 2% and four others (Brazil, Mexico, Indonesia and Egypt) between
0.7 and 1%. All other countries have less than 0.7% of the global AIAs. The TAAI
for China is 112 Mha and for India 101 Mha. The GIAM had an accuracy of
79–91%, with errors of omission not exceeding 21%, and the errors of commission
not exceeding 23%.
Accuracies were determined using two independent databases. The irrigated areas
(all 28 classes put together) were mapped with an accuracy varying between 79 and
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3728
P. S. Thenkabail et al.
91%, with errors of omission less than 21%, and errors of commission less than 23%.
Accuracies were also assessed for irrigation sources: (a) major irrigation (major and
medium surface-water reservoirs) and (b) minor irrigation (groundwater, small
reservoirs and tanks). Minor irrigation classes were generally more difficult to map
with an accuracy of 71–77% when compared with major irrigation, which had an
accuracy of 75–85%. This was mainly due to the intermixing of classes between
major and minor irrigation.
Extensive comparisons were also made between the GIAM statistics with the
FAO/UF and India’s national statistics. The GIAM TAAI of 399 Mha was much
higher than the FAO/UF areas equipped for irrigation (279 Mha). However, the
GIAM TAAI and FAO/UF have a remarkable slope of a nearly perfect 1
(R250.94) for the 154 countries (out of 198), each of which has 10 irrigated areas
of 1 Mha or less. Detailed comparisons were also made between the GIAM
statistics and India’s national census data. The irrigation potential utilized
(IPUutilized2total) of India’s national statistics was 84 Mha and irrigation potential
created (IPCcreated2total) was 111 Mha. The AIA versus IPUutilized2total for the 32
Indian states and union territories showed an R2 value of 0.76 for a 1 : 1 line. The
AIA was 1.34 times the IPUutilized2total. The main causes of differences between
GIAM irrigated areas, when compared with the national statistics and/or FAO/UF
statistics, were due to factors such as: (a) inadequate accounting of informal minor
irrigation (e.g. groundwater, small reservoirs and tanks) statistics in the national
census, (b) uncertainties in IAFs in the GIAM, (c) inconsistencies in the national
census data on how the irrigated areas are compiled, (d) resolutions and/or scales
at which the irrigated area statistics are derived and (e) definition issues, leading to
inclusion of significant supplemental irrigated areas as irrigated areas in GIAM,
whereas most traditional statistics fail to do so. These are issues that need further
investigation.
Particular strengths of this work were: (a) establishing AIAs that consider
intensity in addition to TAAI, which does not consider intensity, (b) mapping
informal minor irrigation (e.g. groundwater, small reservoirs and tanks), in addition
to conventional surface-water major irrigation, (c) determining crop calendars of
irrigated areas and (d) ability to simulate trends in biomass dynamics of irrigated
areas over time. The possibilities for improvements exist by refining IAFs further
through more intensive groundtruth and by calculating irrigated areas of every pixel
by multiplying the IAF of the pixel with the FPA of the pixel in an algorithm.
The irrigated area maps and statistics for the 198 countries of the world are
provided through the GIAM web portal http://www.iwmigiam.org.
