6

Global Food Security Support

Analysis Data at Nominal
1 km (GFSAD1km) Derived
from Remote Sensing in
Support of Food Security in
the Twenty-First Century:
Current Achievements and
Future Possibilities

Pardhasaradhi Teluguntla
U.S. Geological Survey
and
Bay Area Environmental
Research Institute

Prasad S. Thenkabail
U.S. Geological Survey

Jun Xiong
U.S. Geological Survey
and
Northern Arizona University

Murali Krishna Gumma

International Crops Research
Institute for the Semi Arid Tropics

Chandra Giri Acronyms and Definitions .................................................................................................................132

U.S. Geological Survey,
(EROS) Center
6.1 Introduction .............................................................................................................................132
6.2 Global Distribution of Croplands and Other Land Use and Land Cover: Baseline
Cristina Milesi for the Year 2000 ..................................................................................................................... 134
NASA Ames Research Center Existing Global Cropland Maps: Remote Sensing and Non–Remote Sensing Approaches
6.3 Key Remote Sensing–Derived Cropland Products: Global Food Security .....................135
Mutlu Ozdogan 6.4 Definition of Remote Sensing–Based Cropland Mapping Products ...............................138
University of Wisconsin
6.5 Data: Remote Sensing and Other Data for Global Cropland Mapping ..........................138
Russell G. Congalton Primary Satellite Sensor Data • Secondary Data • Field-Plot Data • Very-High-Resolution
University of New Hampshire
Imagery Data • Data Composition: Mega File Data Cube (MFDC) Concept
6.6 Cropland Mapping Methods..................................................................................................142
James Tilton Remote Sensing–Based Cropland Mapping Methods for Global, Regional, and Local
NASA Goddard Space Flight Center Scales • Spectral Matching Techniques (SMTs) Algorithms

131

© 2016 Taylor & Francis Group, LLC

 132 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

Temuulen Tsagaan 6.7 Automated Cropland Classification Algorithm..................................................................142

Sankey 6.8 Remote Sensing–Based Global Cropland Products:
Northern Arizona University Current State-of-the-Art Maps, Their Strengths, and Limitations ..................................143
Global Cropland Extent at Nominal 1 km Resolution
Richard Massey 6.9 Change Analysis.......................................................................................................................148
Northern Arizona University 6.10 Uncertainties of Existing Cropland Products .....................................................................149
Aparna Phalke 6.11 Way Forward ........................................................................................................................... 150
University of Wisconsin 6.12 Conclusions...............................................................................................................................155
Acknowledgments .............................................................................................................................. 156
Kamini Yadav References ............................................................................................................................................ 156
University of New Hampshire

Acronyms and Definitions 6.1 Introduction

ACCA Automated cropland classification algorithm The precise estimation of the global agricultural cropland—
ASTER Advanced spaceborne thermal emission and reflec- extents, areas, geographic locations, crop types, cropping inten-
tion radiometer sities, and their watering methods (irrigated or rain-fed; type of
AVHRR Advanced very-high-resolution radiometer irrigation)—provides a critical scientific basis for the develop-
AWiFS Advanced wide field sensor ment of water and food security policies (Thenkabail et al., 2010,
CDL The Cropland Data Layer (CDL) was created by the 2011, 2012, Turral et al., 2009). By year 2100, the global human
USDA, National Agricultural Statistics Service population is expected to grow to 10.4 billion under median fer-
CEOS Committee on Earth Observing Satellites (CEOS) tility variants or higher under constant or higher fertility vari-
EDS Euclidean distance similarity ants (Table 6.1) with over three-quarters living in developing
FPA Full pixel areas countries and in regions that already lack the capacity to produce
GCAD Global cropland area database enough food. With current agricultural practices, the increased
GCE Global cropland extent demand for food and nutrition would require about 2 billion
GCE V1.0 Global cropland extent version 1.0 hectares of additional cropland, about twice the equivalent to the
GDEM ASTER-derived digital elevation data land area of the United States, and lead to significant increases
GEO Group on Earth Observations in greenhouse gas emissions (GHG) associated with agricultural
GEOSS Global Earth Observation System of Systems practices and activities (Tillman et al., 2011). For example, dur-
GFSAD Global food security support analysis data ing 1960–2010, world population more than doubled from 3 to
GIMMS Global Inventory Modeling and Mapping Studies 7 billion. The nutritional demand of the population also grew
JERS SAR Japanese Earth Resources Satellite-1 (JERS-1) swiftly during this period from an average of about 2000 calories
ISDB IA Ideal Spectra Data Bank on Irrigated Areas per day per person in 1960 to nearly 3000 calories per day per
LEDAPS Landsat Ecosystem Disturbance Adaptive Processing person in 2010. The food demand of increased population along
System with increased nutritional demand during this period was met by
MFDC Mega File Data Cube the “green revolution,” which more than tripled the food produc-
MODIS Moderate-resolution imaging spectroradiometer tion, even though croplands decreased from about 0.43 ha per
MSAS Modified spectral angle similarity capita to 0.26 ha per capita (FAO, 2009; Funk and Brown, 2009).
NASS National Agricultural Statistics Service of USDA The increase in food production during the green revolution
NDVI Normalized difference vegetation index was the result of factors such as: (1) expansion of irrigated crop-
NOAA National Oceanic and Atmospheric Administration lands, which had increased in 2000 from 130 Mha in the 1960s to
SAR Synthetic aperture radar between 278 Mha (Siebert et al., 2006) and 467 Mha (Thenkabail
SCS Spectral correlation similarity et al., 2009a,b,c), with the larger estimate due to consideration of
SIT Strategic Implementation Team cropping intensity; (2) increase in yield and per capita produc-
SMT Spectral matching techniques tion of food (e.g., cereal production from 280 to 380 kg/person
SPA Subpixel areas and meat from 22 to 34 kg/person (McIntyre, 2008); (3) new cul-
SPOT Système Pour l’Observation de la Terre tivar types (e.g., hybrid varieties of wheat and rice, biotechnol-
SSV Spectral similarity value ogy); and (4) modern agronomic and crop management practices
USDA United States Department of Agriculture (e.g., fertilizers, herbicide, pesticide applications).
USGS United States Geological Survey Although modern agriculture met the challenge to increase
VGT Vegetation sensor of SPOT satellite food production last century, lessons learned from the twenti-
VHRI Very-high-resolution imagery eth century “green revolution” and our current circumstances
VHRR Very-high-resolution radiometer impact the likelihood of another such revolution. The intensive

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 133

TABLE 6.1 World Population (Thousands) Under All Variants, for feeding the world (Kumar and Singh, 2005). Increasing per
1950–2100 capita meat consumption is increasing agricultural demands
Medium Constant on land and water (Vinnari and Tapio, 2009). Cropland areas
Fertility High Fertility Low Fertility Fertility are decreasing in many parts of the world due to urbanization,
Year Variant Variant Variant Variant industrialization, and salinization (Khan and Hanjra, 2008).
1950 2,529,346 2,529,346 2,529,346 2,529,346 Ecological and environmental imperatives, such as biodiver-
1955 2,763,453 2,763,453 2,763,453 2,763,453 sity conservation and atmospheric carbon sequestration, have
1960 3,023,358 3,023,358 3,023,358 3,023,358 put a cap on the possible expansion of cropland areas to other
1965 3,331,670 3,331,670 3,331,670 3,331,670 lands such as forests and rangelands (Gordon et al., 2009). Crop
1970 3,685,777 3,685,777 3,685,777 3,685,777 yield increases of the green revolution era have now stagnated
1975 4,061,317 4,061,317 4,061,317 4,061,317 (Hossain et al., 2005). Given these factors and limitations, fur-
1980 4,437,609 4,437,609 4,437,609 4,437,609 ther increase in food production through increase in cropland
1985 4,846,247 4,846,247 4,846,247 4,846,247 areas and/or increased allocations of water for croplands is
1990 5,290,452 5,290,452 5,290,452 5,290,452 widely considered unsustainable or simply infeasible.
1995 5,713,073 5,713,073 5,713,073 5,713,073 Clearly, our continued ability to sustain adequate global
2000 6,115,367 6,115,367 6,115,367 6,115,367 food production and achieve future food security in the
2005 6,512,276 6,512,276 6,512,276 6,512,276 twenty-first century is challenged. So, how does the world con-
2010 6,916,183 6,916,183 6,916,183 6,916,183 tinue to meet its food and nutrition needs? Solutions may come
2015 7,324,782 7,392,233 7,256,925 7,353,522 from biotechnology and precision farming. However, develop-
2020 7,716,749 7,893,904 7,539,163 7,809,497 ments in these fields are not currently moving at rates that will
2025 8,083,413 8,398,226 7,768,450 8,273,410 ensure global food security over the next few decades (Foley
2030 8,424,937 8,881,519 7,969,407 8,750,296 et al., 2011). Further, there is a need for careful consideration
2035 8,743,447 9,359,400 8,135,087 9,255,828 of possible adverse effects of biotechnology. We should not be
2040 9,038,687 9,847,909 8,255,351 9,806,383 looking back 30–50 years from now with regrets, like we are
2045 9,308,438 10,352,435 8,323,978 10,413,537 looking back now at many mistakes made during the green
2050 9,550,945 10,868,444 8,341,706 11,089,178 revolution. During the green revolution, the focus was only on
2055 9,766,475 11,388,551 8,314,597 11,852,474 getting more yield per unit area. Little thought was given to
2060 9,957,399 11,911,465 8,248,967 12,729,809 the serious damage done to our natural environments, water
2065 10,127,007 12,442,757 8,149,085 13,752,494 resources, and human health as a result of detrimental factors
2070 10,277,339 12,989,484 8,016,514 14,953,882 such as uncontrolled use of herbicides, pesticides, and nutri-
2075 10,305,146 13,101,094 7,986,122 15,218,723 ents, drastic groundwater mining, and salinization of fertile
2080 10,332,223 13,213,515 7,954,481 15,492,520 soils due to overirrigation. Currently, there are discussions
2085 10,358,578 13,326,745 7,921,618 15,775,624 of a “second green revolution” or even an “evergreen revolu-
2090 10,384,216 13,440,773 7,887,560 16,068,398 tion,” but definitions of what these terms actually mean are still
2095 10,409,149 13,555,593 7,852,342 16,371,225 debated and are evolving (e.g., Monfreda et al., 2008). One of
2100 10,433,385 13,671,202 7,815,996 16,684,501 the biggest issues that has not been given adequate focus is the
Source: UNDP, Human Development Report 2012: Overcoming Barriers: use of large quantities of water for food production. Indeed,
Human Mobility and Development, New York, United Nations, 2012. an overwhelming proportion (60%–90%) of all human water
use in the World, for example, goes for producing their food
use of chemicals has adversely impacted the environment in (Falkenmark and Rockström, 2006). But such intensive water
many regions, leading to salinization and decreasing water qual- use for food production is no longer sustainable due to increas-
ity and degrading croplands. From 1960 to 2000, worldwide ing competition for water in alternative uses (EPW, 2008), such
phosphorous use doubled from 10 million tons (MT) to 20 MT, as urbanization, industrialization, environmental flows, biofu-
pesticide use tripled from near zero to 3 MT, and nitrogen use as els, and recreation. This has brought into sharp focus the need
fertilizer increased to a staggering 80 MT from just 10 MT (Foley to grow more food per drop of water (or crop water productiv-
et al., 2007; Khan and Hanjra, 2008). Diversion of croplands to ity or crop per drop) leading to the need for a “blue revolution”
biofuels is taking water away from food production (Bindraban in agriculture (Pennisi, 2008).
et al., 2009), even as the economic, carbon sequestration, envi- A significant part of the solution lies in determining how global
ronmental, and food security impacts of biofuel production are croplands are currently used and how they might be better managed
proving to be a net negative (Gibbs et al., 2008; Lal and Pimentel, to optimize the use of resources in food production. This will require
2009; Searchinger et al., 2008). Climate models predict that the development of an advanced global cropland area database (GCAD)
hottest seasons on record will become the norm by the end of the with an ability to map global croplands and their attributes routinely,
century in most regions of the world—a prediction that bodes ill rapidly, consistently, and with sufficient accuracies. This in turn

