Remote Sensing Applied to Irrigation Performance Assessment

in two Irrigation Schemes in India

Authors: Dr. Poolad Karimi (World Bank, Washington DC, US), Dr. Anju Gaur (World Bank, Dehli, India),
Dr. Silvan Ragettli (hydrosolutions GmbH, Zurich, Switzerland)
July 5, 2024

Highlights
• New online earth observations web platform for two pilot irrigation schemes in India
delivers timely high-resolution information on irrigation performance.
• Comprehensive evaluation of irrigation performance in Purna and NLBC irrigation schemes
based on remote sensing reveals fundamental differences in crop and water management
strategies, influencing their overall effectiveness and sustainability.
• Purna demonstrates a productive but less water-intensive approach, contributing to more
equitable water distribution and higher cropping intensity compared to NLBC, which
primarily focuses on water-intensive crops like paddy rice.
• Crop mapping in Purna indicates a potential underreporting of water-intensive crops like
sugarcane. Remote sensing can help improve the reliability of data used for water resource
management and irrigation practices.

Introducing the CropMapper India Web Application

In India, irrigation is playing a pivotal role in ensuring food security and fostering rural development.
Given the vastness and diversity of the country's agricultural landscapes, the management, operations,
and maintenance (MOM) of its large-scale irrigation and drainage (I&D) infrastructure present
significant challenges. These systems are crucial for optimizing agricultural productivity and
sustainability, underlining the importance of efficient water resource management in one of the
world's largest agrarian economies.
Using earth observations (EO) via satellite remote sensing and geographic information systems,
irrigation performance mapping can help prioritize sectoral MOM investments by directing them to
locations with the highest needs. The World Bank has supported the development of a pilot online
earth observations web platform for irrigation performance mapping in two pilot schemes: Purna in
Maharashtra and Narayanpur Left Bank Canal (NLBC) in Karnataka. It is available online and can be
accessed here CropMapper India. Figure 1 shows a screenshot and an explanation of the main
elements of the user interface.
The CropMapper India is a Google Earth Engine (GEE) Application (Gorelick, et al., 2017). The Web
App allows access, visualization, summarizing, and downloading crop-disaggregated annual maps of
irrigated areas at 10 meters (m) resolution. The resulting annual and seasonal maps are available
between June 2018 and May 2023. The App also provides access to crucial aggregate statistics for a
user-selected year and irrigation scheme. Statistics on the total cropped and irrigated area, the crop-
specific agricultural productivity (total biomass production TBP), and the crop-specific water balance
(precipitation P (mm), actual total evapotranspiration ET (mm), evaporation from precipitation (ET
green; mm), evapotranspiration from irrigation (ET blue; mm) are calculated for the corresponding
scheme.
1
This report presents selected critical results from applying the CropMapper India web application to
the assessment of irrigation performance in the region. A separate, detailed report provides further
information on the methodologies applied (World Bank, 2024).

Figure 1: Screenshot of the CropMapper India web application. (1) User input and selection panel, (2) Crop map legend showing color-
coded crop list, (3) Map comparison selection panel, (4) Interactive map window with crop-disaggregated crop map shown in the left
part and a user-selected dataset shown on the right side (5). (6) shows the output panels with a cropped area bar chart for the chosen
domain, a table containing a crop-specific water balance (7), and signature chart data that can be selected by the user (8). All data can
be exported as georeferenced maps (9).

Development of Irrigated Areas and Cropping Patterns, 2018 - 2023

The basis for the crop-based disaggregation of the irrigated area maps is an unsupervised clustering
of Sentinel-1 and Sentinel-2 bi-monthly image data within the command areas of irrigation schemes
and adjacent areas. The machine learning approach can automatically distinguish pixels with different
phenologies, vegetation structure or soil moisture. Examples are classes with a characteristic Kharif-
Rabi growth phase (e.g., cotton), Rabi seasonal crops (e.g., sorghum) hot season (Zaid) crops (e.g.,
groundnut), perennial crops such as sugarcane, or two consecutive growth periods of the same crop
(paddy rice). To facilitate the labelling of automatically identified classes, in-situ crop type samples are
available (NLBC) or data on crop wise, type wise irrigated area per scheme (Purna) that can be
correlated with the frequency of unlabelled crop classes. The following class labels are used in the final
crop maps: cotton, paddy, wheat, redgram, chickpea, sorghum, chili, groundnut, and sugarcane. Pixels
that could not unambiguously be attributed to one of the 9 main crop categories were divided into a
“other rainfed” class and two “other irrigated” classes (high water intensive and less water intensive
crops, respectively).