Acknowledgements
The authors would like to gratefully acknowledge the support and guidance
provided by Prof. Frank Rijsberman, former Director General of the IWMI. It is
mainly due to his vision that a project of this complexity was possible. We would like
to also thank Mr Sarath Abayawardana, former Head of the Global Research
Division of the IWMI, who was instrumental in supporting the remote sensing and
GIS unit initially, which later developed into a well-respected unit globally. The
authors would like to thank Mr Sarath Gunasinge and Mr Ranjith Alankara for
data compilation support and all the hard work relating to compilation of
groundtruth data. Secretarial support from Jacintha Navaratne, Arosha Ranasinghe
and Samanmali Jayathilake is much appreciated. The AVHRR pathfinder data used
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
3729
by the authors in this study include data produced through funding from the Earth
Observing System Pathfinder Program of NASA’s Mission to Planet Earth in
cooperation with National Oceanic and Atmospheric Administration. The data were
provided by the Earth Observing System Data and Information System (EOSDIS),
the Distributed Active Archive Center at the Goddard Space Flight Center, which
archives, manages and distributes this data set. This project would not have been
possible without the availability of high-quality data made available for free. In this
regard, we would like to thank SPOT Image (France) for SPOT VGT data, USGS
for GTOPO30 and numerous other data, Dr Tim Mitchell of the CRU of the East
Anglian University (UK) for precipitation data, the global rain forest mapping
project (GRFM) team of NASA/JPL for the JERS-1 SAR data, the Enterprise of
Google Earth visionaries in making available sub-metre to 4 m data to large parts
of the world, the thousands of volunteers for the DCP, the ESRI, University of
Maryland, NASA and Earth Satellite Corl. (now renamed MDA Federal Inc.) for
the Landsat Geocover mosaic of the world and the global land cover facility of the
University of Maryland for the forest cover map (DeFries et al.). Interactions,
discussions and insights from the FAO/UF team (Dr Stefan Siebert et al.) are much
appreciated. Finally, we are very grateful to our partners in China (Dr Mei Xurong
and Dr Hai Weiping from the Chinese Academy of Agricultural Sciences and Dr
Songcai You from the Chinese Academy of Sciences) and India (Dr Mangala Rai,
Dr J.S. Samra, Dr A.K. Maji and Dr Obi Reddy of the Indian Council for
Agricultural Research and Dr Bharat Sharma of the IWMI Delhi office) for making
available national data, and for facilitating, support, discussions and insights. This
paper is not internally reviewed by USGS Geological Survey (USGS) and hence the
views expressed in this paper are not endorsed by USGS.
References
AGBU, P.A. and JAMES, M.E., 1994, NOAA/NASA Pathfinder AVHRR Land Data Set User’s
Manual (Greenbelt: Goddard Distributed Active Archive Center, NASA Goddard
Space Flight Center).
AGRAWAL, S., JOSHI, P.K., SHUKLA, Y. and ROY, P.S., 2004, Spot-vegetation multi temporal
data for classifying vegetation in South Central Asia. Current Science, 84, pp.
1440–1448.
BARTHOLOME, E. and BELWARD, A.S., 2005, GLC2000: a new approach to global land cover
mapping from Earth observation data. International Journal of Remote Sensing, 26,
pp. 1959–1977.
BELWARD, A.S., ESTES, J.E. and KLINE, K.D., 1999, The IGBP-DIS 1 km land cover data set
DISCover: a project overview. Photogrammetric Engineering and Remote Sensing, 65,
pp. 1013–1020.
BIGGS, T., THENKABAIL, P.S., KRISHNA, M., GANGADHARARAO, P. and TURRAL, H., 2006,
Vegetation phenology and irrigated area mapping using combined MODIS timeseries, ground surveys, and agricultural census data in Krishna River Basin, India.
International Journal of Remote Sensing, 27, pp. 4245–4266.
BING, Z., JIANGU, L., XIANGJUN, W. and CHANGSHAN, W., 1998, Study on the classification of
hyperspectral data in urban area. SPIE, 3502.0277-786x/98.
BIRADAR, C.M., THENKABAIL, P.S., NOOJIPADY, P., LI, Y.J., DHEERAVATH, V., VELPURI, M.,
TURRAL, H., GUMMA, M.K., REDDY, O.G.P., CAI, X.L., SCHULL, M.A.,
ALANKARA, R.D., GUNASINGHE, S., MOHIDEEN, S. and XIAO, X., 2009, A global
map of rain-fed cropland areas (GMRCA) at the end of the last millennium using
remote sensing. International Journal of Applied Earth Observation and
Geoinformation, 11, pp. 114–129.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3730
P. S. Thenkabail et al.
CANISIUS, F., TURRAL, H. and MOLDEN, D., 2007, Fourier analysis of historical NOAA time
series data to estimate bimodal agriculture. International Journal of Remote Sensing,
28, pp. 5503–5522.
CENTRAL BOARD OF IRRIGATION AND POWER (CBIP), 1994, Central Board of Irrigation and
Power Irrigated Area Map of India for 1994 (New Delhi, India: CBIP).