© 2016 Taylor & Francis Group, LLC

 134 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

requires the creation of a framework of best practices for cropland with other land use and land cover classes (Figure 6.1). This
mapping and an advanced global geospatial information system on provides a first view of where global croplands are concentrated
global croplands. Such a system would need to be consistent across and helps us to focus on the appropriate geographic locations
nations and regions by providing information on issues such as the for detailed cropland studies. Water and snow (Class 8 and 9,
composition and location of cropping, cropping intensities (e.g., sin- respectively) have zero croplands and occupy 44% of the total
gle, double crop), rotations, crop health/vigor, and irrigation status. terrestrial land surface. Further, forests (Class 6) occupy 17% of
Opportunities to establish such a global system can be achieved by the terrestrial area and deserts (Class 7) an additional 12%. In
fusing advanced remote sensing data from multiple platforms and these two classes, <5% of the total croplands exist. Therefore,
agencies (e.g., http://eros.usgs.gov/ceos/satellites_midres1.shtml; in order to study croplands systematically and intensively, one
http://www.ceos-cove.org/index.php) in combination with national must prioritize mapping in the areas of Classes 1–5 (26% of the
statistics, secondary data (e.g., elevation, slope, soils, temperature, terrestrial area) where >95% of all global croplands exist, with
and precipitation), and the systematic collection of field level obser- the first 3 classes (Class 1, 2, and 3) having ∼75% and the next 2
vations. An example of such a system on a regional scale is USDA, ∼20%. In the future, it is likely some of the noncroplands may
NASS Cropland Data Layer (CDL), which is a raster, georeferenced, be converted to croplands (e.g., especially in Africa where large
crop-specific land cover data layer with a ground resolution of 30 m farmlands are introduced in recent years in otherwise over-
(Johnson and Mueller, 2010). The GCAD will be a major contribu- whelmingly small-holder dominant farming) or vice versa,
tion to Group on Earth Observations (GEO) Global Agricultural highlighting the need for repeated and systematic global map-
Monitoring Initiative (GLAM), to the overarching vision of GEO ping of croplands. Segmenting the world into cropland versus
Agriculture and Water Societal Beneficial Areas (GEO Ag. SBAs), noncropland areas routinely will help us understand and study
G20 Agriculture Ministers initiatives, and ultimately to the Global these change dynamics better.
Earth Observation System of Systems (GEOSS). These initiatives
are also supported by the Committee on Earth Observing Satellites
6.2.1 Existing Global Cropland Maps:
(CEOS) Strategic Implementation Team (SIT).
Remote Sensing and Non–Remote
Within the context of the above facts, the overarching
Sensing Approaches
goal of this chapter is to provide a comprehensive overview
of the state-of-art of global cropland mapping procedures There are currently six major global cropland maps:
using remote sensing as characterized and envisioned by (1) Thenkabail et al. (2009a,b), (2) Ramankutty and Foley
the “Global Food Security Support Analysis Data @ 30 m (1998), (3) Goldewijk et al. (2011), (4) Portmann et al. (2010),
(GFSAD30)” project working group team. First, the chapter (5) Pittman et al. (2010), and (6) Yu et al. (2013). These studies
will provide an overview of existing cropland maps and their estimated the total global cropland area to be around 1.5 to 1.7
characteristics along with establishing the gaps in knowl- billion hectares for the year 2000 as a baseline. However, there
edge related to global cropland mapping. Second, definitions are two significant differences in these products: (1) spatial dis-
of cropland mapping along with key parameters involved in agreement on where the actual croplands are, and (2) irrigated
cropland mapping based on their importance in food security to rain-fed cropland proportions and their precise spatial loca-
analysis, and cropland naming conventions for standardized tions. Globally, cropland areas have increased from around 265
cropland mapping using remote sensing will be presented. Mha in year 1700 to around 1471 Mha in year 1990, while the
Third, existing methods and approaches for cropland mapping area of pasture has increased approximately sixfold from 524 to
will be discussed. This will include the type of remote sens- 3451 Mha (Foley et al., 2011). Ramankutty and Foley (1998) esti-
ing data used in cropland mapping and their characteristics mated the cropland and pasture to represent about 36% of the
along with discussions on the secondary data, field-plot data, world’s terrestrial surface (148,940,000 km2), of which, accord-
and cropland mapping algorithms. Fourth, currently existing ing to different studies, roughly 12% is croplands and 24% pas-
global cropland products derived using remote sensing will ture. Multiple studies (Goldewijk et al., 2011; Portmann et al.,
be presented and discussed. Fifth, a synthesis of all existing 2010; Ramankutty et al., 2008) integrated agricultural statistics
products leading to a composite global cropland extent ver- and census data from the national systems with spatial mapping
sion 1.0 (GCE V1.0) is presented and discussed. Sixth, a way technologies involving geographic information systems (GIS) to
forward for advanced global cropland mapping is visualized. derive global cropland maps.
Thenkabail and others (2009a,b, 2011) produced the first remote
sensing–based global irrigated and rain-fed cropland maps and
6.2 Global Distribution of Croplands statistics through multisensor remote sensing data fusion along
and Other Land Use and Land with secondary data and in situ data. They further used five domi-
Cover: Baseline for the Year 2000 nant crop types (wheat, rice, corn, barley, and soybeans) using par-
cel-based inventory data (Monfreda et al., 2005, 2008; Portmann
The first comprehensive global map of croplands was created by et al., 2010; Ramankutty et al., 2008) to produce a classification of
Ramankutty et al. in 1998. A more current version for the year global croplands with crop dominance (Thenkabail et al., 2012).
2000 shows the spatial distribution of global croplands along The five crops account for about 60% of the total global cropland

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 135

180°0΄0˝ 120°0΄0˝W 60°0΄0˝W 0°0΄0˝ 60°0΄0˝E 120°0΄0˝E 180°0΄0˝

60°0΄0˝N 60°0΄0˝N

30°0΄0˝N 30°0΄0˝N

0°0΄0˝ 0°0΄0˝

N
30°0΄0˝S 30°0΄0˝S

km
60°0΄0˝S 0 3,000 6,000 12,000 60°0΄0˝S

180°0΄0˝ 120°0΄0˝W 60°0΄0˝W 0°0΄0˝ 60°0΄0˝E 120°0΄0˝E 180°0΄0˝

05 Savanna, grassland, and shrublands (low % cropland) (7%)

01 Irrigated croplands (2%) 06 Forest (very low % cropland) (17%)

02 Rainfed croplands (7%) 07 Barren lands, deserts (very low % cropland) (12%)

03 Rainfed fragments with grassland and shrubland (5%) 08 Snow, Ice, and tundra (no cropland) (41%)

04 Rainfed fragments with woodland and forest (5%) 09 Water body (no cropland) (3%)

FIGURE 6.1 Global croplands and other land use and land cover: Baseline.

areas. The precise spatial location of these crops is only an approx- (2) watering methods (e.g., irrigated, supplemental irrigated, and
imation due to the coarse resolution (approximately 1 km 2) and rain-fed), (3) crop types, and (4) cropping intensities (e.g., single
fractional representation (1%–100% crop in a pixel) of the crop crop, double crop, and continuous crop). Although not the focus
data in each grid cell of all the maps from which this compos- of this chapter, many other parameters are also derived in local
ite map is produced (Thenkabail et al., 2012). The existing global regions, such as: (5) precise location of crops, (6) cropping calen-
cropland datasets also differ from each other due to inherent dar, (7) crop health/vigor, (8) flood and drought information, (9)
uncertainties in establishing the precise location of croplands, the water use assessments, and (10) yield or productivity (expressed per
watering methods (rain-fed versus irrigated), cropping intensities, unit of land and/or unit of water). Remote sensing is specifically
crop types and/or dominance, and crop characteristics (e.g., crop suited to derive the four key products over large areas using fusion
or water productivity measures such as biomass, yield, and water of advanced remote sensing (e.g., Landsat, Resourcesat, MODIS)
use). Improved knowledge of the uncertainties (Congalton and in combination with national statistics, ancillary data (e.g., eleva-
Green, 2009) in these estimates will lead to a suite of highly accu- tion, precipitation), and field-plot data. Such a system, at the global
rate spatial data products (Goodchild and Gopal, 1989) in support level, will be complex in data handling and processing and requires
of crop modeling, food security analysis, and decision support. coordination between multiple agencies leading to development of
a seamless, scalable, transparent, and repeatable methodology. As a
result, it is important to have a systematic class labeling convention
6.3 Key Remote Sensing–Derived as illustrated in Figure 6.3. A standardized class identifying and
Cropland Products: Global labeling process (Figure 6.3) will enable consistent and systematic
Food Security labeling of classes, irrespective of analysts. First, the area is sepa-
rated into cropland versus noncropland. Then, within the cropland
The production of a repeatable global cropland product requires a class, labeling will involve (Figure 6.3): (1) cropland extent (crop-
standard set of metrics and attributes that can be derived consis- land versus noncropland), (2) watering source (e.g., irrigated versus
tently across the diverse cropland regions of the world. Four key rain-fed), (3) irrigation source (e.g., surface water, ground water),
cropland information systems attributes that have been identified (4) crop type or dominance, (5) scale (e.g., large or contiguous,
for global food security analysis and that can be readily derived small or fragmented), and (6) cropping intensity (e.g., single crop,
from remote sensing include (Figure 6.2): (1) cropland extent/areas, double crop). The detail at which one maps at each stage and each

© 2016 Taylor & Francis Group, LLC

 136 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

1. Global cropland extent/area

@ nominal 30 m through landsat
30 m (Gutman et al, 2008) + MODIS 250 m +
secondary data fusion
2010

180°0΄0˝ 120°0΄0˝W 60°0΄0˝W

1990
0°0΄0˝ 60°0΄0˝E 120°0΄0˝E

60°0΄0˝N 60°0΄0˝N

30°0΄0˝N 30°0΄0˝N

0°0΄0˝ 0°0΄0˝

30°0΄0˝S 30°0΄0˝S
01 Croplands

02 Non-croplands
60°0΄0˝S 60°0΄0˝S

90°0΄0˝ 90°0΄0˝
120°0΄0˝W 60°0΄0˝W 0°0΄0˝ 60°0΄0˝E 120°0΄0˝E

2. Crop type
3. Irrigated versus rainfed
Eight major crops + others
180°0΄0˝ 120°0΄0˝W 60°0΄0˝W 0°0΄0˝ 60°0΄0˝E 120°0΄0˝E

60°0΄0˝N 60°0΄0˝N
Crop Type Crop Area (ha) Proportion (%)
30°0΄0˝N 30°0΄0˝N
Wheat 402,800,000 22
0°0΄0˝ 0°0΄0˝ Corn 227,100,000 13
30°0΄0˝S
01 Irrigated
02 Rainfed
30°0΄0˝S Rice 195,600,000 11
03 Non-croplands
60°0΄0˝S 60°0΄0˝S
Barley 158,000,000 9
90°0΄0˝ 90°0΄0˝
120°0΄0˝W 60°0΄0˝W 0°0΄0˝ 60°0΄0˝E 120°0΄0˝E Soybeans 92,700,000 5
Pulses 79,400,000 4
Cotton 53,400,000 3
Potatoes 50,100,000 3

1 Irrigated, double crop

MODIS NDVI

0.8 Irrigated, triple crop

0.6
0.6
MODIS NDVI

0.5
0.4 0.4
0.3
0.2
0.2
0 0.1
1 2 3 4 5 6 7 8 9 10 11 12
0
Month
1 2 3 4 5 6 7 8 9 10 11 12
Month
1 Irrigated, continuous crop
4. Cropping intensities
MODIS NDVI

0.8
180°0΄0˝ 120°0΄0˝W 60°0΄0˝W 0°0΄0˝ 60°0΄0˝E 120°0΄0˝E 0.8 Irrigated, double crop
0.6
MODIS NDVI

0.6
0.4 60°0΄0˝N 60°0΄0˝N

30°0΄0˝N 0.4
0.2 30°0΄0˝N

0 0°0΄0˝ 0°0΄0˝ 0.2

1 2 3 4 5 6 7 8 9 10 11 12
01 Irrigated
Month 30°0΄0˝S
02 Rainfed
30°0΄0˝S
0
1 2 3 4 5 6 7 8 9 10 11 12
60°0΄0˝S 03 Non-croplands 60°0΄0˝S
Month
90°0΄0˝ 90°0΄0˝
120°0΄0˝W 60°0΄0˝W 0°0΄0˝ 60°0΄0˝E 120°0΄0˝E

Rainfed, single crop

0.8 Rainfed, single crop 0.5
Rainfed, single crop
MODIS NDVI

0.8 0.4
MODIS NDVI

MODIS NDVI

0.6 0.6 0.3

0.2
0.4 0.4
0.1
0.2 0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12
0 Month
1 2 3 4 5 6 7 8 9 10 11 12
1 2 3 4 5 6 7 8 9 10 11 12
Month Month

FIGURE 6.2 Key global cropland area products that will support food security analysis in the twenty-first century.

© 2016 Taylor & Francis Group, LLC

© 2016 Taylor & Francis Group, LLC

Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km)

Delta
Level I: Watering method + Irrigation type + Crop type* + Scale + Intensity + SSD +

Irrigated Rainfed LULC e.g., Rice, Wheat, Maize or others* Single Double For example,
crop crop Type1..,v

SW GW Conjunctive use Supplemental Large scale Small scale Continuous

crop
Fragment

Level II: Watering method + Irrigation type + Crop type* + LS, SC or SS SC or LS, DC or SS, SC

Irrigated Rainfed LULC e.g., Rice, Wheat, Maize, rice-wheat or others*

Level VI: Watering method
SW GW Conjunctive use Supplemental

Irrigated LULC
Level III: Watering method + Irrigation type + Crop type*

Rainfed

Irrigated Rainfed LULC e.g., Rice, Wheat, Maize, rice-wheat or others* Abrreviations
SW: Surface water
SW GW Conjunctive use Supplemental GW: Ground water
LULC: Land use, land cover
LS: Large scale
SS: Small scale
Level IV and V: Watering method + Irrigation type SC: Single scale
DC: Double scale
CC: Continuous
SSD: Spectral signature difference
Irrigated Rainfed LULC
*: others = mention crop type and/or crop
dominant
SW GW Conjunctive use Supplemental

FIGURE 6.3 Cropland class naming convention at different levels. Level I is most detailed and Level IV is least detailed.

137
 138 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

parameter would depend on many factors such as resolution of the 6.5 Data: Remote Sensing and Other
imagery, available ground data, and expert knowledge. For exam-
ple, if there is no sufficient knowledge on whether the irrigation is
Data for Global Cropland Mapping
by surface water or ground water, but it is clear that the area is irri- Cropland mapping using remote sensing involves multiple types
gated, one could just map it as irrigated without mapping greater of data: satellite data with a consistent and useful global repeat
details on the type of irrigation. But, for every cropland class, one cycle, secondary data, statistical data, and field plot data. When
has the potential to map the details as shown in Figure 6.3. these data are used in an integrated fashion, the output products
achieve highest possible accuracies (Thenkabail et al., 2009b,c).
6.4 Definition of Remote Sensing–
Based Cropland Mapping Products 6.5.1 Primary Satellite Sensor Data
Key to effective mapping is a precise and clear definition of what Cropland mapping will require satellite sensor data across spa-
will be mapped. It is the first and primary step, with different tial, spectral, radiometric, and temporal resolutions from a wide
definitions leading to different products. For example, irrigated array of satellite/sensor platforms (Table 6.2) throughout the
areas are defined and understood differently in different appli- growing season. These satellite sensors are “representative” of
cations and contexts. One can define them as areas that receive hyperspectral, multispectral, and hyperspatial data. The data
irrigation at least once during their crop growing period. points per hectare (Table 6.2, last column) will indicate the spa-
Alternatively, they can be defined as areas that receive irriga- tial detail of agricultural information gathered. In addition to
tion to meet at least half their crop water requirements during satellite-based sensors, it is always valuable to gather ground-
the growing season. One other definition can be that these are based hand-held spectroradiometer data from hyperspectral
areas that are irrigated throughout the growing season. In each sensors (Thenkabail et al., 2013), and/or imaging spectroscopy
of these cases, the extent of irrigated area mapped will vary. from ground-based, airborne, or space borne sensors for vali-
Similarly, croplands can be defined as all agricultural areas irre- dation and calibration purposes (Thenkabail et al., 2011). Much
spective of the types of crops grown or they may be limited to greater details of a wide array of sensors available to gather
food crops (and not the fodder crops or plantation crops). So, it data are presented in Chapters 1 and 2 of Remotely Sensed Data
is obvious that having a clear understanding of the definitions Characterization, Classification, and Accuracies.
of what we map is extremely important for the integrity of the
products developed. We defined cropland products as follows:
6.5.2 Secondary Data
• Minimum mapping unit: The minimum mapping unit of a
particular crop is an area of 3 by 3 (0.81 ha) Landsat pixels There is a wide array of secondary or ancillary data such as the
identified as having the same crop type. ASTER-derived digital elevation data (GDEM), long (50–100
• Cropland extent: All cultivated plants harvested for food, years) records of precipitation and temperature (CRU), digital
feed, and fiber, including plantations (e.g., orchards, vine- maps of soil types, and administrative boundaries. Many sec-
yards, coffee, tea, rubber). ondary data are known to improve crop classification accuracies
• What is a cropland pixel?: sub-pixel composition is used to (Thenkabail et al., 2009a,b). The secondary data will also form
calculate area. This involves multiplying full pixel area (FPA) core data for the spatial decision support system and final visu-
with cropland area fraction (CAF). CAF provides what % of alization tool in many systems.
pixel is cropped. So, sub-pixel area/actual area = FPA*CAF
• Irrigated areas: Irrigation is defined as artificial applica-
6.5.3 Field-Plot Data
tion of any amount of water to overcome crop water stress.
Irrigated areas are those areas that are irrigated one or Field-plot data (e.g., Figure 6.4) will be used for purposes such
more times during crop growing season. as: (1) class identification and labeling; (2) determining irrigated
• Rain-fed areas: Areas that have no irrigation whatsoever area fractions (AFs), and (3) establishing accuracies, errors, and
and are precipitation dependent. uncertainties. At each field point (e.g., Figure 6.3), data such as
• Cropping intensity: Number of cropping cycles within a cropland or noncropland, watering method (irrigated or rain-
12-month period. fed), crop type, and cropping intensities are recorded along
• Crop type: Eight crops (wheat, corn, rice, barley, soybeans, with GPS locations, digital photographs, and other information
pulses, cotton, and potatoes), that occupy approx. 70% (e.g., yield, soil type) as needed. Field plot data will also help in
global cropland areas are considered. The rest of the crops gathering an ideal spectral data bank of croplands. One could
are under “others”. However, in particular continents where use the precise locations and the crop characteristics and gener-
other crops like sugarcane or cassava etc. are important, ate coincident remote sensing data characteristics (e.g., MODIS
they will be mapped as well. time-series monthly NDVI).