2
Figure 2: a): Location of pilot schemes. b) Crop map, Rabi season 2022. c) Irrigated area map, Rabi season 2022.
Irrigated areas are identified within the cropped area pixels based on a water balance approach. In a
first step the mean actual evapotranspiration (ETa) of clearly non-irrigated classes (identified through
satellite image interpretation) is calculated per month and scheme. Considering the vegetation period
of each cropped pixel, the corresponding total ETa is compared to rainfed ETa (also named ET green)
during the same period. If ETa of cropped pixels is substantially higher than ET green (ETa – ET green
> 10 mm) the difference is called ET blue and is equivalent to crop irrigation water use. Such pixels can
therefore be labelled as irrigated. This thresholding approach has been made possible thanks to recent
advances in satellite-based ET mapping. For ET blue calculations the Landsat Collection 2 Level-3
Provisional Science Product is used. This product uses the most advanced version of the Operational
Simplified Surface Energy Balance (SSEBop) model (Senay et al., 2023) to retrieve the daily total of ETa.
A recent accuracy assessment (Volk et al., 2024) shows that the bean bias of this product is less than
10% for cropland sites. The final outputs of the crop-based disaggregation are seasonal crop and
irrigated area maps with a resolution of 10 meters.
Seasonal crop and irrigated area maps for Kharif (June – September), Rabi (October – January) and
Zaid (February – May) seasons were generated for each of the two schemes Purna and NLBC and every
year between June 2018 and May 2023. Figure 2 presents the results for the Rabi season of the year
2022. In addition, cropping intensity maps were generated for each of the five years of analysis. The
cropping intensity maps indicate for each pixel the number of harvests per year and associates each
crop to one season or indicates if the crop was perennial.
Within NLBC command area boundaries a total arable area of 593,000 hectares (ha) has been
identified. The arable area is defined as the area that was classified as cropped at least once during
the 5-year period. In Purna, this area amounts to 78,000 ha. The total irrigable land classified is 531,000
ha in NLBC (67,000 ha in Purna). This means that 90% (85%) of the arable land was identified as
irrigated at least once during the 5-year period.
In NLBC on average 65% of the annual irrigated area is used for single-crop Kharif crop rotations (also
including two-seasonal Kharif/Rabi crops). On average 22.5% of the annual irrigated area is used for
double crop Kharif/Zaid crop rotations. The most common Kharif crop was paddy rice (95,000 ha or
21% of the cropped area), followed by cotton (66,000 ha or 14%) and redgram (37,000 ha or 8%).
During Zaid 68,000 ha (55%) of the irrigated land is used on average for paddy rice cultivation.
In Purna the annual cropping patterns are more heterogeneous. About one fourth of the irrigated area
(25%) is used each for Kharif single-crop rotations, Kharif/Rabi double-crop rotations and Kharif/Zaid
double-crop rotations, respectively. Due to persistent cloud cover it is difficult to label cropped pixels
with a short vegetation period during Kharif. Therefore, during this season, 55% of the irrigated crops
(on average 4,900 ha) are just labelled as Other Irrigated (Less Water Intensive), and another 10% (900
3
ha) as Other Irrigated (High Water Intensive). In absolute terms, Rabi sorghum is the most common
irrigated crop class (7,000 ha on average), followed by sugarcane (perennial crop; 5,000 ha on average)
and Kharif-Rabi two-seasonal cotton (4000 ha).
A strong decline in total irrigated area was identified across both schemes and all season between
June 2018 and May 2019. This decline in agricultural production is confirmed by media reports1 and
by crop statistics available from Purna irrigation scheme. According to the remote sensing analysis, the
annual irrigated area in that year was 25% lower than average in NLBC and 39% lower than average in
Purna. Irrigated areas during Zaid dropped by 77% and 69%, respectively. During other years, the
annual irrigated area varied by less than ±11% in both schemes, and the Zaid irrigated areas varied by
less than ±22%.
Table 1. Summary of main irrigated crops in Purna and NLBC schemes. The values represent annual averages over the five-year period
2018-2022, considering the crop-specific vegetation period of each crop. TBP is total biomass production, ET is total actual
evapotranspiration and ET blue is the portion of ET supplied by irrigation. Missing values for specific crops in the data indicate that these
crops appear infrequently in a given scheme and could not be reliably classified as distinct categories.