COMPREHENSIVE ASSESSMENT OF WATER MANAGEMENT IN AGRICULTURE (CA), 2007, Water
for Food and Water for Life: A Comprehensive Assessment of Water Management in
Agriculture, D. Molden (Ed.) (London: Earthscan and Colombo: IWMI).
CONGALTON, R.G., 1994, Accuracy Assessment of Remotely Sensed Data: Future Needs and
Directions. In Proceedings of Pecora 12 Land Information from Space-Based Systems
(Bethesda: ASPRS), pp. 383–388.
DEFRIES, R.S. and TOWNSHEND, J.R.G., 1994, Global land cover: comparison of ground
based data sets to classifications with AVHRR data. In Environmental Remote Sensing
from Regional to Global Scales, G.M. Foody and P.J. Curran (Eds), pp. 84–110 (New
York: Wiley).
DEFRIES, R., HANSEN, M. and TOWNSHEND, J., 1995, Global discrimination of land cover
types from metrics derived from AVHRR Pathfinder data. Remote Sensing of
Environment, 54, pp. 209–222.
DEFRIES, R., HANSEN, M., TOWNSEND, J.G.R. and SOHLBERG, R., 1998, Global land cover
classifications at 8 km resolution: the use of training data derived from Landsat
imagery in decision tree classifiers. International Journal of Remote Sensing, 19, pp.
3141–3168.
DEFRIES, R., HANSEN, M., TOWNSHEND, J.R.G., JANETOS, A.C. and LOVELAND, T.R., 2000a,
Continuous Fields 1 Km Tree Cover (College Park, Maryland: The Global Land Cover
Facility).
DEFRIES, R., HANSEN, M., TOWNSHEND, J.R.G., JANETOS, A.C. and LOVELAND, T.R., 2000b,
A new global 1 km data set of percentage tree cover derived from remote sensing.
Global Change Biology, 6, pp. 247–255.
DHEERAVATH, V., THENKABAIL, P.S., NOOJIPADY, P., BIRADAR, C.B., TURRAL, H.,
CHANDRAKATHA, G., GUMMA, M., REDDY, G.P.O. and VELPURI, M., 2009,
Irrigated areas of India derived using MODIS 500 m data for years 2001–2003.
ISPRS Journal of Photogrammetry and Remote Sensing (in review).
DÖLL, P. and SIEBERT, S., 2000, A digital global irrigated area map. ICID Journal, 49, pp.
55–66.
DROOGERS, P., 2002, Global Irrigated Area Mapping: Overview and Recommendations,
Working Paper 36 (Colombo, Sri Lanka: IWMI).
EIDENSHINK, J.C. and FAUNDEEN, J.L., 1994, The 1 km AVHRR global land data set: first
stages implementation. International Journal of Remote Sensing, 15, pp. 3443–3462.
ENDERSBEE,, 2005, A Voyage of Discovery: A History of Ideas About the Earth, With a New
Understanding of the Global Resources of Water and Petroleum, and the Problems of
Climate Change (Frankston, Victoria, Australia: Lance Monash University
Bookshop).
FARRAND, W.H. and HARSANYI, J.C., 1997, Mapping the distribution of mine tailings in the
Coeur d’Alene River Valley, Idaho, through the use of a constrained energy
minimization technique. Remote Sensing of Environment, 59, pp. 64–76.
FLEIG, A.J., HEATH, D.F., KLENK, K.F., OSLIK, N., LEE, K.D., PARK, H., BARTIA, P.K. and
GORDON, D., 1984, User’s Guide for the Solar Backscattered Ultraviolet (SBUV) and
the Total Ozone Mapping Spectrum (TOMS) RUT-S and RUT-T Data Sets: October
31 1978 to November 1980. NASA Reference Publication 1112. Report Number:
NAS1.61:1112, NASA-RP-1112, REPT-910.
FOODY, G., 2002, Status of land cover classification accuracy assessment. Remote Sensing of
Environment, 80, pp. 185–201.
FRAMJI, K., GARG, B. and LUTHRA, S., 1981, Irrigation and Drainage in the World, volume I,
p. 82 (New Delhi: ICID).