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 139

TABLE 6.2 Characteristics of Some of the Key Satellite Sensor Data Currently Used in Cropland Mapping
Wavelength Spatial Spectral Temporal Radiometric
Satellite Sensor Range (μm) Resolution (m) Bands (#) (days) (bits) Data Points (per ha)
A. Hyperspectral
EO-1 Hyperion 196 16 16 11.1 points for 30 m pixel
VNIR 0.43–0.93 30 (0.09 ha per pixel)
SWIR 0.93–2.40 30
B. Advanced multispectral
Landsat TM 7/8 16 8
Multispectral
Band 1 0.45–0.52 30 44.4 points for 15 m pixel
Band 2 0.53–0.61 30 11.1 points for 30 m pixel
Band 3 0.63–0.69 30 2.77 points for 60 m pixel
Band 4 0.78–0.90 30 0.69 points for 120 m pixel
Band 5 1.55–1.75 30
Band 6 10.40–12.50 120/60
Band 7 2.09–2.35 30
Panchromatic 0.52–0.90 15
EO-1 ALI 10 16 16
Multispectral
Band 1 0.43–0.45 30
Band 2 0.45–0.52 30
Band 3 0.52–0.61 30
Band 4 0.63–0.69 30
Band 5 0.78–0.81 30
Band 6 0.85–0.89 30
Band 7 1.20–1.30 30
Band 8 1.55–1.75 30
Band 9 2.08–2.35 30
Panchromatic 0.48–0.69 10
ASTER 14 16 8
VNIR 15
Band 1 0.52–0.60
Band 2 0.63–0.69
Band 3N/3B 0.76–0.86
SWIR 30
Band 4 1.600–1.700
Band 5 2.145–2.185
Band 6 2.185–2.225
Band 7 2.235–2.285
Band 8 2.295–2.365
Band 9 2.360–2.430
TIR 90 1.23 points for 90 m
Band 10 8.125–8.475
Band 11 8.475–8.825
Band 12 8.925–9.275
Band 13 10.25–10.95
Band 14 10.95–11.65
MODIS
MOD09Q1 250 2 1 12 0.16 points for 250 m
Band 1 0.62–0.67
Band 2 0.84–0.876
(Continued )

© 2016 Taylor & Francis Group, LLC

 140 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

TABLE 6.2 (Continued ) Characteristics of Some of the Key Satellite Sensor Data Currently Used in Cropland Mapping
Wavelength Spatial Spectral Temporal Radiometric
Satellite Sensor Range (μm) Resolution (m) Bands (#) (days) (bits) Data Points (per ha)
MOD09A1 500 7a/36 1 12 0.04 points for 500 m
Band 1 0.62–0.67
Band 2 0.84–0.876
Band 3 0.459–0.479
Band 4 0.545–0.565
Band 5 1.23–1.25
Band 6 1.63–1.65
Band 7 2.11–2.16
C. Hyperspatial
GeoEye-1
Multispectral 1.65 5 <3 11
Band 1 0.45–0.52 59,488 points for 0.41 m
Band 2 0.52–0.60 26,874 points for 0.61 m
Band 3 0.63–0.70 10,000 points for 1 m
Band 4 0.76–0.90 3673 points for 1.65 m
Panchromatic 0.45–0.90 0.41 1679 points for 2.44 m
IKONOS 5 3 11
Multispectral 4
Band 1 0.45–0.52 625 points for 4 m
Band 2 0.51–0.60 400 points for 5 m
Band 3 0.63–0.70 236 points for 6.5 m
Band 4 0.76–0.85 100 points for 10 m
Panchromatic 0.53–0.93 1 44.4 points for 15 m
QuickBird 5 1–6 11
Multispectral 2.44
Band 1 0.45–0.52
Band 2 0.52–0.60
Band 3 0.63–0.69
Band 4 0.76–0.90
Panchromatic 0.45–0.90 0.61
RapidEye 5–6.5 5 1–6 16
Band 1 0.44–0.51
Band 2 0.52–0.59
Band 3 0.63–0.68
Band 4 0.69–0.73
Band 5 0.76–0.85
a MODIS has 36 bands, but we considered only the first 7 bands (Mod09A1).

6.5.4 Very-High-Resolution Imagery Data 6.5.5 Data Composition: Mega File

Data Cube (MFDC) Concept
Very-high-resolution (submeter to 5 m) imagery (VHRI; see
hyperspatial data characteristics in Table 6.2) is widely avail- Data preprocessing requires that all the acquired imagery is
able these days from numerous sources. These data can be harmonized and standardized in known time intervals (e.g.,
used as ground samples in localized areas to classify as well monthly, biweekly). For this, the imagery data is either acquired
as verify classification results of the coarser resolution imag- or converted to at-sensor reflectance (see Chander et al., 2009;
ery. For example, in Figure 6.5, VHRI tiles identify uncertain- Thenkabail et al., 2004) and then converted to surface reflec-
ties existing in cropland classification of coarser resolution tance using Landsat Ecosystem Disturbance Adaptive Processing
imagery. VHRI is specifically useful for identifying croplands System (LEDAPS) codes for Landsat (Masek et al., 2006) or similar
versus noncroplands (Figure 6.5). They can also be used for codes for other sensors. All data are processed and mosaicked to
identifying irrigation based on associated features such as required geographic levels (e.g., global, continental). One method
canals and tanks. to organize these disparate but colocated datasets is through the

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 141

Ground reference data points (Global collection: Total 125,796 points)

N
150°0΄0˝W 120°0΄0˝W90°0΄0˝W 60°0΄0˝W 30°0΄0˝W 0°0΄0˝ 30°0΄0˝E 60°0΄0˝E 90°0΄0˝E 120°0΄0˝E 150°0΄0˝E W E
90°0΄0˝ 90°0΄0˝S

60°0΄0˝N 60°0΄0˝N

30°0΄0˝N 30°0΄0˝N
Total points : 125,796
Data source:
0°0΄0˝ 0°0΄0˝
IWMI–collection (678)
Gumma et al.–collection (3,561)
Thenkabail et al.–collection (1,312)
30°0΄0˝S 30°0΄0˝S
Degree–confluence–Pts (973)
Geo–wiki–validation (11,453)
60°0΄0˝S Geo–wiki–competition (30,359) 60°0΄0˝S
USAID–Project (77,480)

90°0΄0˝ 90°0΄0˝
150°0΄0˝W120°0΄0˝W90°0΄0˝W 60°0΄0˝W 30°0΄0˝W 0°0΄0˝ 30°0΄0˝E 60°0΄0˝E 90°0΄0˝E 120°0΄0˝E 150°0΄0˝E

FIGURE 6.4 Field plot data for cropland studies collected over the globe.

Crop Crop No crop No crop

Crop Crop No crop No crop

FIGURE 6.5 Very-high-resolution imagery used to resolve uncertainties in cropland mapping of Australia.

use of a MFDC. Numerous secondary datasets are combined in distinctive segments of MFDCs and analyzing them separately for
an MFDC, which is then stratified using image segmentation into croplands will enhance accuracy. For example, the likelihood of
distinct precipitation-elevation-temperature-vegetation zones. croplands in a temperature zone of <280°K is very low. Similarly,
Data within the MFDC can include ASTER-derived refined digital croplands in elevation above 1500 m will be of distinctive charac-
elevation from SRTM (GDEM), monthly long-term precipitation, teristics (e.g., patchy, on hilly terrain most likely plantations of cof-
monthly thermal skin temperature, and forest cover and density. fee or tea). Every layer of data is geolinked (having precisely same
This segmentation allows cropland mapping to be focused; creating projection and datum and are georeferenced to one another).

© 2016 Taylor & Francis Group, LLC

 142 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

The purpose of MFDC (MFDC; see Thenkabail et al., 2009b distance similarity (EDS)—a distance measure; and (4) modified
for details) is to ensure numerous remote sensing and second- spectral angle similarity (MSAS)—a hyperangle measure.
ary data layers are all stacked one over the other to form a
6.6.2.1 Generating Class Spectra
data cube akin to hyperspectral data cube. This approach has
been used by X to map croplands in Y (reference). The MFDC The MFDC (Section 6.4.5) of each of segment (Figures 6.6 and 6.7a)
allows us to have the entire data stack for any geographic loca- is processed using ISOCLASS K-means classification to produce
tion (global to local) as a single file available for analysis. For a large number of class spectra with a unsupervised classification
example, one can classify 10s or 100s or even 1000s of data layers technique that are then interpreted and labeled. In more localized
(e.g., monthly MODIS NDVI time series data for a geographic applications, it is common to undertake a field-plot data collection
area for an entire decade along with secondary data of the same to identify and label class spectra. However, at the global scale, this
area) stacked together in a single file and classify the image. The is not possible due to the enormous resources required to cover vast
classes coming out of such a MFDC inform us about the phenol- areas to identify and label classes. Therefore, SMTs (Thenkabail
ogy along with other characteristics of the crop. et al., 2007a) to match similar classes or to match class spectra from
the unsupervised classification with a library of ideal or target spec-
tra (e.g., Figure 6.6a) will be used to identify and label the classes.
6.6 Cropland Mapping Methods
6.6.2.2 Creating Ideal Spectra Data Bank (ISDB)
6.6.1 Remote Sensing–Based Cropland The term “ideal or target” spectra refers to time-series spectral
Mapping Methods for Global, reflectivity or NDVI generated for classes for which we have pre-
Regional, and Local Scales cise location-specific ground knowledge. From these locations,
There is a growing literature on cropland mapping across resolutions signatures are extracted using MFDC, synthesized, and aggre-
for both irrigated and rain-fed crops (Friedl et al., 2002; Gumma gated to generate a few hundred signatures that will constitute
et al., 2011; Hansen et al., 2002; Kurz and Seelan, 2007; Loveland an ISDB (e.g., Figures 6.6 and 6.7a).
et al., 2000; Olofsson et al., 2011; Ozdogan and Woodcock, 2006;
6.6.2.3 Matching Class Spectra with Ideal Spectra
Thenkabail et al., 2009a,c; Wardlow and Egbert, 2008; Wardlow
Using Spectral Matching Techniques (SMTs)
et al., 2006, 2007). Based on these studies, an ensemble of meth-
ods that is considered most efficient include: (1) spectral matching Once the class spectra are generated, they are compared with
techniques (SMTs) (Thenkabail et al., 2007a, 2009a,c); (2) decision ideal spectra to match, identify, and label classes. Often quan-
tree algorithms (DeFries et al., 1998); (3) Tassel cap brightness- titative spectral matching techniques like spectral correlation
greenness-wetness (Cohen and Goward, 2004; Crist and Cicone, similarity R-square (SCS R-square) and spectral similarity value
1984; Masek et al., 2008); (4) space-time spiral curves and change (SSV) are used (Thenkabail et al., 2007a).
vector analysis (Thenkabail et al., 2005); (5) phenology (Loveland
et al., 2000; Wardlow et al., 2006); and (6) climate data fusion with 6.7 Automated Cropland
MODIS time-series spectral indices using decision tree algorithms
and subpixel classification (Ozdogan and Gutman, 2008). More
Classification Algorithm
recently, cropland mapping algorithms that analyze end-member The first part of the automated cropland classification algorithm
spectra have been used for global mapping by Thenkabail et al. (ACCA) method involves knowledge capture to understand and
(2009a, 2011). map agricultural cropland dynamics by: (1) identifying croplands
versus noncroplands and crop type/dominance based on SMTs,
decision trees tassel cap bispectral plots, and very-high-resolution
6.6.2 Spectral Matching Techniques
imagery; (2) determining watering method (e.g., irrigated or rain-
(SMTs) Algorithms
fed) based on temporal characteristics (e.g., NDVI), crop water
SMTs (Thenkabail et al., 2007a, 2009a, 2011) are innovative methods requirement (water use by crops), secondary data (elevation, pre-
of identifying and labeling classes (see illustration in Figures 6.6 and cipitation, temperature), and irrigation structure (e.g., canals and
6.7a). For each derived class, this method identifies its characteris- wells); (3) establishing croplands that are large scale (i.e., contigu-
tics over time using MODIS time-series data (e.g., Figure 6.6). NDVI ous) versus small scale (i.e., fragmented); (4) characterizing crop-
time-series or other metrics (Biggs et al., 2006; Dheeravath et al., ping intensities (single, double, triple, and continuous cropping);
2010; Thenkabail et al., 2005, 2007a) are analogous to spectra, where (5) interpreting MODIS NDVI temporal bispectral plots to identify
time is substituted for wavelength. The principle in SMT is to match and label classes; and (6) using in situ data from very-high-resolution
the shape, or the magnitude or both to an ideal or target spectrum imagery, field-plot data, and national statistics (see Figure 6.7b for
(pure class or “end-member”). The spectra at each pixel to be clas- details). The second part of the method establishes accuracy of the
sified is compared to the end-member spectra and the fit is quanti- knowledge-captured agricultural map (Congalton, 1991 and 2009)
fied using the following SMTs (Thenkabail et al., 2007a): (1) spectral and statistics by comparison with national statistics, field-plot data,
correlation similarity (SCS)—a shape measure; (2) spectral similar- and very-high-resolution imagery. The third part of the method
ity value (SSV)—a shape and magnitude measure; (3) Eucledian makes use of the captured knowledge to code and map cropland