Crop Type (irrigated) Area (ha) Rainfall (mm) TBP (kg/ha) ET (mm) ET blue (mm)
Purna NLBC Purna NLBC Purna NLBC Purna NLBC Purna NLBC
Cotton 4,168 61,737 707 553 13,428 9,403 465 445 95 113
Paddy (one harvest) 21,546 550 8,315 422 130
Paddy (two harvests) 73,638 560 13,244 945 556
Wheat 2,964 90 8,250 291 90
Red gram 32,089 545 7,511 373 64
Chickpea 3,442 55 6,562 207 48
Sorghum 8,317 18,840 62 123 7,017 5,490 241 262 75 61
Chili 11,056 510 12,943 647 261
Groundnut 4,248 32 4,372 424 264
Sugarcane 4,681 32,414 852 675 17,087 13,876 984 889 452 422

Validation of Irrigated Areas and Cropping Patterns, 2018 - 2023

The crop statistics from CropMapper India align well with the official figures from Purna (Figure 3). The
discrepancy in the multi-annual average of irrigated areas, as derived from remote sensing versus
reported figures, is +9% for the Kharif season, +13% for Rabi, and -7% for Zaid. The mean absolute
annual bias is highest for the Kharif season at 75%, followed by Rabi at 55%, and Zaid at 26%. This
indicates a greater uncertainty in remote sensing during the monsoon season due to the scarcity of
cloud-free satellite images. The notably high average bias for the Rabi season can be attributed to the
discrepancies observed in the 2018 Rabi season data. Although both data sets recorded lower than
average irrigated areas for that year, remote sensing depicted a significantly less severe decline
compared to official statistics. From 2019 to 2022, the mean average bias for the total irrigated area
during the Rabi season was only 16%.
Remote sensing effectively captures the relative distribution of crop types within irrigated areas
(Figure 4). It tends to identify larger sugarcane areas compared to official reports (Figure 3b). During
the Zaid season, remote sensing recognized nearly 100% of the sugarcane area as irrigated, exceeding
the reported irrigated sugarcane area by 55%. Among the crop types that could not be identified

1
https://www.indiatoday.in/india/story/drought-maharashtra-fall-kharif-rabi-crops-sugarcane-
production-1550882-2019-06-18
4
through remote sensing, the most common in Purna are soybean and turmeric/chili (37% and 15% of
Kharif irrigated area, respectively, according to official data), and grass (20% of Zaid irrigated area).
a) 40'000 Remote Sensing
Validation
b) 10'000 Validation, Irrigated
Remote Sensing, Rainfed
Remote Sensing, Irrigated
Irrigated Area [ha]

30'000

Sugarcane [ha]
20'000 5'000

10'000

0
0

Kh rif 2 8
Kh rif 2 9
Kh rif 2 0
1
2
8
9
0
1
id 22
id 1 8
id 1 9
id 2 0
id 2 1
22
a 01
a 01
a 02
ar 02
ab 2
ab 1
ab 1
ab 2
ab 2
id 1 8
id 1 9
id 2 0
id 2 1
22
Kh rif 2 8
Kh rif 2 9
Kh rif 2 0
1
2
8
9
0
i 1
id 22

R f 20
R 20
R 20
R i 20
R 20
Za 20
Za 20
Za 20
Za 20
Za 20
20
a 01
a 01
a 02
ar 02
ab 2
ab 1
ab 1
ab 2
ab 2