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
3731
GRANAHAN, J.C. and SWEET, J.N., 2001, An evaluation of atmospheric correction techniques
using the spectral similarity scale. In IEEE 2001 International Geoscience and Remote
Sensing Symposium, 5, pp. 2022–2025.
GUMMA, M.K., THENKABAIL, P., FUJII, H. and NAMARA, R., 2009, Spatial models for
selecting the most suitable areas of rice cultivation in the Inland Valley Wetlands of
Ghana using remote sensing and geographic information systems. Journal of Applied
Remote Sensing, 3, 033537.
HOLBEN, B.N., 1986, Characteristics of maximum-value composite images from temporal
AVHRR data. International Journal of Remote Sensing, 7, pp. 1417–1434.
HOMAYOUNI, S. and ROUX, M., 2003, Material mapping from hyperspectral images using
spectral matching in urban area. Submitted to IEEE Workshop in honour of Prof.
Landgrebe, October 2004, Washington, DC, USA.
INTERNATIONAL GEOSPHERE–BIOSPHERE PROGRAMME (IGBP), 1990, Global Change, Report
no. 12 (Stockholm, Sweden: IGBP).
KAUTH, R.J. and THOMAS, G.S., 1976, The tasseled cap – a graphic description of the spectral–
temporal development of agricultural crops as seen by Landsat. In Proceedings of the
Symposium on Machine Processing of Remotely Sensed Data, Purdue University, West
Layfayette, Indiana, pp. 4B-41–4B-50.
KIDWELL, K., 1991, NOAA Polar Orbiter Data User’s Guide (NCDC/SDSD, Washington, DC:
National Climatic Data Center).
KOGAN, F.N. and ZHU, X., 2001, Evolution of long-term errors in NDVI time-series: 1985–
1999. Advances in Space Research, 28, pp. 149–154.
LEICA, 2005, Earth Resources Data Analysis System (ERDAS Imagine 8.0) Leica Geosystems,
A.G. Heinrich-Wild-Strasse, CH-9435, Heerbrugg, Switzerland.
LEICA, 2007, Earth Resources Data Analysis System (ERDAS Imagine8.x) Leica Geosystems,
A.G. Heinrich-Wild-Strasse, CH-9435, Heerbrugg, Switzerland.
LISSENS, G., KEMPENEERS, P., FIERENS, F. and VAN RENSBERGEN, J., 2000, Development of
cloud, snow, and shadow masking algorithm for VEGETATION imagery. In
Proceedings of IGARSS 2000, Hawaii.
LIU, J.B., 2000, Study of spatial-temporal feature of modern land-use change in China: using
remote sensing techniques. Quaternary Sciences, 20, pp. 229–239 (in Chinese).
LOVELAND, T.R., REED, B.C., BROWN, J.F., OHLEN, D.O., ZHU, J., YANG, L. and
MERCHANT, J.W., 2000, Development of a global land cover characteristics
Database and IGBP DISCover from 1 km AVHRR Data. International Journal of
Remote Sensing, 21, pp. 1303–1330.
LOVELAND, T.R., ZHU, Z., OHLEN, D.O., BROWN, J.F., REED, B.C. and YANG, L., 1999, An
analysis of the IGBP global land-cover characterization process. Photogrammetric
Engineering and Remote Sensing, 65, pp. 1021–1032.
MERRETT, S., 2002, Water for Agriculture: Irrigation Agriculture in International Perspectives .
Spon’s Environmental Science and Engineering Series (New York, NY: Taylor &
Francis).
MINISTRY OF WATER RESOURCES (MOWR), 2005, Third Census of Minor Irrigation Schemes
(2000–2001). Available online at: http://mowr.gov.in/micensus/mi3census/index.htm
(New Delhi: MoWR, Government of India).
MITCHELL, T.D., CARTER, T.R., JONES, P.D., HULME, M. and NEW, M., 2003, A
comprehensive set of high-resolution grids of monthly climate for Europe and the
globe: the observed record (1901–2000) and 16 scenarios (2001–2100). Journal of
Climate, submitted.
NATIONAL GEOPHYSICAL DATA CENTER (NGDC), 1994, 5 Minute Gridded World Elevation.