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 143

1.0 1.0

0.8 0.8

0.6 0.6

NDVI
NDVI

0.4 0.4

0.2 0.2
Ideal spectral signature for Irrigated-SW-rice-DC Similar class spectra signatures
0.0
0.0
Jun-00 Sep-00 Dec-00 Mar-01 Jun-00 Sep-00 Dec-00 Mar-01
Date
Date
15 16 17 18 32 33 34 19
(a) Irrigated-SW-rice-DC (b)

1.0 1.0

0.8 0.8

0.6 0.6
NDVI

NDVI
0.4 0.4

0.2 0.2
Ideal spectral signature match with similar
class spectra Grouping similar classes
0.0 0.0
Jun-00 Sep-00 Dec-00 Mar-01 Jun-00 Sep-00 Dec-00 Mar-01
Date
15 16 17 Date
18 32 33
34 19 Irrigated-SW-rice-DC 17 33 Irrigated-SW-rice-DC
(c) (d)

FIGURE 6.6 SMT. In SMTs, the class temporal profile (NDVI curves) are matched with the ideal temporal profile (quantitatively based on tem-
poral profile similarity values) in order to group and identify classes as illustrated for a rice class in this figure. (a) Ideal temporal profile illustrated
for “irrigated- surface-water-rice-double crop”; (b) some of the class temporal profile signatures that are similar; (c) ideal temporal profile signature
(Figure 6.6a) matched with class temporal profiles (Figure 6.6b); and (d) the ideal temporal profile (Figure 6.6a, in deep green) matches with class
temporal profiles of Classes 17 and 33 perfectly. Then one can label Classes 17 and 33 to be same as the ideal temporal profile (“irrigated-surface-
water-rice-double crop”). This is a qualitative illustration of SMTs. For quantitative methods, refer to Thenkabail et al. (2007a).

dynamics through an automated algorithm. The fourth part of the been developed using remote sensing techniques. In addition,
method compares the agricultural cropland map derived using an we also considered a recent MODIS global land cover and land
automated algorithm (classified data) with that derived based on use map where croplands are included. We examined these maps
knowledge capture (reference map). The fifth part of the method to identify their strengths and weaknesses, to see how well they
applies the tested algorithm on an independent dataset of the same compare with each other, and to understand the knowledge gaps
area to automatically classify and identify agricultural cropland that need to be addressed. These maps were produced by:
classes. The sixth part of the method assesses accuracy and vali-
1. Thenkabail et al. (2009b, 2011; Biradar et al., 2009)
dates the classes derived from independent dataset using an auto-
2. Pittman et al. (2010)
mated algorithm (Thenkabail et al., 2012; Wu et al., 2014a,b).
3. Yu et al. (2013)
4. Friedl et al. (2010)
6.8 Remote Sensing–Based Global Thenkabail et al. (2009b, 2011; Figure 6.8; Table 6.3) used a com-
Cropland Products: Current bination of AVHRR, SPOT VGT, and numerous secondary (e.g.,
State-of-the-Art Maps, Their precipitation, temperature, and elevation) data to produce a global
Strengths, and Limitations irrigated area map (Thenkabail et al., 2009b, 2011) and a global
map of rain-fed cropland areas (Biradar et al., 2009; Thenkabail
Remote sensing offers the best opportunity to map and charac- et al., 2011; Figure 6.8; Table 6.3). Pittman et al. (2010; Figure 6.9;
terize global croplands most accurately, consistently, and repeat- Table 6.4) used MODIS 250 m data to map global cropland extent.
edly. Currently, there are three global cropland maps that have More recently, Yu et al. (2013; Figure 6.10; Table 6.5) produced a

© 2016 Taylor & Francis Group, LLC

 144 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

2.2 A Cropland Classification Algorithm (ACCA)

Thenkabail et al. (2012)

Irrigated and Rainfed ACCA illustration

Slope, SRTM MODIS Yearly Elevation, Landsat B3 MODIS Elevation,

derived Total NDVI from SRTM (chlorophyll Feb SRTM
absorption) NDVI derived

>18 ≤18
and or ≤130 >130 ≤1300 m >1300 m
>1.5 and ≤30 >30
>2.5% <1950 ≥1950 ≤510 m >510 m % %
≤2.5%
reflec reflec
tance tance

Irrigated Others Rainfed Others

Landsat B5 (moisture sensitivity) Landsat B3 (chlorophyll absorption) Elevation, SRTM derived

>15 and ≤25% reflectance ≤15 or >25% reflectance ≤16% reflectance >16% reflectance ≤900 m >900 m

Irrigated Others

2.3 Finalizing, after algorithm runs: (a) cropland

extent\areas, (b) 8 major crops, (c) irrigated vs.
rainfed, and (d) cropping intensities. Also, their 1 Double crop
Continuous crop

MODIS NDVI
Accuracy Assessments
0.5
Single crop
Field plot data (approx. 0
10,000 points already 1 2 3 4 5 6 7 8 9 10 11 12
Month
Available; approx. 10,000
Cropping intensities
during project)
Class identification,
labeling, and
accuracy assessment
process Congalton
(1991, 2009)
1–5 m imagery
(e.g., IKONOS, Quickbird) Geocover Landsat 150 m high-res.
over 1000s in USGS archive data of the world

Secondary data from national

system (e.g., Central Board Of Irrigation
Space-time spiral and Power, India; USDA, USA)
Google Earth high-res.
data of the World Bi-spectral plots curve (ST-SC)
National and International
statistical Data:
1. FAO Aquastat
2. USDA FAS and CDL,
Is the class 3. National statistics;
Yes No
identified 4. National maps
Note: apart from point
field-plot data these used for
Google Panaromia high-res. training algorithms and
Mixed class accuracy assessments
data of the World
Mask image area of mixed
Final class names class, re-run the STM and\or
ACCA algorithms and go through
class identification and labeling
process till all areas are resolved

(a)

FIGURE 6.7 (a) Cropland mapping method illustrated here for a global scale (see Thenkabail et al., 2009b, 2011). The flowchart demonstrates
comprehensive global cropland mapping methods using multisensor, multidate remote sensing, secondary, field plot, and very-high-resolution
imagery data. (Continued )

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 145

1.0 Image masks and segments of mega-file data cube of the world
Image masks to segment the mega-file that includes landsat + MODIS data into precipitation, temperature,
elevation, forest, and desert zones of the world, Apply Cropland Mapping Algorithms (CMAs) to each zone,
And discern: (a) cropland extent, (b) crop types, (c) Irrigated vs. rainfed, and (d) cropping intensities

Croplands and Croplands and Insignificant Croplands and

Non-croplands zone cropland zone Insignificant Mountain croplands
non-croplands mix non-croplands mix cropland zone non-croplands mix
Precipitation Temperature Forest Cover Elevation All other Areas
Precipitation <280° Kelvin Density Areas (DAs)
<360 mm/year >2400 mm/year Density >75% >2500 m Of the world
(PLT 360) (TLT 280) (FGT 75) (FGT 2500)
(PLT 2400) (AOAW)

Note: A. Global Mega-file data-cube (GMFDC) 2010: 47 data layers

PLT = Precipitation less than Global Landsat ETM+mosaic for year 2010 (GLS2010)
PGT = Precipitation grater than
TLT = Temperature less than -6 non-thermal bands Data Normalization
FGT = Forest cover grater than Global SRTM DEM 90 m Global coverage (GDEM) and mosaicking (@
FSAR = Forest cover from -1 bands NASA Ames super
Synthetic aperture radar Global MODIS NDVI MVC, monthly 2009–2011
FGT = Forest cover grater than -36 bands (1 per month for 3 years) computer)
AOAD = All other area of the world Global Secondary Data: (average global: 1 layer per data) -Surface reflectance
- Precipitation, 40 years Average (source: CRU) -Global mosaicking
GIAM and GMRCA Cropland Masks
- Forest Cover 1 km, one time (Defries et al.)
- Skin temperature, 20 years mean (source: AVHRR)
- Evapotranspiration 100 year (source: IWMI Water Portal)
B. Global Mega-file data-cube (GMFDC) 1990: 47 data layers Data Mining and Reduction
(a) 6 Landsat 30 m bands from GLS1990; (b) 1 band of GDEM, Techniques
(c) 36 bands AVHRR NDVI, monthly 1989–1991; (d) 4 secondary layers -JPEG2010 lossless
Analyze areas outside GIAM & Compression
GMRCA 1 km separately -Wavelet compression

Segments based on: precip., temp., elev., forests, and deserts

PLT 360 PGT 2400 TLT 280 FGT 75 DA FGT 2500 AOAW
Segment Segment Segment Segment Segment Segment Segment
47 data Layers 47 data Layers 47 data Layers 47 data Layers 47 data Layers 47 data Layers 47 data Layers
for 2010, 1990 for 2010, 1990 for 2010, 1990 for 2010, 1990 for 2010, 1990 for 2010, 1990 for 2010, 1990

2.0 Cropland Mapping Algorithms (CMAs)

2.1 Spectral Matching Techniques (SMTs)

Thenkabail et al. (2007a) http://www.iwmigiam.org/info/main/index.asp
SMTquantitative Algorithm provides rapid, automated computation of croplands
2.1.1 Generate class spectra of each segment

PLT 360 PGT 2400 TLT 280 FGT 75 DA FGT 1500 AOAW

Spectral Matching Technique (SMT):

SMTQualitative 2.2 Match class spectra with ideal spectra
SMTIdeal
Rice double crop ideal Rice double crop class
spectra (n = 275) spectra, several classes SMT Quantitative
1.0 1.0
(a) (b)
0.8
0.8
Shape measure
0.6 0.6 2.1.2 Develop ideal
NDVI

NDVI

0.4 0.4
2.2.1 Spectral correlation similarity spectral data bank
0.2
Ideal spectral signature for Irrigated-SW-rice-DC
0.2
Similar class spectra signatures R. Squared value (SCS-R2) (ISDB) of:
0.0
0.0
Jun-00 Sep-00 Dec-00 Mar-01
Jun-00 Sep-00 Dec-00 Mar-01 (1) agricultural
Date
Irriggated-SW-rice-DC 15 16 17 18
Date
32 33 34 19 crops, (2) 8 major
crop types, (3)
1.0
(c)
1.0
(d)
watering method
0.8 0.8 (irrigated vs.
0.6 0.6 rainfed), and (4)
NDVI
NDVI

0.4 0.4
crop intensities
0.2 Ideal spectral signature match with similar
class spectra
0.2
Grouping similar classes
Shape & magnitude measure using approx., 7500
0.0
Jun-00 Sep-00 Dec-00
Date
Mar-01
0.0
Jun-00 Sep-00 Dec-00 Mar-01 field-plot data
15
18
34
16
32
19
17
33
Irriggated-SW-rice-DC 17 33
Date
Irriggated-SW-rice-DC
2.2.2 Spectral similarity value (SSV)
points
Match class spectra with ideal spectra to
identify and label classes

(b)

FIGURE 6.7 (Continued ) (b) Cropland mapping methods illustrated for a global scale. Top half shows ACCA (see Thenkabail and Wu, 2012; Wu
et al., 2014a) and bottom half shows class identification and labeling process.

© 2016 Taylor & Francis Group, LLC

 146 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

N
W E
S
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E
90°0΄0˝ 90°0΄0˝

70°0΄0˝N 70°0΄0˝N

50°0΄0˝N 50°0΄0˝N

30°0΄0˝N 30°0΄0˝N

10°0΄0˝N 10°0΄0˝N

10°0΄0˝S 10°0΄0˝S

30°0΄0˝S 30°0΄0˝S

50°0΄0˝S 50°0΄0˝S

70°0΄0˝S 70°0΄0˝S

90°0΄0˝ 90°0΄0˝
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E

1. Croplands, irrigated dominance 3. Natural vegetation with minor cropland fractions

2. Croplands, rainfed dominance 4. Natural vegetation dominance with very minor

cropland fractions

FIGURE 6.8 Global cropland product by Thenkabail et al. (2011, 2009b) using the method illustrated in Figure 6.7 and described in Section 6.1.1
(details in Thenkabail et al., 2011, 2009b). This includes irrigated and rain-fed areas of the world. The product is derived using remotely sensed data
fusion (e.g., NOAA AVHRR, SPOT VGT, JERS SAR), secondary data (e.g., elevation, temperature, and precipitation), and in situ data. Total area
of croplands is 2.3 billion hectares.

nominal 30 m resolution cropland extent of the world. These three these products differ extensively. As a result, the cropland extents
global cropland extent maps are the best available current state- mapped by these products also vary significantly. The areas in
of-the-art products. Friedl et al. (2010; Figure 6.11; Table 6.6) used Tables 6.3 through 6.6 only show the full pixel areas (FPAs) and
500 m MODIS data in their global land cover and land use product not subpixel areas (SPAs). SPAs are actual areas, which can be esti-
(MCD12Q1) where croplands were one of the land cover classes. mated by reprojecting these maps to appropriate projections and
The methods, approaches, data, and definitions used in each of calculating the areas. For the purpose of this chapter, we did not
estimate SPAs. However, a comparison of the FPAs of the four
TABLE 6.3 Global Cropland Extent at Nominal 1-km Based
on Thenkabail et al. (2009b, 2011)a maps (Figures 6.8 through 6.11) shows significant differences in the
cropland areas (Tables 6.3 through 6.6) as well as significant differ-
Pixels
ences in the precise locations of the croplands (Figures 6.8 through
Class # Class Description (Names) (1 km) Percent (%)
6.11), the reasons for which are discussed in the next section.
1 Croplands, irrigated dominance 9,359,647 40
2 Croplands, rain-fed dominance 14,273,248 60
3 Natural vegetation with minor 5,504,037 6.8.1 Global Cropland Extent at
cropland fractions Nominal 1 km Resolution
4 Natural vegetation dominance with 44,170,083
very minor cropland fractions We synthesized the four global cropland products discussed
23,632,895 100 and produced a unified global cropland extent map GCE V1.0
a
at nominal 1 km (Table 6.7a; Figure 6.12a). The process involved
Total of approximately 2.3 billion hectares; Note that these are FPAs.
Actual area is SPA. The SPA is not estimated here. See Thenkabail et al. (2007b)
resampling each global cropland product to a common resolu-
for the methods for calculating SPAs; % calculated based on Class 1 and 2. tion of 1 km and then performing GIS data overlays to determine
Class 3 and 4 are very small cropland fragments. where the cropland extents matched and where they differed.