Kh rif 2
R f 20

Za 20
Za 20
Za 20
Za 20
20
R 20
R 20
R i 20
R i 20
Za 20

i
i

i
i
Kh rif 2

i
a
i
i
i

Kh
a
Kh

Figure 3. a) Validation of total irrigated area per season and year identified through remote sensing against reported data from the
Central Water Commission (CWC). b) Validation of sugarcane area per season and year against CWC data. Seasonal rainfed sugarcane
areas are not available from CWC.

a) 100%
20% 14% 22% 23%
Irrigated Area [%]

75% 13%
14% 16%
65% 64% 28%
24%
50% 26%
27%
19% 27%
14%
25% 19% 13%
7% 10% 24%
14% 34%
16% 17% 16%
6% 8%
0%
i
if

d
i

id
rif

ab

ab
ar

ai
Za
ha

.Z
.R
Kh

.K

S
S

lid
lid
S

R
R
lid

Va
R

Va
Va

Cotton Wheat Chickpea Sorghum

Groundnut Sugarcane Other
Figure 4. a) Distribution of crop types within Purna irrigated area per season, as determined by remote sensing (RS) and validated against
CWC data. b) Average annual volume of irrigation water consumed by crops (in million cubic-meters of water) and its distribution among
different crop types.
Less Water Intensive
a) Purna High Water Intensive
b) NLBC Less Water Intensive
High Water Intensive
3% 400'000 74%
30'000 4%
Irrigated Area [ha]

Irrigated Area [ha]

6%4% 73%
300'000 53% 56%
98% 2%
20'000 32% 97% 3%3% 60%
98% 200'000 5% 97% 97%
14%6% 99% 45% 98% 98%
10'000 12% 13%0% 100'000
98% 100%
0
0
18

20

22
18

20

22

1
01

02
id 1 8
id 1 9
id 2 0
id 2 1
22
Kh rif 2 8
Kh rif 2 9
Kh rif 2 0
1
2
8
9
0
1
id 22

20

20

20

20

20

20
a 01
a 01
a 02
ar 02
ab 2
ab 1
ab 1
ab 2
ab 2

i2

i2
R f 20

Za 20
Za 20
Za 20
Za 20
20
R 20
R 20
R i 20
R 20
Za 20

id

id

id
if

if

if
Kh rif 2

ab

ab
ar

ar

ar

Za

Za

Za
i
i

i
i
i

R
Kh

Kh

Kh
a
Kh

Figure 5. Area of high and less water intensive irrigation, respectively, per season and year (2018-2022) in a) Purna) and b) NLBC. Labels
indicate the fraction of high water intensive irrigated area.