NGDC data announcement DA 93-MGG-01 (Boulder, CO: NGDC).
OLSON, J.S., 1994a, Global Ecosystem Framework: Definitions. USGS EROS Data Center
Internal Report (Sioux Falls, SD: USGS).
OLSON, J.S., 1994b, Global Ecosystem Framework: Translation Strategy. USGS EROS Data
Center Internal Report (Sioux Falls, SD: USGS).
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
3732
P. S. Thenkabail et al.
OZDOGAN, M. and WOODCOCK, C.E., 2006, Resolution dependent errors in remote sensing of
cultivated areas. Remote Sensing of Environment, 103, pp. 203–217.
PEUQUET, D.J. and MARBLE, D.F., 1990, Introductory Readings in Geographic Information
Systems (New York: Taylor & Francis).
POSTEL, S., 1999, Pillar of Sand: Can the Irrigation Miracle Last? (New York: W.W. Norton).
RAO, C.R.N., 1993a, Nonlinearity Corrections for the Thermal Infrared Channels of the
Advanced Very High Resolution Radiometer: Assessment and Recommendations.
NOAA Technical Report NESDIS-69 (Washington, DC: NOAA/NESDIS).
RAO, C.R.N., 1993b, Degradation of the Visible and Near-infrared Channels of the Advanced
Very High Resolution Radiometer on the NOAAP9 Spacecraft: Assessment
and Recommendations for Corrections. NOAA Technical Report NESDIS-70
(Washington, DC: NOAA/NESDIS).
SAATCHI, S.S., NELSON, B., PODEST, E. and HOLT, J., 2000, Mapping land cover types in the
Amazon Basin using 1 km JERS-1. International Journal of Remote Sensing, 21, pp.
1201–1235.
SCHWARZ, J. and STAENZ, K., 2001, Adaptive threshold for spectral matching of hyperspectral
data. Canadian Journal of Remote Sensing, 27, pp. 216–224.
SECKLER, D., AMARASINGHE, U., DE FRAITURE, C., KELLER, A., MOLDEN, D. and
SAKTHIVADIVEL, R. (Eds), 2000, World Water Supply and Demand: 1995 to 2025
(Colombo, Sri Lanka: IWMI).
SHAH, T., SINGH, O.P. and MUKHERJEE, A., 2004, Groundwater irrigation and South Asian
Agriculture: empirical analyses from a large-scale survey of India, Pakistan, Nepal
Terai, and Bangladesh. In IWMI-Tata Annual Partners’ Meeting, 17–19 February,
Anand, India.
SHAH, T., ROY, A.D., QURESHI, A.S. and WANG, J., 2003, Sustaining Asia’s groundwater
boom: an overview of issues and evidence. Natural Resources Forum, 27, pp. 130–140.
SHIPPERT, P., 2001, Spectral and hyperspectral analysis with ENVI. ASPRS ENVI user’s
group notes. In Annual Meeting of the American Society of Photogrammetry and
Remote Sensing, 22–28 April, St. Louis, Missouri, USA.
SIEBERT, S., FEICK, S. and HOOGEVEEN, J., 2005a, A digital global map of irrigation areas – an
update for Asia. In Frankfurt Hydrology Paper 01 (Germany: Frankfurt University).
SIEBERT, S., HOOGEVEEN, J. and FRENKEN, K., 2006, Irrigation in Africa, Europe, and Latin
America. Update of the Digital Global Map of Irrigation Areas to Version 4. In
Frankfurt Hydrology Paper 05 (Germany: University of Frankfurt).
SIEBERT, S., DÖLL, P., HOOGEVEEN, J., FAURES, J.-M., FRENKEN, K. and FEICK, S., 2005b,
Development and validation of the global map of irrigation areas. Hydrology and
Earth System Sciences, 9, pp. 535–547.
SINHA, S., 2003, Rs 24,500 cr plan for harvesting rain water. Times of India, 6 January 2003.
SMITH, P.M., KALLURI, S.N.V., PRINCE, S.D. and DEFRIES, R.S., 1997, The NOAA/NASA
Pathfinder AVHRR 8 km land data set. Photogrammetric Engineering and Remote
Sensing, 63, pp. 12–31.