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 147

W E

160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E 180°0΄0˝ S
90°0΄0˝

70°0΄0˝N 70°0΄0˝N

50°0΄0˝N 50°0΄0˝N

30°0΄0˝N 30°0΄0˝N

10°0΄0˝N 10°0΄0˝N

10°0΄0˝S 10°0΄0˝S

30°0΄0˝S 30°0΄0˝S

50°0΄0˝S 50°0΄0˝S

70°0΄0˝S 70°0΄0˝S

90°0΄0˝ 90°0΄0˝
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E

1. Croplands

FIGURE 6.9 Global cropland extent map by Pittman et al. (2010) derived using MODIS 250 m data. There is only one cropland class, which
includes irrigated and rain-fed areas of the world. There is no discrimination between rain-fed and irrigated areas. Total area of croplands is
0.9 billion hectares.

TABLE 6.4 Global Cropland Extent at Nominal 250 m Based products match as croplands and then added irrigation status or
on Pittman et al. (2010)a other indicators (e.g., irrigation dominance, rain-fed; Table 6.7b)
Class # Class Description (Names) Pixels (1 km) Percent (%) from the product by Thenkabail et al. (2009b, 2011).
Table 6.7b and Figure 6.12b show 12 classes of which Classes 1
1 Croplands 8,948,507 100
and 2 are croplands with irrigated agriculture, Classes 3 and 4
a Total of approximately 0.9 billion hectares. Note that these are FPAs.
are croplands with rain-fed agriculture, Classes 5 and 6 are crop-
Actual area is SPA. SPA is not estimated here. See Thenkabail et al. (2007b) for
lands where irrigated agriculture dominates, Classes 7 and 8 are
the methods for calculating SPAs; % calculated based on Class 1.
croplands where rain-fed agriculture dominates, and Classes
Figure 6.12a shows the aggregated global cropland extent 9–12 are areas with minor or very minor cropland fractions.
map with its statistics in Table 6.7a. Class 1 in Figure 6.12a and Classes 9–12 are those with large areas of natural vegetation
Table 6.7a provides the global cropland extent included in all and\or desert lands and other lands.
four maps. Actual area of this extent is not calculated yet, but Interestingly, and surprisingly as well, only 20% (Class 1 and 3;
it includes approximately 2.3 billion hectares FPAs (Table 6.7a). Table 6.7b; Figure 6.12b) of the total cropland extent are matched
The spatial distribution of these 2.3 billion hectares is demon- precisely in all four products. Further, 49% (Class 1, 2, 3, 4, and 7;
strated as Class 1 in Figure 12a. Classes 2 and 3 are areas with Table 6.7b; Figure 6.12b) of the total cropland areas match in at
minor or very minor cropland fractions. Class 2 and Class 3 are least three of the four products. This implies that all the four
classes with large areas of natural vegetation and/or desert lands products have considerable uncertainties in determining the
and other lands. precise location of the croplands. The great degree of uncertainty
Figure 6.12b and Table 6.7b demonstrate where and by how in the cropland products can be attributed to factors including
much the four products match with one another. For example,
2,802,397 pixels (Class 1, Table 6.7b; Figure 6.12b) are croplands 1. Coarse resolution of the imagery used in the study
that are irrigated. Some of the products do not separately clas- 2. Definition of mapping products of interest
sify irrigated versus rain-fed croplands, although all four prod- 3. Methods and approaches adopted
ucts show where croplands are. We first identified where all four 4. Limitations of the data

© 2016 Taylor & Francis Group, LLC

 148 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

W E
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E
S
90°0΄0˝ 90°0΄0˝

70°0΄0˝N 70°0΄0˝N

50°0΄0˝N 50°0΄0˝N

30°0΄0˝N 30°0΄0˝N

10°0΄0˝N 10°0΄0˝N

10°0΄0˝S 10°0΄0˝S

30°0΄0˝S 30°0΄0˝S

50°0΄0˝S 50°0΄0˝S

70°0΄0˝S 70°0΄0˝S

90°0΄0˝ 90°0΄0˝
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E

1. Croplands (10–14) 2. Bare-cropland (94 and 24)

FIGURE 6.10 Global cropland extent map by Yu et al. (2013) derived at nominal 30 m data. Total area of croplands is 2.2 billion hectares. There
is no discrimination between rain-fed and irrigated areas.

TABLE 6.5 Global Cropland Extent at Nominal 30 m Based The product (Figure 6.12c; Table 6.7c) does not show where the
on Yu et al. (2013)a crop types are or even the crop dominance. However, cropping
Class # Class Description (Names) Pixels (1 km) Percent intensity can be gathered using multitemporal remote sensing
1 Croplands (Classes 10–14) 7,750,467 35
over these cropland areas.
2 Bare-cropland (Classes 94 and 24) 14,531,323 65
22,281,790 100
a
Total of approximately 2.2 billion hectares. Note that these are FPAs.
6.9 Change Analysis
Actual area is SPA. SPA is not estimated here. See Thenkabail et al. (2007b) for
Once the croplands are mapped (Figure 6.13), we can use the
the methods for calculating SPAs; % calculated based on Class 1 and 2.
time-series historical data such as continuous global cover-
age of remote sensing data from NOAA very-high-resolution
Table 6.7c and Figure 6.12c show five classes of which Classes 1 radiometer (VHRR) and advanced VHRR (AVHRR), Global
and 2 are croplands with irrigated agriculture, Class 3 is crop- Inventory Modeling and Mapping Studies (GIMMS; 1982–
land with rain-fed agriculture, Classes 4 and 5 have ONLY 2000), and MODIS time-series (2001–present) to help build
minor or very minor cropland fractions. We recommend the use an inventory of historical agricultural development (e.g.,
of this aggregated five class global cropland map (Figure 12c and Figures 6.13 and 6.14). Such an inventory will provide infor-
Table 6.7c) produced based on the four major cropland mapping mation including identifying areas that have switched from
efforts [i.e., Thenkabail et al. (2009a, 2011), Pittman et al. (2010), rain-fed to irrigated production (full or supplemental), and
Yu et al. (2013), and Friedl et al. (2010)] using remote sensing. noncropped to cropped (and vice versa). A complete history
This map (Figure 6.12c; Table 6.7c) provides clear consensus will require systematic analysis of remotely sensed data as
view on of four major studies on global: well as a systematic compilation of all routinely populated
cropland databases from the agricultural departments of
• Cropland extent location all countries throughout the world. The differences in pixel
• Cropland watering method (irrigation versus rain-fed) sizes in AVHRR versus MODIS will: (1) inf luence class

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 149

160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E
90°0΄0˝ 90°0΄0˝

70°0΄0˝N 70°0΄0˝N

50°0΄0˝N 50°0΄0˝N

30°0΄0˝N 30°0΄0˝N

10°0΄0˝N 10°0΄0˝N

10°0΄0˝S 10°0΄0˝S

30°0΄0˝S 30°0΄0˝S

50°0΄0˝S 50°0΄0˝S

70°0΄0˝S 70°0΄0˝S

90°0΄0˝ 90°0΄0˝
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E

GLC croplands (class 12 and class 14)

FIGURE 6.11 Global cropland classes (Class 12 and Class 14) extracted from MODIS Global land use and land cover (GLC) 500 m product
MCD12Q2 by Friedl et al. (2010). Total area of croplands is 2.7 billion hectares. There is no discrimination between rain-fed and irrigated crop-
land areas.

TABLE 6.6 Global Cropland Extent at Nominal 500 m Based 6.10 Uncertainties of Existing
on Friedl et al. (2010)1
Cropland Products
Class # Class Description (Names) Pixels (1 km) Percent
Currently, the main causes of uncertainties in areas reported
1 Global croplands (Class 12 and 14) 27,046,084 100
in various studies (Ramankutty et al., 2008; Thenkabail et al.,
a Approximately, total 2.7 billion hectares based on Class 12 and 14. Note
2009a,c) can be attributed to, but not limited to: (1) reluctance
that these are FPAs. Actual area is SPA. SPA is not estimated here. See
of national and state agencies to furnish the census data on irri-
Thenkabail et al. (2007b) for the methods for calculating SPAs.
gated area and concerns of their institutional interests in sharing
of water and water data; (2) reporting of large volumes of census
data with inadequate statistical analysis; (3) subjectivity involved
identification and labeling, and (2) cause different levels in the observation-based data collection process; (4) inadequate
of uncertainties. We will address these issues by determin- accounting of irrigated areas, especially minor irrigation from
ing SPAs and uncertainties involved in class accuracies and groundwater, in national statistics; (5) definitional issues involved
uncertainties in areas at various spatial resolutions using in mapping using remote sensing as well as national statistics; (6)
methods detailed in recent work of this team (Ozdogan and difficulties in arriving at precise estimates of AFs using remote
Woodcock, 2006; Thenkabail et al., 2007b; Velpuri et al., sensing; (7) difficulties in separating irrigated from rain-fed crop-
2009). Change analyses (Tomlinson, 2003) are conducted in lands; and (8) imagery resolution in remote sensing. Other limita-
order to investigate both the spatial and temporal changes in tions include (Thenkabail et al., 2009a, 2011)
croplands (e.g., Figures 6.13 and 6.14) that will help estab-
lish: (1) change in total cropland areas, (2) change in spatial 1. Absence of precise spatial location of the cropland areas
location of cropland areas, (3) expansion on croplands into for training and validation
natural vegetation, (4) expansion of irrigation, (5) change 2. Uncertainties in differentiating irrigated areas from rain-
from croplands to biofuels, and (6) change from croplands to fed areas
urban. Massive reductions in cropland areas in certain parts 3. Absence of crop types and cropping intensities
of the world will be detected, including cropland lost as a 4. Inability to generate cropland maps and statistics, routinely
result of reductions in available ground water supply due to 5. Absence of dedicated web\data portal for dissemination
overdraft (Jiang, 2009; Rodell et al., 2009; Wada et al., 2012). cropland products

© 2016 Taylor & Francis Group, LLC

 150 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

TABLE 6.7 Global Cropland Extent at Nominal 1-km Based on Four Major Studies: Thenkabail et al.
(2009b, 2011), Pittman et al. (2010), Yu et al. (2013), and Friedl et al. (2010).
Class # Class Description (Names) Pixels (1 km) Percent (%)
(a) Three class map a

1 Croplands 23,493,936 100

2 Cropland minor fractions 13,700,176
3 Cropland very minor fractions 44,662,570
(b) Twelve class mapb
1 Croplands all 4, irrigated 2,802,397 12
2 Croplands 3 of 4, irrigated 289,591 1
3 Croplands all 4, rain-fed 1,942,333 8
4 Croplands 3 of 4, rain-fed 427,731 2
5 Croplands, 2 of 4, irrigation dominance 3,220,330 14
6 Croplands, 2 of 4, irrigation dominance 1,590,539 7
7 Croplands, 3 of 4, rain-fed dominance 6,206,419 26
8 Croplands, 2 of 4, rain-fed dominance 3,156,561 13
9 Croplands, minor fragments, 2 of 4 3,858,035 17
10 Croplands, very minor fragments, 2 of 4 6,825,290
11 Croplands, minor fragments, 1 of 4 6,874,886
12 Croplands, very minor fragments, 1 of 4 44,662,570
Class 1–9 total 23,493,936 100
(c) Five class mapc
1 Croplands, irrigation major 3,091,988 13
2 Croplands, irrigation minor 4,810,869 21
3 Croplands, rain-fed 11,733,044 50
4 Croplands, rain-fed minor fragments 3,858,035 16
5 Croplands, rain-fed very minor fragments 13,700,176
Classes 1–4 total 23,493,936 100.0%
a Approximately 2.3 billion hectares (Class 1) of cropland is estimated. But this is full pixel area (FPA). Actual area is sub
pixel area (SPA). SPA is not estimated here. See Thenkabail et al. (2007b) for the methods for calculating SPAs; % calculated
based on Class 1; Class 2 and 3 are minor/very minor cropland fragments.
b Approximately 2.3 billion hectares (Class 1–9) of cropland is estimated. But this is FPA. Actual area is SPA. SPA is not

estimated here. See Thenkabail et al. (2007b) for the methods for calculating SPAs; % calculated based on Class 1–9; Classes
10, 11, and 12 are minor cropland fragments; All 4 means, all 4 studies agreed.
c Approximately 2.3 billion hectares (Class 1–4) of cropland is estimated. But this is FPA. Actual area is SPA. SPA is not

estimated here. See Thenkabail et al. (2007b) for the methods for calculating SPAs; % calculated based on Class 1–4; Class
5 is very minor cropland fragments.

These limitations are a major hindrance in accurate/reliable global, methods. Previous research has shown that at finer spatial resolu-
regional, and country-by-country water use assessments that in tion, the accuracy of irrigated and rain-fed area class delineations
turn support crop productivity (productivity per unit of land, improves, because at finer spatial resolution, more fragmented
kg/m2) studies, water productivity (productivity per unit of water, and smaller patches of irrigated and rain-fed croplands can be
kg/m3) studies, and food security analyses. The higher degrees of delineated (Ozdogan and Woodcock, 2006; Velpuri et al., 2009).
uncertainty in coarser resolution data are a result of an inability to Further, greater details of crop characteristics such as crop types
capture fragmented, smaller patches of croplands accurately, and (e.g., Figure 6.15) can be determined at finer spatial resolutions.
the homogenization of both crop and noncrop land within areas of Crop type mapping will involve the use of advanced methods of
patchy land cover distribution. In either case, there is a strong need analysis such as data fusion of higher spatial resolution images
for finer spatial resolution to resolve the confusion. from sensors such as Resourcesat\Landsat and AWiFS\MODIS
(e.g., Table 6.2) supported by extensive ground surveys and ideal
6.11 Way Forward spectral data bank (ISDB) (Thenkabail et al., 2007a). Harmonic
analysis is often adopted to identify crop types (Sakamoto et al.,
Given the aforementioned issues with existing maps of global crop- 2005) using methods such as the conventional Fourier analysis
lands, the way forward will be to produce global cropland maps at and adopting a Fourier filtered cycle similarity (FFCS) method.
finer spatial resolution and applying a suite of advanced analysis Mixed classes are resolved using hierarchical crop mapping

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 151

W E

160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E S
90°0΄0˝ 90°0΄0˝

70°0΄0˝N 70°0΄0˝N

50°0΄0˝N 50°0΄0˝N

30°0΄0˝N 30°0΄0˝N

10°0΄0˝N 10°0΄0˝N

10°0΄0˝S 10°0΄0˝S

30°0΄0˝S 30°0΄0˝S

50°0΄0˝S 50°0΄0˝S

70°0΄0˝S 70°0΄0˝S

90°0΄0˝ 90°0΄0˝
160°0΄0˝W140°0΄0˝W120°0΄0˝W100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E

(a)
1. Global cropland extent 2. Cropland minor fractions 3. Cropland insignificant fractions