5
Assessment of Irrigation Performance
The remote sensing information on irrigated agricultural areas is combined with state-of-the-art
datasets on evapotranspiration and biomass production to perform an irrigation performance
assessment (IPA). To ensure maximum accuracy, the biomass production and evapotranspiration
layers have been downscaled to match the resolution of the crop and irrigated area maps (Onačillová
et al., 2022). The total volume of water used by the crops from irrigation has been determined based
on the seasonal and annual maps of ET blue. Based on these calculations, the total volume of irrigation
water used by crops in Purna annually is 65 m3 and 844 m3 in NLBC (Figure 4b). High water intensive
irrigation (> 1000 m3/ha crop irrigation water use) specifically takes place during the hot season in
both schemes (Figure 5). In NLBC, high water intensive irrigation is also necessary during Rabi season
(Figure 5b) due to the crop water requirements of paddy rice, which consumes 50% of the total
irrigation water in that scheme (Figure 4b). It is important to recognize that total irrigation water use
exceeds the actual water consumption by crops due to losses that occur during transport and at the
field level. Crop irrigation water use as determined from remote sensing can be instrumental in
calculating irrigation efficiency when combined with water use data. However, no water use data were
available for this particular study.
The goal of assessing irrigation performance is to comprehensively evaluate and improve various
aspects related to irrigation practices. This involves analyzing cost-effectiveness and equitable access
to irrigation water, productivity of water use in irrigation, and adequacy and reliability and availability
of water for irrigation in a sustained hydrological ecosystem. To facilitate a comprehensive assessment
of irrigation performance, a set of eight indicators has been defined, ensuring coverage of various
aspects. The IPA scheme, initially developed by a World Bank team in 2020 to assess irrigation
performance across ten large-scale schemes in India (World Bank, 2020), has been customized for this
project. Adaptations were made to meet stakeholder needs and to optimize the use of available
remote sensing layers. Table 2 presents the indicators, along with their descriptions, associated
benchmark measures, and units of measurement.
Table 2. Summary of irrigation performance assessment indicators and their associated scores for Purna and NLBC, respectively. The
‘Critial Value' and the 'Target Value' represent the global range of values considered for each indicator. Values calculated for each scheme
represent the annual average over the period 2018-2022. A score of 1 corresponds to the critical value, while a score of 10 is assigned to
the target value. The score for ‘Productivity’ represents the average of the productivity scores for cotton, sorghum, paddy rice and
sugarcane, respectively.
Critical Target Value Value Score Score
Indicator Description Units Value Value Purna NLBC Purna NLBC
1 Cropping Average number of harvests per year - 1.00 1.50 1.49 1.19 10 4
Intensity
2 Irrigated Area Fraction of irrigable area %/100 0.25 0.80 0.56 0.65 6 8
3 Irrigation Fraction of less water intensive irrigated %/100 0.00 1.00 0.55 0.28 6 4
Intensity crops, considering potential ET (ETp) of
each crop and a threshold of 500 mm.
4 Adequacy RWD=1-(ETa/ETp) - 0.40 0.00 0.24 0.23 5 5
5 Equity Standard deviation of ETa over irrigated - 0.40 0.00 0.17 0.23 6 5
areas divided by average ETa
6 Productivity Water Productivity (in terms of yield) 7.7 4.0
Cotton kg/m3 0.26 0.42 0.37 0.28 7 2
Sorghum kg/m3 0.53 0.82 0.75 0.61 8 3
Paddy kg/m3 0.63 0.72 0.69 7
Sugarcane kg/m3 4.14 5.23 4.96 4.45 8 4
7 Reliability Range of average crop-specific monthly %/100 0.40 0.00 0.10 0.14 8 7
Relative ET (ET/ETp)
8 Yield Gap due (Ypotential−Yactual) / (Ypotential) %/100 0.40 0.00 0.18 0.23 6 5
to water stress
AVERAGE SCORE 6.8 5.3

6
The following insights were drawn from the assessment of irrigation performance based on indicators
presented in Table 2:
• Double cropping is more common in Purna than in NLBC. On average, 49% of the annual
agricultural area in Purna sees two harvest per year, as compared to 19% in NLBC. Transitioning from
single to double cropping could enhance agricultural land productivity but necessities system
improvements to fulfil additional irrigation requirements.
• The irrigation infrastructure has not been used to full capacity in recent years. On average, 56%
of the irrigable area within Purna scheme has been used annually for irrigation. In NLBC, this fraction
is slightly higher (65%).
• Purna scheme has a higher fraction of less water intensive irrigated crops as compared to NLBC
(55% as compared to 28%). This means that the irrigation water demand in NLBC is high due to a larger
share of high-water intensive crops. Replacing high water intensive crops by less water intensive crops
could alleviate crop water stress during periods of water scarcity and facilitate a more equitable spatial
distribution of irrigation water.
• To assess irrigation adequacy, actual crop water use was compared with potential crop water
use. This assessment has revealed that the relative water deficit is on average 23% in both schemes,
which results in an average score of 5 on a scale from 1 to 10. This indicates that while the availability
of irrigation water is adequate, there is room for improvement.
• A more uniform distribution of water consumption within a scheme can help to mitigate water
stress. This can be achieved by substituting irrigation intensive cultivations such as paddy rice, which
require more than two times the irrigation volume of sorghum or red gram (Table 1). The equity of the
spatial distribution of crop water use can be improved in both schemes.
• The crop water productivities per command area revealed significantly higher productivity for
cotton, sorghum, and sugarcane in Purna compared to NLBC. However, NLBC displayed relatively high
productivity for paddy rice. Water management strategies could be optimized by adjusting crop
choices based on water productivity data. For instance, promoting crops like cotton in regions like
Purna where they are more water-efficient than in NLBC could lead to more sustainable water use
across these schemes.
• The reliability of service indicator shows if water supply is reliable over the course of the
irrigation season. Both assessed schemes get relatively high scores for this indicator, meaning that the
irrigation water supply can be considered consistent throughout the season. This suggests that the
infrastructure and management practices in place are effective in ensuring water availability for
agricultural needs during critical growth periods.
• In both schemes, well-irrigated crops, where crop water use matches potential crop water use,
show an average yield increase of 18%-23%. This demonstrates the benefits of optimal water
management in maximizing agricultural output.
The assessment of irrigation performance has resulted in an overall higher score for Purna than for
NLBC (6.8 as compared to 5.3, see Table 2). The difference in overall score is due to a more productive
and less water intensive irrigated agriculture in Purna, allowing for a relatively equitable distribution
of water ressources and a high cropping intensity. In NLBC, the irrigation infrastructure is well utilized
(resulting in a high score for indicator 2), but there is a strong emphasis on high water intensive crops
such as paddy rice. While this approach benefits rice cultivation, it may not promote sustainable water
usage. Addressing this could enhance NLBC's water management strategies and possibly improve their
overall irrigation performance score by diversifying crops and optimizing water use.