THENKABAIL, P.S., GANGADHARARAO, P., BIGGS, T., KRISHNA, M. and TURRAL, H., 2007a,
Spectral matching techniques to determine historical land use/land cover (LULC)
and irrigated areas using time series AVHRR Pathfinder Data sets in the Krishna
River Basin, India. Photogrammetric Engineering and Remote Sensing, 73, pp.
1029–1040.
THENKABAIL, P.S., BIRADAR, C.M., NOOJIPADY, P., CAI, X.L., DHEERAVATH, V., LI, Y.J.,
VELPURI, M., GUMMA, M. and PANDEY, S., 2007b, Sub-pixel irrigated area calculation
methods. Sensors Journal, Special Issue: Remote Sensing of Natural Resources and the
Environment, 7, pp. 2519–2538, Available online at: http://www.mdpi.org/sensors/
papers/s7112519.pdf.
THENKABAIL, P.S., SCHULL, M. and TURRAL, H., 2005, Ganges and Indus River Basin land
use/land cover (LULC) and irrigated area mapping using continuous streams of
MODIS data. Remote Sensing of Environment, 95, pp. 317–341.
Downloaded By: [US Geological Survey Library] At: 15:49 24 July 2009
Global irrigated area map
3733
THENKABAIL, P.S., ENCLONA, E.A., ASHTON, M.S. and VAN DER MEER, V., 2004a, Accuracy
assessments of hyperspectral waveband performance for vegetation analysis applications. Remote Sensing of Environment, 91, pp. 354–376.
THENKABAIL, P.S., ENCLONA, E.A., ASHTON, M.S., LEGG, C. and JEAN DE DIEU, M., 2004b,
Hyperion, IKONOS, ALI, and ETM + sensors in the study of African rain forests.
Remote Sensing of Environment, 90, pp. 23–44.
THENKABAIL, P.S., BIRADAR, C.M., TURRAL, H., NOOJIPADY, P., LI, Y.J., VITHANAGE, J.,
DHEERAVATH, V., VELPURI, M., SCHULL, M., CAI, X.L. and DUTTA, R., 2006, An
Irrigated Area Map of the World (1999) Derived From Remote Sensing, Research
report 105 (IWMI). Available online at: http://www.iwmigiam.org.
TOMLIN, D., 1990, Geographic Information Systems and Cartographic Modeling (New Jersey:
Prentice Hall College Division).
TOMLINSON, R., 2003, Thinking About Geographic Information Systems Planning for Managers
(California: ESRI Press).
TOU, J.T. and GONZALEZ, R.C., 1975, Pattern Recognition Principles (Reading,
Massachusetts: Addison-Wesley Publishing Company).
TUCKER, C.J., GRANT, D.M. and DYKSTRA, J.D., 2005, NASA’s global orthorectified Landsat
data set. Photogrammetric Engineering and Remote Sensing, 70, pp. 313–322.
US GEOLOGICAL SURVEY (USGS), 1993, Digital Elevation Models, Data User Guide 5 (Reston,
Virginia: USGS).
VAN SCHILFGAARDE, J., 1994, Irrigation – a blessing or a curse. Agricultural Water
Management, 25, pp. 203–219.
VELPURI, M., THENKABAIL, P.S., GUMMA, M.K., BIRADAR, C.B., DHEERAVATH, V.,
NOOJIPADY, P. and YUANJIE, L., 2009, Influence of resolution or scale in irrigated
area mapping and area estimations. Photogrammetric Engineering and Remote Sensing,
accepted, in press.
VERDIN, K. and JENSON, S., 1996, Development of continental scale DEMs and extraction of
hydrographic features. In Proceedings of the Third Conference on GIS and
Environmental Modeling (Santa Fe, New Mexico: NCGIA).
VERDIN, K.L. and GREENLEE, S.K., 1996, Development of continental scale digital elevation
models and extraction of hydrographic features. In Proceedings of Third International
Conference/Workshop on Integrating GIS and Environmental Modeling, Santa Fe, New
Mexico, 21–26 January 1996 (Santa Barbara, CA: National Center for Geographic
Information and Analysis).