N
W E
S
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E
90°0΄0˝ 90°0΄0˝

70°0΄0˝N 70°0΄0˝N

50°0΄0˝N 50°0΄0˝N

30°0΄0˝N 30°0΄0˝N

10°0΄0˝N 10°0΄0˝N

10°0΄0˝S 10°0΄0˝S

30°0΄0˝S 30°0΄0˝S

50°0΄0˝S 50°0΄0˝S

70°0΄0˝S 70°0΄0˝S

90°0΄0˝ 90°0΄0˝
160°0΄0˝W 140°0΄0˝W 120°0΄0˝W 100°0΄0˝W 80°0΄0˝W 60°0΄0˝W 40°0΄0˝W 20°0΄0˝W 0°0΄0˝ 20°0΄0˝E 40°0΄0˝E 60°0΄0˝E 80°0΄0˝E 100°0΄0˝E 120°0΄0˝E 140°0΄0˝E 160°0΄0˝E
(b)
1. Croplands all 4, irrigated 5. Croplands, 2 of 4, irrigation dominance 9. Croplands, minor fragments, 2 of 4
2. Croplands 3 of 4, irrigated 6. Croplands, 2 of 4, irrigation dominance 10. Croplands, insignificant fragments, 2 of 4
3. Croplands all 4, rainfed 7. Croplands, 3 of 4, rainfed dominant 11. Croplands, minor fragments, 1 of 4
4. Croplands 3 of 4, rainfed 8. Croplands, 2 of 4, rainfed dominance 12. Croplands, insignificant fragments, 1 of 4

FIGURE 6.12 (a) An aggregated three class global cropland extent map at nominal 1 km based on four major studies: Thenkabail et al. (2009a,
2011), Pittman et al. (2010), Yu et al. (2013), and Friedl et al. (2010). Class 1 is total cropland extent; total cropland extent is 2.3 billion hectares
(FPAs). Class 2 and Class 3 have ONLY minor fractions of croplands. Refer to Table 6.7a for cropland statistics of this map. (b) A disaggregated
twelve class global cropland extent map derived at nominal 1-km based on four major studies: Thenkabail et al. (2009a, 2011), Pittman et al. (2010),
Yu et al. (2013), and Friedl et al. (2010). Classes 1–9 are cropland classes that are dominated by irrigated and rain-fed agriculture. Classes 10–12 have
ONLY minor or very minor fractions of croplands. Refer to Table 6.7b for cropland statistics of this map. (Continued )

© 2016 Taylor & Francis Group, LLC

 152 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

N
W E
S
140°0΄0˝W 100°0΄0˝W 60°0΄0˝W 20°0΄0˝W 0°0΄0˝ 40°0΄0˝E 80°0΄0˝E 120°0΄0˝E 160°0΄0˝E
90°0΄0˝ 90°0΄0˝

70°0΄0˝N 70°0΄0˝N

50°0΄0˝N 50°0΄0˝N

30°0΄0˝N 30°0΄0˝N

10°0΄0˝N 10°0΄0˝N

10°0΄0˝S 10°0΄0˝S

30°0΄0˝S 30°0΄0˝S

50°0΄0˝S 50°0΄0˝S

70°0΄0˝S 70°0΄0˝S

0 1500 3000 6000 Miles

90°0΄0˝ 90°0΄0˝
140°0΄0˝W 100°0΄0˝W 60°0΄0˝W 20°0΄0˝W 0°0΄0˝ 40°0΄0˝E 80°0΄0˝E 120°0΄0˝E 160°0΄0˝E

(c)
Class–names 3. Croplands, rainfed
1. Croplands, irrigation major 4. Croplands, rainfed minor fragments
2. Croplands, irrigation minor 5. Croplands, rainfed very minor fragments

FIGURE 6.12 (Continued ) (c) A disaggregated five class global cropland extent map derived at nominal 1-km based on four major studies:
Thenkabail et al. (2009a, 2011), Pittman et al. (2010), Yu et al. (2013), and Friedl et al. (2010). Classes 1–5 are cropland classes, that are dominated
by irrigated and rain-fed agriculture. However, Class 4 and Class 5 have ONLY minor or very minor fractions of croplands. Refer to Table 6.7c
for cropland statistics of this map. Note: Irrigation major: areas irrigated by large reservoirs created by large and medium dams, barrages, and
even large ground water pumping. Irrigation minor: areas irrigated by small reservoirs, irrigation tanks, open wells, and other minor irrigation.
However, it is very hard to draw a strict boundary between major and minor irrigations and in places, there can be significant mixing. Major irri-
gated areas such as the Ganges basin, California’s central valley, Nile basin, etc., are clearly distinguishable as major irrigation, and in other areas
major and minor irrigation may be intermixed.

protocol based on decision tree algorithm (Wardlow and Egbert, double crop, and continuous year round crop). Numerous other
2008). Irrigated versus rain-fed croplands will be distinguished methods and approaches exist. But, the ultimate goal using mul-
using spectral libraries (Thenkabail et al., 2007b) and ideal spec- tisensor remote sensing is to produce croplands products such as
tral data banks (Thenkabail et al., 2007a, 2009a). Similar classes
will be grouped by matching class spectra with ideal spectra based 1. Cropland extent\area
on SMTs (SMTs; Thenkabail et al., 2007a). Details such as crop 2. Crop types (initially focused on eight crops that occupy
types are crucial for determining crop water use, crop productiv- 70% of global croplands)
ity, and water productivity, leading to providing crucial informa- 3. Irrigated versus rain-fed croplands
tion needed for food security studies. However, the high spatial 4. Cropping intensities\phenology (single, double, triple, and
resolution must be fused with high temporal resolution data in continuous cropping)
order to obtain time-series spectra that are crucial for monitoring 5. Cropped area computation
crop growth dynamics and cropping intensity (e.g., single crop, 6. Cropland change over space and time

© 2016 Taylor & Francis Group, LLC

© 2016 Taylor & Francis Group, LLC

Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km)

180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E
January 2000 February 2000 March 2000 April 2000

30°N 60°N

30°N 60°N
30°N 60°N
30°N 60°N

30°N 60°N

30°N 60°N
30°N 60°N

30°N 60°N
NDVI NDVI NDVI NDVI
–1.0–0.10 –1.0–0.10 –1.0–0.10 –1.0–0.10
>0.10–0.20 >0.10–0.20 >0.10–0.20 >0.10–0.20
>0.20–0.30 >0.20–0.30 >0.20–0.30 >0.20–0.30

0°

0°
0°
0°

0°

0°
0°

0°
>0.30–0.40 >0.30–0.40 >0.30–0.40 >0.30–0.40
>0.40–0.50 >0.40–0.50 >0.40–0.50 >0.40–0.50

30°S

30°S
30°S
30°S

30°S

30°S
30°S

30°S
>0.50–0.60 >0.50–0.60 >0.50–0.60 >0.50–0.60
>0.60–0.70 >0.60–0.70 >0.60–0.70 >0.60–0.70
>0.70–0.80 N
>0.70–0.80 N
>0.70–0.80 N
>0.70–0.80 N

60°S

60°S
60°S
60°S

60°S

60°S
60°S

60°S
>0.80–1.0 W

S
E
>0.80–1.0 W

S
E
>0.80–1.0 W

S
E
>0.80–1.0 W

S
E

3,300 1,650 0 3,300 3,300 1,650 0 3,300 3,300 1,650 0 3,300 3,300 1,650 0 3,300
Country boundaries km Country boundaries km Country boundaries km Country boundaries km

Note: areas in gray color are other land use/land cover

Rainfed areas Note: areas in gray color are other land use/land cover
Rainfed areas Note: areas in gray color are other land use/land cover
Rainfed areas Note: areas in gray color are other land use/land cover
Rainfed areas
180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E
180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 180°W 120°W 60°W 0° 60°E 120°E 180°E
60°W 0° 60°E 120°E 180°E
December 2000 May 2000

30°N 60°N
30°N 60°N

30°N 60°N
30°N 60°N

NDVI NDVI

60°N
60°N
–1.0–0.10 –1.0–0.10
>0.10–0.20 >0.10–0.20
>0.20–0.30 >0.20–0.30
Class 3

0°
0°

0°
0°

1
>0.30–0.40 >0.30–0.40
>0.40–0.50
0.8
1 Class 6 >0.40–0.50

NDVI
0.6

30°S
30°S >0.50–0.60

30°S
30°S

>0.50–0.60 0.8

30°N
30°N
0.4

NDVI
>0.60–0.70 0.6 >0.60–0.70
>0.70–0.80
0.2
0.4
1
Class 2 >0.70–0.80 N
N
0

60°S
60°S

60°S
0.2 >0.80–1.0
60°S

W E

>0.80–1.0 W

S
E
J M A N
0.8 3,300 1,650
S
0 3,300
3,300 1,650 0 3,300 Date 0 Country boundaries
Legend

NDVI
Country boundaries km J M A N 0.6
km

Date 0.4
Note: areas in gray color are other land use / land cover
Rainfed areas 01. Irrigated: wheat and rice dominant 0.2 Note: areas in gray color are other land use/land cover
Rainfed areas
180°W 120°W 60°W 0° 60°E 120°E 180°E 02. Irrigated mixed crops 1: wheat, rice, 0 180°W 120°W 60°W 0° 60°E 120°E 180°E
0°

0°
D M J S J
180°W 120°W 60°W 0° 60°E 120°E 180°E barley and soybeans Date 180°W 120°W 60°W 0° 60°E 120°E 180°E
November 2000 03. Irrigated mixed crops 2: corn, wheat, 1 Class 1 June 2000
rice, cotton and orchards
30°N 60°N

30°N 60°N
Class 7 0.8
30°N 60°N

30°N 60°N
1

NDVI
0.8 0.6
04. Rainfed: wheat, rice, soybeans,

NDVI
0.4
0.6

30°S
30°S

sugarcane, corn and cassava 0.4

0.2
NDVI 0
NDVI
–1.0–0.10 05. Rainfed: wheat and barley dominant 0.2 D M J S J –1.0–0.10
0 Date
>0.10–0.20 06. Rainfed: corn and soybeans dominant J M A N
>0.10–0.20
Date N
>0.20–0.30 >0.20–0.30
0°

0°
0°

0°
>0.30–0.40 07. Rainfed mixed crops 1: wheat, W E
>0.30–0.40
corn, rice, barley and soybeans
60°S

>0.40–0.50 S >0.40–0.50

60°S
3,300 1,650 0 3,300
08. Minor fractions of mixed crops: wheat,
30°S

30°S
30°S

30°S
>0.50–0.60 km >0.50–0.60
>0.60–0.70 maize, rice, barley and soybeans >0.60–0.70
>0.70–0.80 N Countries >0.70–0.80 N
60°S

60°S
60°S

60°S
>0.80–1.0 W

S
E
>0.80–1.0 W

S
E

3,300 1,650 0 3,300 3,300 1,650 0 3,300

Country boundaries km Country boundaries km

Note: areas in gray color are non-croplands Global Irrigated and Rainfed Croplands @ nominal 1 km (Source: Thenkabail et al., 2012)
Note: areas in gray color are other land use / land cover
Rainfed areas Note: areas in gray color are other land use/land cover
Rainfed areas
180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E
180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E
October 2000 September 2000 August 2000 July 2000
30°N 60°N

30°N 60°N
30°N 60°N
30°N 60°N

30°N 60°N

30°N 60°N
30°N 60°N

30°N 60°N
NDVI NDVI NDVI NDVI
–1.0–0.10 –1.0–0.10 –1.0–0.10 –1.0–0.10
>0.10–0.20 >0.10–0.20 >0.10–0.20 >0.10–0.20
>0.20–0.30 >0.20–0.30 >0.20–0.30 >0.20–0.30
0°

0°
0°
0°

0°

0°
0°

0°
>0.30–0.40 >0.30–0.40 >0.30–0.40 >0.30–0.40
>0.40–0.50 >0.40–0.50 >0.40–0.50 >0.40–0.50
30°S

30°S
30°S
30°S

30°S

30°S
>0.50–0.60 >0.50–0.60 30°S >0.50–0.60 >0.50–0.60

30°S
>0.60–0.70 >0.60–0.70 >0.60–0.70 >0.60–0.70
>0.70–0.80 N
>0.70–0.80 N
>0.70–0.80 N
>0.70–0.80 N
60°S

60°S
60°S
60°S

60°S

60°S
60°S

60°S
>0.80–1.0 >0.80–1.0 >0.80–1.0 >0.80–1.0
W E W E W E W E

S S S S
3,300 1,650 0 3,300 3,300 1,650 0 3,300 3,300 1,650 0 3,300 3,300 1,650 0 3,300
Country boundaries km Country boundaries km Country boundaries km Country boundaries km

Note: areas in gray color are other land use/land cover Rainfed areas Note: areas in gray color are other land use/land cover
Rainfed areas Note: areas in gray color are other land use/land cover
Rainfed areas Note: areas in gray color are other land use/land cover
Rainfed areas
180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E 180°W 120°W 60°W 0° 60°E 120°E 180°E

FIGURE 6.13 Center image of global cropland (irrigated and rainfed) areas @ 1 km for year 2000 produced by overlying the remote sensing derived product of the International Water
Management Institute (IWMI; Thenkabail et al., 2012, 2011, 2009a,b; http://www.iwmigiam.org) over five dominant crops (wheat, rice, maize, barley, and soybeans) of the world produced by
Ramankutty et al. (2008). The five crops constitute about 60% of all global cropland areas. The IWMI remote sensing product is derived using remotely sensed data fusion (e.g., NOAA AVHRR,
SPOT VGT, and JERS SAR), secondary data (e.g., elevation, temperature, and precipitation), and in situ data. Total area of croplands is 1.53 billion hectares, of which 399 million hectares is
total area available for irrigation (without considering cropping intensity) and 467 million hectares is annualized irrigated areas (considering cropping intensity). Surrounding NDVI images
of irrigated areas: From January to December irrigated area NDVI dynamics is produced using NOAA AVHRR NDVI. The irrigated areas were determined by Thenkabail et al. (2011, 2009a,b).