7
The mapping of crops has revealed that irrigation intensive crops such as sugarcane are likely
underreported in Purna (Figure 3b). Remote sensing can help to identify such discrepancies. This
capability enhances the reliability of agricultural data, which is crucial for informed water resource
management and ensuring the efficiency of irrigation practices.

Conclusions
The delivery of timely, accurate, continuous high-resolution information of land surface variables and
water balance components using open-source satellite-based remote sensing revolutionizes irrigation
management in the global drylands. EO data in irrigation support scheme MOM for the evaluation &
increase of on-farm water use efficiency and crop productivity and enable monitoring irrigation policy
compliance.
India can benefit similarly from EO-derived intelligence. The advantage of a technology like the
CropMapper India is that no local infrastructure, beyond a computer connected to the internet, is
required to access data and intelligence relevant for irrigation improvements. The software can be
specifically tailored to stakeholders' specific needs and wants at all levels, for water consumers and
service providers at district, state, and national levels alike.

References
Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., Moore, R., 2017. Google Earth Engine: Planetary-scale geospatial analysis
for everyone. Remote Sens. Environ. 202, 18–27. https://doi.org/10.1016/j.rse.2017.06.031
Onačillová, K., Gallay, M., Paluba, D., Péliová, A., Tokarčík, O., & Laubertová, D. (2022). Combining landsat 8 and sentinel-2 data in google
earth engine to derive higher resolution land surface temperature maps in urban environment. Remote Sensing, 14(16), 4076.
https://doi.org/10.3390/rs14164076
Senay, G.B., Parrish, G.E., Schauer, M., Friedrichs, M., Khand, K., Boiko, O., Kagone, S., Dittmeier, R., Arab, S. and Ji, L., 2023. Improving
the operational simplified surface energy balance evapotranspiration model using the forcing and normalizing operation. Remote
Sensing, 15(1):260. https://doi.org/10.3390/rs15010260
Volk, J.M., Huntington, J.L., Melton, F.S., Allen, R., Anderson, M., Fisher, J.B., Kilic, A., Ruhoff, A., Senay, G.B., Minor, B. and Morton, C.,
2024. Assessing the accuracy of OpenET satellite-based evapotranspiration data to support water resource and land management
applications. Nature Water, 1-13. https://doi.org/10.1038/s44221-023-00181-7
World Bank., 2020. Towards a Remote Sensing Guided Performance Measurement System for the Indian Irrigation Sector, Project
Report. Maria Iskandarani, Mutlu Ozdogan, and IJsbrand H. de Jong, https://icid-ciid.org/icid_data_web/ReportTRSGP-
MeasurementSystem.pdf
World Bank, 2024. Remote Sensing Applied to Irrigation Performance Assessment in India – Technical Report. Poolad Karimi, Silvan
Ragettli.