153
© 2016 Taylor & Francis Group, LLC

154
1.0 1 1.0
Class 6: Rainfed single crop Class 5: Rainfed single crop Class 7: Rainfed single crop
0.9 0.9 0.9
Consistent crop productivity Severe drought years in Spain Massive crop failure resulted in China importing
0.8 0.8 0.8
0.7
over 2 decades, year after year 0.7 0.7 20 million metric ton of grain in 1995-96
0.6 0.6 0.6
NDVI

NDVI

NDVI
0.5 0.5 0.5
0.4 0.4 0.4
0.3 0.3 0.3
0.2 0.2 0.2
0.1 0.1 0.1
0.0 0 0.0
Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00

Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00

Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00
Month Month Month

180°W 120°W 60°W 0° 60°E 120°E 180°E Class 2: Irrigated single crop
1.0 Class 3: irrigated mixed crops 1.0
0.9 0.9
Increase in tree\fruit\nut crops in later half of Massive crop failure resulted in China importing
0.8 0.8
1980s and 1990s as opposed to greater %age 20 million metric ton of grain in 1995-96

60°N
0.7 0.7

60°N
0.6 of vegetables and diary in early 1980s 0.6
NDVI

NDVI
0.5 1 Class 3 0.5
0.4
0.8
1 Class 6 0.4

NDVI
0.6

30°N
0.8

30°N
0.4
0.3

NDVI
0.3 0.6
0.2
0.4 Class 2
0.2 0
J M A N 0.2
1 0.2
Date 0.8

Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

0
Legend

NDVI
0.1 J M A N 0.6 0.1
Date 0.4
0.0 01. Irrigated: wheat and rice dominant 0.2 0.0
Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00

Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00
0°
0
02. Irrigated mixed crops 1: wheat, rice,
0°
D M J S J
Date
barley and soybeans
03. Irrigated mixed crops 2: corn, wheat, 1 Class 1
Month rice, cotton and orchards 1 Class 7 0.8 Month

NDVI
0.8 0.6
04. Rainfed: wheat, rice, soybeans,

30°S
NDVI
0.4
30°S

0.6
sugarcane, corn and cassava
1.0 Class 4: 1.0 Class 4: 0.4
0.2
0
1.0 Class 1: 1.0 Class 1:
05. Rainfed: wheat and barley dominant 0.2 D M J S J
0.8 June 1982 to May 1983 0.8 June 1998 to May 1999 06. Rainfed: corn and soybeans dominant
0
J M A N
Date 0.8 June 1982 to May 1983 0.8 June 1998 to May 1999
Date N
0.6 0.6

NDVI
0.6 0.6
NDVI

NDVI

NDVI
07. Rainfed mixed crops 1: wheat, W E

60°S
corn, rice, barley and soybeans
60°S

0.4 0.4 3,300 1,650

S
0 3,300
0.4 0.4
08. Minor fractions of mixed crops: wheat,
km
0.2 0.2 maize, rice, barley and soybeans 0.2 0.2
0.0 0.0 Countries 0.0 0.0

Jun-82
Jul-82
Aug-82
Sep-82
Oct-82
Nov-82
Dec-82
Jan-83
Feb-83
Mar-83
Apr-83

Jun-98
Jul-98
Aug-98
Sep-98
Oct-98
Nov-98
Dec-98
Jan-99
Feb-99
Mar-99
Apr-99
May-83

May-99
May-83
Jun-82
Jul-82
Aug-82
Sep-82
Oct-82

Dec-82
Jan-83
Feb-83
Mar-83
Apr-83
Nov-82

May-99
Jun-98
Jul-98
Aug-98
Sep-98
Oct-98
Nov-98
Dec-98
Jan-99
Feb-99
Mar-99
Apr-99

Note: areas in gray color are non-croplands Global Irrigated and Rainfed Croplands @ nominal 1km (Source: Thenkabail et al., 2012)
Month Month 180°W 120°W 60°W 0° 60°E 120°E 180°E Month Month

1 1 1.0
Class 4: Rainfed single crop Class 7: Rainfed single crop Class 1: Irrigated double crop
0.9 0.9 0.9
0.8 0.8 0.8 Irrigation double crop throughout in 2 decades
0.7 0.7 0.7
0.6 0.6
Severe drought years 0.6
NDVI

NDVI

NDVI
0.5 0.5 0.5
0.4 0.4 0.4
0.3 High cropping intensity throughout 2 decades 0.3 0.3
0.2 (see Class 4 plots for June 1982 to may 1983 and 0.2 0.2
0.1 June 1988 to May 1999 above) 0.1 0.1
0.0 0.0 0.0
Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00

Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00

Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jul-84
Jan-85
Jul-85
Jan-86
Jul-86
Jan-87
Jul-87
Jan-88
Jul-88
Jan-89
Jul-89
Jan-90
Jul-90
Jan-91
Jul-91
Jan-92
Jul-92
Jan-93
Jul-93
Jan-94
Jul-94
Jan-95
Jul-95
Jan-96
Jul-96
Jan-97
Jul-97
Jan-98
Jul-98
Jan-99
Jul-99
Jan-00
Jul-00
Month Month Month

FIGURE 6.14 Global agricultural dynamics over two decades illustrated here for some of the most significant agricultural areas of the World. Once we establish GCAD2010 and GCAD1990
at nominal 30 m resolution for the entire world, we will use AVHRR-MODIS monthly MVC NDVI time-series from 1982 to 2017 to provide a continuous time history of global irrigated and
rain-fed croplands, establish their spatial and temporal changes, and highlight the hot spots of change. The GCAD2010, GCAD1990, and GCAD four decade’s data will be made available on
USGS global cropland data portal (currently under construction): http://powellcenter.usgs.gov/current_projects.php#GlobalCroplandsAbstract. Further, the need to map accurately specific
cropland characteristics such as crop types and watering methods (e.g., irrigated versus rain-fed) is crucial in food security analysis. For example, the importance of irrigation to global food
security is highlighted in a recent study by Siebert and Döll (2010) who show that without irrigation, there would be a decrease in production of various foods including dates (60%), rice (39%),
cotton (38%), citrus (32%), and sugarcane (31%) from their current levels. Globally, without irrigation, cereal production would decrease by a massive 43%, with overall cereal production, from
irrigated and rain-fed croplands, decreasing by 20%.
 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 155

60°0΄0˝E 70°0΄0˝E 80°0΄0˝E 90°0΄0˝E 100°0΄0˝E

6000
Rice areas from MODIS 2000-01 (area in km2)
y = 0.8871x
40°0΄0˝N

40°0΄0˝N
R2 = 0.9469

5000
4000
3000
2000
1000
30°0΄0˝N

30°0΄0˝N
0
0 1000 2000 3000 4000 5000 6000
t an Rice areas from National Statistics 2000-01 (area in km2)
kis
Pa Nep
al
Bhutan

Bangladesh

dia
In

20°0΄0˝N
20°0΄0˝N

Irrigated, single crop

South Asia rice classes 0.9
01. Irrigated rice
02. Rainfed rice 0.6

NDVI
03. Water bodies
04. Urban areas 0.3

05. Other land use/land cover

10°0΄0˝N
10°0΄0˝N

0.0
South Asia countries
Jun-00 Sep-00 Dec-00 Mar-01
Month

Irrigated, double crop Sri Lanka Rainfed, single crop

0.9 0.9

0.6 0.6
NDVI
NDVI

0.3 0.3

0°0΄0˝
0°0΄0˝

0.0 0.0
Jun-00 Sep-00 Dec-00 Mar-01 Jun-00 Sep-00 Dec-00 Mar-01
0 250 500 1000 km

Month Month

60°0΄0˝E 70°0΄0˝E 80°0΄0˝E 90°0΄0˝E 100°0΄0˝E

FIGURE 6.15 Rice map of south Asia produced using the method illustrated in Figure 6.6. (From Gumma, M. K. et al., J. Appl. Rem. Sens., 5,
053547, September 1, 2011, doi:10.1117/1.3619838, 2011.)

6.12 Conclusions used at global, regional, and local levels. Some of the remote sens-
ing methods for global cropland mapping have been illustrated.
This chapter provides an overview of the importance of global The current state-of-the-art provides four-key global cropland
cropland products in food security analysis. It is obvious that products (-e.g., Figure 6.12) derived from remote sensing, based
only remote sensing from Earth-observing (EO) satellites pro- on the work conducted by four major studies (Thenkabail et al.
vides consistent, repeated, high-quality data for characterizing (2009a, 2011, Pittman et al. 2010, Yu et al. 2013, and Friedl et al.
and mapping key cropland parameters for global food security 2010). These studies were conducted using: (1) time-series of mul-
analysis. Importance of definitions and class naming conventions tisensor data and secondary data, (2) 250 m MODIS time-series
in cropland mapping has been reiterated. Typical EO systems data, (3) 30 m Landsat data, and (4) a MODIS 500 m time-series
and their spectral, spatial, temporal, and radiometric character- derived cropland classes from a land use\land cover product
istics useful for cropland mapping have been highlighted. The has been used. These four studies help synthesized, at nominal
chapter provides a review of various cropland mapping methods 1 km, to obtain a consensus cropland mask of the world (global

© 2016 Taylor & Francis Group, LLC

 156 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

cropland extent version 1.0 or GCE V1.0). It was demonstrated Chander, G., Markham, B. L., and Helder, D. L. (2009). Summary
from these products that the uncertainty in location of croplands of current radiometric calibration coefficients for Landsat
in any one given product is quite high and no single product MSS, TM, ETM+, and EO-1 ALI sensors. Remote Sensing
maps croplands particularly well. Therefore, a synthesis identi- of Environment, 13 (5), 893–903. ISSN 0034-4257,
fies where some or all of these products agree and where they http://dx.doi.org/10.1016/j.rse.2009.01.007. http://www.
disagree. This provides a starting point for the next level of more sciencedirect.com/science/article/pii/S0034425709000169.
detailed cropland mapping at 250 and 30 m (see ongoing efforts Last accessed March 15th, 2014.
at: http://geography.wr.usgs.gov/science/croplands/ and https:// Cohen, W. B. and Goward, S. N. (2004). Landsat’s role in eco-
www.croplands.org/). The five key cropland products identified logical applications of remote sensing. BioScience, 54,
to be derived from remote sensing are: (1) cropland extent\areas, 535–545.
(2) cropping intensities, (3) watering method (irrigated versus Congalton, R. (1991). A review of assessing the accuracy of
rain-fed), (4) crop type, and (5) cropland change over time and classifications of remotely sensed data. Remote Sensing of
space. From these primary products, one can derive crop produc- Environment, 37, 35–46.
tivity and water productivity. Such products have great impor- Congalton, R. (2009). Accuracy and error analysis of global and
tance and relevance in global food security analysis. local maps: Lessons learned and future considerations. In:
Authors recommend the use of composite global cropland Remote Sensing of Global Croplands for Food Security, P.
map (see Figure 6.12c; Table 6.7c) that provides clear consensus Thenkabail, J. Lyon, H. Turral, and C. Biradar (eds). CRC/
view on of four major cropland studies on global Taylor & Francis, Boca Raton, FL, pp. 441–458.
Congalton, R. and Green, K. (2009). Assessing the Accuracy of
• Cropland extent location
Remotely Sensed Data: Principles and Practices, 2nd edn.
• Cropland watering method (irrigation versus rain-fed)
CRC/Taylor & Francis, Boca Raton, FL, 183pp.
The nominal 1 km product (Figure 6.12c and Table 6.7c) does Crist, E. P. and Cicone, R. C. (1984). Application of the tasseled cap
not show where the crop types are or even where the crop domi- concept to simulated Thematic Mapper data. Photogrammetric
nance occur. However, cropping intensity can be generated Engineering and Remote Sensing, 50, 343–352.
using multitemporal remote sensing for every pixel over these DeFries, R., Hansen, M., Townsend, J. G. R., and Sohlberg, R.
cropland areas. (1998). Global land cover classifications at 8 km resolution:
The use of training data derived from Landsat imagery in
Acknowledgments decision tree classifiers. International Journal of Remote
Sensing, 19, 3141–3168.
The authors thank NASA Making Earth Science Data Records Dheeravath, V., Thenkabail, P. S., Chandrakantha, G,
for Use in Research Environments (MEaSUREs) solicitation Noojipady, P., Biradar, C. B., Turral, H., Gumma, M. I.,
for funding this research. Support by USGS Powell Center for Reddy, G. P. O., and Velpuri, M. (2010). Irrigated areas of
a working group on global croplands is much appreciated. We India derived using MODIS 500m data for years 2001–
thank the global food security support analysis data @ 30 m 2003. ISPRS Journal of Photogrammetry and Remote
(GFSAD30) project team for inputs. Figures 6.1 and 6.2 were Sensing, 65 (1), 42–59. http://dx.doi.org/10.1016/j.
produced by Dr. Zhuoting Wu, USGS Mendenhall Post Doctoral isprsjprs.2009.08.004.
Researcher, we thank her for it. EPW (2008). Food Security Endangered: Structural changes
in global grain markets threaten India’s food security.
References Economic and Political Weekly, 43, 5.
FAO (2009). Food Outlook: Outlook Global Market Analysis.
Biggs, T., Thenkabail, P. S., Krishna, M., GangadharaRao, P., and Rome: FAO.
Turral, H. (2006). Vegetation phenology and irrigated area Foley, J. A., DeFries, R., Asner, G. P., Barford, C., Bonan, G.,
mapping using combined MODIS time-series, ground surveys, Carpenter, S. R., Chapin, F. S. et al. (2011). Solutions for a
and agricultural census data in Krishna River Basin, India. cultivated planet. Nature, 478 (7369), 337–342.
International Journal of Remote Sensing, 27 (19), 4245–4266. Foley, J. A., Monfreda, C., Ramankutty, N., and Zaks, D. (2007).
Bindraban, P. S., Bulte, E. H., and Conijn, S. G. (2009). Can large- Our share of the planetary pie. PNAS, 104, 12585–12586.
scale biofuel production be sustained by 2020? Agricultural Falkenmark, M. and Rockström, J. (2006). The new blue and
Systems, 101, 197–199. green water paradigm: Breaking new ground for water
Biradar, C. M., Thenkabail, P. S., Noojipady, P., Li, Y., Dheeravath, resources planning and management. Journal of Water
V., Turral, H., Velpuri, M. et al. (2009). A global map of Resource Planning and Management, 132, 1–15.
rainfed cropland areas (GMRCA) at the end of last millen- Friedl, M. A., McIver, D. K., Hodges, J. C. F., Zhang, X. Y.,
nium using remote sensing. International Journal of Applied Muchoney, D., and Strahler, A. H. (2002). Global land
Earth Observation and Geoinformation, 11 (2), 114–129. cover mapping from MODIS: Algorithms and early results.
doi:10.1016/j.jag.2008.11.002. Remote Sensing of Environment, 83, 287–302.

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 157

Friedl, M. A., Sulla-Menashe, D., Tan, B., Schneider, A., Kurz, B. and Seelan, S. K. Use of remote sensing to map irrigated
Ramankutty, N. S., and Huang, X. M. (2010). MODIS agriculture in areas overlying the Ogallala Aquifer, U.S. In
Collection 5 global land cover: Algorithm refinements Remote Sensing of Global Croplands for Food Security;
and characterization of new datasets. Remote Sensing of Thenkabail, P. S., Turral, H., Lyon, J. G., Biradar, C., Eds.;
Environment, 114 (1), 168–182. CRC Press: Boca Raton, FL, 2009.
Funk, C. and Brown, M. (2009). Declining global per capita Lal, R. and Pimentel, D. (2009). Biofuels from crop residues. Soil
agricultural production and warming oceans threaten and Tillage Research, 93, 237–238.
food security. Food Security, 1 (3), 271–289. doi:10.1007/ Loveland, T. R., Reed, B. C., Brown, J. F., Ohlen, D. O., Zhu, Z.,
s12571–009–0026-y. and Yang, L. (2000). Development of global land cover
Gibbs, H. K., Johnston, M., Foley, J. A., Holloway, T., Monfreda, C., characteristics database and IGBP DISCover from 1 km
Ramankutty, N., and Zaks, D. (2008). Carbon payback times AVHRR data. International Journal of Remote Sensing, 21,
for crop-based biofuel expansion in the tropics: The effects of 1303–1330.
changing yield and technology. Environment Research Letters, McIntyre, B. D. (2008). International assessment of agricul-
3(2008), 034001, 10pp. doi:10.1088/1748–9326/3/3/034001. tural knowledge, science and technology for development
Goldewijk, K., Beusen, A., de Vos, M., and van Drecht, G. (IAASTD): Global report. Includes Bibliographical References
(2011). The HYDE 3.1 spatially explicit database of human and Index. ISBN 978-1-59726-538-6, Oxford, U.K.
induced land use change over the past 12,000 years. Masek, J. G., Huang, C., Wolfe, R., Cohen, W., Hall, F., Kutler, J.,
Global Ecology and Biogeography, 20 (1), 73–86. and Nelson, P. (2008). North American forest disturbance
doi:10.1111/j.1466–8238.2010.00587.x. mapped from a decadal Landsat record. Remote Sensing of
Goodchild, M. and Gopal, S. (eds.) (1989). The Accuracy of Spatial Environment, 112, 2914–2926.
Databases. Taylor & Francis, New York, 290pp. Masek, J. G., Vermote, E. F., Saleous, N., Wolfe, R., Hall, F. G.,
Gordon, L. J., Finlayson, C. M., and Falkenmark, M. (2009). Huemmrich, Gao, F., Kutler, J., and Lim, T. K. (2006). A
Managing water in agriculture for food production and Landsat surface reflectance data set for North America,
other ecosystem services. Agricultural Water Management, 1990–2000. Geoscience and Remote Sensing Letters, 3, 68–72.
97, 512–519. doi: 10.1016/j.agwat.2009.03.017. Monfreda, C., Patz, J. A., Prentice, I. C., Ramankutty, N., and
Gumma, M. K., Nelson, A., Thenkabail, P. S., and Singh, A. N. Snyder, P. K. (2005). Global consequences of land use.
(2011). Mapping rice areas of South Asia using MODIS Science July 22; 309 (5734), 570–574.
multi temporal data. Journal of Applied Remote Sensing, 5, Monfreda, C., Ramankutty, N., and Foley, J. A. (2008). Farming
053547 (September 1, 2011). doi:10.1117/1.3619838. the planet: 2. Geographic distribution of crop areas, yields,
Gutman, G., Byrnes, R., Masek, J., Covington, S., Justice, C., physiological types, and net primary production in the year
Franks, S., and Headley, R. (2008). Towards monitoring 2000. Global Biogeochemical Cycles 22 (1), GB1022.
Land-cover and land-use changes at a global scale: The Olofsson, P., Stehman, S. V., Woodcock, C. E., Sulla-Menashe,
global land survey 2005. Photogrammetric Engineering and D., Sibley, A.M., Newell, J.D. et al. (2012). A global land
Remote Sensing, 74 (1): 6–10. cover validation dataset, I: Fundamental design principles.
Hansen, M. C., DeFries, R. S., Townshend, J. R. G., Sohlberg, R., International Journal of Remote Sensing, 33, 5768–5788.
Dimiceli, C., and Carroll, M. (2002). Towards an opera- Ozdogan, M. and Gutman, G. (2008). A new methodology to map
tional MODIS continuous field of percent tree cover algo- irrigated areas using multi-temporal MODIS and ancillary
rithm: Examples using AVHRR and MODIS data. Remote data: An application example in the continental US. Remote
Sensing of Environment, 83, 303–319. Sensing of Environment, 112 (9), 3520–3537.
Hossain, M., Janaiah, A., and Otsuka, K. (2005). Is the produc- Ozdogan, M. and Woodcock, C. E. (2006). Resolution dependent
tivity impact of the Green Revolution in rice vanishing? errors in remote sensing of cultivated areas. Remote Sensing
Economic and Political Weekly 5595–9600. of Environment, 103, 203–217.
Jiang, Y. (2009). China’s water scarcity. Journal of Environmental Pennisi, E. (2008). The Blue revolution, drop by drop, gene
Management, 90, 3185–3196. by gene. Science, 320 (5873): 171–173. doi:10.1126/
Johnson, D. M. and Mueller, R. (2010). The 2009 cropland data science.320.5873.171
layer. Photogrammetric Engineering and Remote Sensing, 76 Pittman, K., Hansen, M. C., Becker-Reshef, I., Potapov, P. V.,
(11), 1201–1205. and Justice, C. O. (2010). Estimating global cropland
Khan, S. and Hanjra, M. A. (2008). Sustainable land and water extent with multi-year MODIS data. Remote Sensing 2 (7),
management policies and practices: A pathway to envi- 1844–1863.
ronmental sustainability in large irrigation systems. Land Portmann, F., Siebert, S., and Döll, P. (2010). MIRCA2000—
Degradation and Development, 19, 469–487. Global monthly irrigated and rainfed crop areas around the
Kumar, M. D. and Singh, O. P. (2005). Virtual water in global food year 2000: A new high-resolution data set for agricultural
and water policy making: Is there a need for rethinking? and hydrological modelling. Global Biogeochemical Cycles,
Water Resources Management, 19, 759–789. 24 GB0003435. 24pp.

© 2016 Taylor & Francis Group, LLC

 158 Land Resources Monitoring, Modeling, and Mapping with Remote Sensing

Ramankutty, N., Evan, A. T., Monfreda, C., and Foley, J. A. (2008). Thenkabail, P. S., Hanjra, M. A., Dheeravath, V., Gumma, M.
Farming the planet: 1. Geographic distribution of global (2011). Global croplands and their water use remote sens-
agricultural lands in the year 2000. Global Biogeochemical ing and non-remote sensing perspectives. In: Weng, Q.
Cycles, 22. 19pp. doi:10.1029/2007GB002952. (ed.), Advances in Environmental Remote Sensing: Sensors,
Ramankutty, N. and Foley, J. A. (1998). Characterizing patterns of Algorithms, and Applications, Taylor & Francis, pp. 383–419.
global land use: An analysis of global croplands data. Global Thenkabail, P. S., Hanjra, M. A., Dheeravath, V., Gumma, M. A.
Biogeochemical Cycles 12 (4), 667–685. (2010). A holistic view of global croplands and their water use
Rodell, M., Velicogna, I., and Famiglietti, J. S. (2009). Satellit- for ensuring global food security in the 21st century through
based estimates of groundwater depletion in India. Nature, advanced remote sensing and non-remote sensing approaches.
460, 999–1002. doi:10.1038/nature08238. Journal Remote Sensing, 2 (1), 211–261. doi:10.3390/
Sakamoto, T., Yokozawa, M., Toritani, H., Shibayama, M., rs2010211. http://www.mdpi.com/2072–4292/2/1/211.
Ishitsuka, N., Ohno, H. (2005), A crop phenology detection Thenkabail, P. S., Knox, J. W., Ozdogan, M., Gumma, M.
method using time-series MODIS data. Remote Sensing of K., Congalton, R. G., Wu, Z., Milesi, C. et al. (2012).
Environment, 96, 366–374. Assessing future risks to agricultural productivity, water
Searchinger, T., Heimlich, R., Houghton, R. A., Dong, F., Elobeid, resources and food security: How can remote sensing
A., Fabiosa, J., Tokgoz, S., Hayes, D., and Yu, T. H. (2008). help? Photogrammetric Engineering and Remote Sensing,
Use of U.S. croplands for biofuels increases greenhouse August 2012 Special Issue on Global Croplands: Highlight
gases through emissions from land-use change. Science, 319, Article. 78(8), 773–782.
1238–1240. Thenkabail, P. S., Lyon, G. J., Turral, H., and Biradar, C. M. (2009a).
Siebert, S. and Döll, P. (2010). Quantifying blue and green Remote Sensing of Global Croplands for Food Security. CRC
virtual water contents in global crop production as Press-Taylor & Francis group, Boca Raton, FL, London, New
well as potential production losses without irrigation. York. pp. 556 (48 pages in color). Published in June, 2009.
Journal of Hydrology. 384(3–4), 198–217. doi:10.1016/j. Thenkabail, P. S., Lyon, G. J., Turral, H., and Biradar, C. M. (2009c).
jhydrol.2009.07.031. Remote Sensing of Global Croplands for Food Security. Boca
Siebert, S., Hoogeveen, J., and Frenken, K. (2006). Irrigation Raton, FL, London, New York: CRC Press/Taylor & Francis
in Africa, Europe and Latin America—Update of the digi- Group, Published in June.
tal global map of irrigation areas to version 4, Frankfurt Thenkabail, P. S., Mariotto, I., Gumma, M. K., Middleton, E. M.,
Hydrology Paper 05, 134 pp. Institute of Physical Geography, Landis, and D. R., Huemmrich, F. K. (2013). Selection of
University of Frankfurt, Frankfurt am Main, Germany and hyperspectral narrowbands (HNBs) and composition of
Rome, Italy. hyperspectral twoband vegetation indices (HVIs) for bio-
Thenkabail, P. S., Biradar, C. M., Noojipady, P., Cai, X. L., physical characterization and discrimination of crop types
Dheeravath, V., Li, Y. J., Velpuri, M., Gumma, M., and using field reflectance and Hyperion/EO-1 data. IEEE
Pandey, S. (2007b). Sub-pixel irrigated area calculation Journal of Selected Topics in Applied Earth Observations and
methods. Sensors Journal (special issue: Remote Sensing of Remote Sensing 6 (2), 427–439, April 2013. doi:10.1109/
Natural Resources and the Environment (Remote Sensing JSTARS.2013.2252601.
Sensors Edited by Assefa M. Melesse). 7: 2519–2538. http:// Thenkabail, P. S., Schull, M., and Turral, H. 2005. Ganges and
www.mdpi.org/sensors/papers/s7112519.pdf.) Indus River Basin Land Use/Land Cover (LULC) and irri-
Thenkabail, P. S., Biradar, C. M., Noojipady, P., Dheeravath, V., gated area mapping using continuous streams of MODIS
Li, Y. J., Velpuri, M., Gumma, M. et al. (2009b). Global data. Remote Sensing of Environment, 95 (3), 317–341.
irrigated area map (GIAM), derived from remote sensing, Thenkabail, P. S. and Wu, Z. (2012). An automated cropland clas-
for the end of the last millennium. International Journal of sification algorithm (ACCA) for Tajikistan by combining
Remote Sensing, 30 (14), 3679–3733. July, 20, 2009. landsat, MODIS, and secondary data. Remote Sensing, 4
Thenkabail, P. S., Enclona, E. A., Ashton, M. S., Legg, C., Jean De (10), 2890–2918.
Dieu, M. (2004). Hyperion, IKONOS, ALI, and ETM+ sen- Tillman, D., C. Balzer, J. Hill, and B. L. Befort, 2011. Global
sors in the study of African rainforests. Remote Sensing of food demand and the sustainable intensification of agri-
Environment, 90, 23–43. culture, Proceedings of the National Academy of Sciences of
Thenkabail, P. S., GangadharaRao, P., Biggs, T., Krishna, M., and the United States of America, November 21. doi:10.1073/
Turral, H., (2007a). Spectral matching techniques to deter- pnas.1116437108. Tomlinson, R. 2003. Thinking about
mine historical land use/land cover (LULC) and irrigated Geographic Information Systems Planning for Managers,
areas using time-series AVHRR pathfinder datasets in the 283 ESRI Press.
Krishna river Basin, India. Photogrammetric Engineering Turral, H., Svendsen, M., and Faures, J. (2009). Investing in
and Remote Sensing, 73 (9), 1029–1040. (Second Place irrigation: Reviewing the past and looking to the future.
Recipients of the 2008 John I. Davidson ASPRS President’s Agricultural Water Management. 97(2010), 551–560.
Award for Practical papers). doi:101.1016/j.agwt.2009.07.012.

© 2016 Taylor & Francis Group, LLC

 Global Food Security Support Analysis Data at Nominal 1 km (GFSAD1km) 159

UNDP (2012). Human Development Report 2012: Overcoming Wardlow, B. D., Kastens, J. H., and Egbert, S. L. (2006). Using USDA
Barriers: Human Mobility and Development. New York: crop progress data for the evaluation of greenup onset date
United Nations. calculated from MODIS 250-meter data. Photogrammetric
Velpuri, M., Thenkabail, P. S., Gumma, M. K., Biradar, C. B., Engineering and Remote Sensing, 72, 1225–1234.
Dheeravath, V., Noojipady, P., and Yuanjie, L. (2009). Wu, Z., Thenkabail, P. S., and Verdin, J. (2014b). An auto-
Influence of resolution or scale in irrigated area mapping and mated cropland classification algorithm (ACCA) using
area estimations. Photogrammetric Engineering and Remote Landsat and MODIS data combination for California.
Sensing (PE&RS). 75 (12), December 2009, 1383–1395. Photogrammetric Engineering and Remote Sensing, 80 (1):
Vinnari, M. and Tapio, P. (2009). Future images of meat con- 81–90.
sumption in 2030. Futures, 41 (5), 269–278. doi:10.1016/j. Wu, Z., Thenkabail, P. S., Zakzeski, A., Mueller, R., Melton, F.,
futures.2008.11.014. Rosevelt, C., Dwyer, J., Johnson, J., and Verdin, J. P. (2014a).
Wada, Y., van Beek, L. P. H., and Bierkens, M. F. P. (2012). Seasonal cultivated and fallow cropland mapping using
Nonsustainable groundwater sustaining irrigation: A modis-based automated cropland classification algorithm.
global assessment. Water Resources Research, 48, W00L06. Journal of Applied Remote Sensing 0001;8 (1), 083685.
doi:10.1029/2011WR010562. doi:10.1117/1.JRS.8.083685.
Wardlow, B. D. and Egbert, S. L. (2008). Large-area crop mapping Yu, L., Wang, J., Clinton, N., Xin, Q., Zhong, L., Chen, Y.,
using time-series MODIS 250 m NDVI data: An assess- and Gong, P. (2013). FROM-GC: 30 m global cropland
ment for the U.S. Central Great Plains. Remote Sensing of extent derived through multisource data integration.
Environment, 112, 1096–1116. International Journal of Digital Earth. 12pp. doi:10.1080/
Wardlow, B. D., Egbert, S. L., and Kastens, J. H. (2007). Analysis 17538947.2013.822574.
of time-series MODIS 250 m vegetation index data for
crop classification in the U.S. Central Great Plains. Remote
Sensing of Environment, 108, 290–310.

© 2016 Taylor & Francis Group, LLC