Remotely sensed time- series (2000-2018) estimation of evapotranspiration in the Indus Basin Implementation, evaluation and analysis

Abstract

Evapotranspiration is the largest water balance component in semi-arid irrigated systems. The Indus Basin Irrigation System (IBIS, ~140,000 km2) is the largest irrigated system in the world. Remote sensing can provide consistent and robust spatial estimates of evapotranspiration at spatiotemporal scales (<1000 m and monthly) that can be used to estimate the water balance and the performance of irrigated systems at the canal command scale. This study evaluated the skill of the CMRSET (CSIRO MODIS ReScaled EvapoTranspiration) algorithm to estimate actual evapotranspiration (ETa) in the 56 canal command areas of the IBIS in Pakistan over the period 2000 to 2018 at 10-day temporal and 500 m spatial resolution. This algorithm was selected as it only requires multi-temporal remote sensing imagery to derive a crop factor and meteorological data for its implementation. To implementation was facilitated by pre-processing satellite reflectance data using the geospatial analysis tool and remote sensing data repository Google Earth Engine. Unlike previous studies of limited duration or spatial domain, these time-series provide the first long-term (>15 years) consistent ETa time-series for the entire IBIS at spatial and temporal resolutions that are useful to assess irrigation systems at the canal command scale.
To assess CMRSET’s accuracy and therefore its usefulness for water balance modelling and other applications, its estimates were evaluated against existing estimates from two remotely sensed ETa products, SEBAL and ETLook at the pixel and canal command scale, and against ground ETa measurements at the pixel scale – to assess its accuracy. SEBAL and ETLook were implemented for the year October 2004 to September 2005 and for the calendar year 2007, respectively, and had a comparable spatial and temporal resolution as CMRSET. SEBAL was implemented in the northern part (covering mainly Punjab canal commands) of the study region, and ETLook for the whole Indus Basin (which includes the entire IBIS), therefore the assessment was conducted in 40 canal commands for SEBAL, and 56 canals commands for ETLook. Generally, CMRSET compared well against both datasets, both in terms of magnitude and temporal patterns. CMRSET agreed better to SEBAL in terms of magnitude, with a mean Pearson’s correlation coefficient r of 0.85 (min of 0.54 and max of 0.97), a mean absolute percentage bias of 7.6% (min of -12.5% and max of 27.8%), and a mean RMSD of 18.0 mm/mo (min of 9.4 mm/mo and max of 26.7 mm/mo), and no canal command had a bias greater than ±30%. For ETLook, the mean Pearson’s correlation coefficient r was 0.93 (min of 0.68 and max of 1.00), the mean absolute percentage bias was 21.4% (min of -23% and max of 77%), and the mean RMSD was 20.5 mm/mo (min of 7.3 mm/mo and max of 49.8 mm/mo). Whilst the temporal patterns were well captured, the resulting magnitudes seemed to be mixed, with 14 canal commands having a percentage bias larger than +30%. CMRSET ETa estimates were also assessed against two in situ Bowen ratio surface energy balance ETa measurements, that were set up from July 2000 to March 2001 (9 months, during the growing season) in two locations in the Punjab Province. In these locations, CMRSET was also implemented using Landsat (30 m) bands and the same CMRSET model parameters to assess scale differences related to pixel smearing and averaging in the coarser MODIS data (500 m). MODIS CMRSET showed reasonable agreement both in magnitude and seasonality considering the scale differences. For both locations, the Pearson’s correlation coefficient r was greater than 0.92, the percentage bias less than 20% and the RMSD less than 17 mm/mo. Results were markedly better for Landsat CMRSET estimates at both locations with bias in both cases being lower than 5%, although with some seasonal compensation reducing bias and RMSD errors.
Considering the differences between CMRSET and the existing remotely sensed products, particular care is required in the use of ETa for a quantitative water assessment or water balance analyses that uses absolute values, such as the assessment of sustainable groundwater use. Ideally, the results of related applications should be cross-checked for the presence of biases or inconsistencies in relation to the use of this or any other remotely sensed ETa products. This report is companion to a report that assesses future scenarios impact on irrigated agriculture using the remotely sensed ETa products for scenario exploration in a way that the differences do not exert an undue influence in the interpretation of results.
Monthly CMRSET ETa estimates were used to assess how ETa changed spatially and temporally in the lower IBIS canal commands during the 2000‒2018 period. Temporal resolution was annual (April–March water year) and seasonal – wet Kharif (April to September) and dry Rabi (October to March). The assessment showed that ETa in most irrigated areas within the canal commands exceeded 600 mm/y, with some areas that exceeded 1000 mm/y, particularly in rice canal commands in the Sindh Rice Wheat agro-climatic zone. Lower mean annual and Kharif ETa occurred in the Sindh Cotton Wheat South agro-climatic zone and Sindh Rice Wheat South agro-climatic zone canal commands, with around 400 mm to 500 mm (about 100 to 200 mm less) during Kharif. On the other hand, during Rabi, most irrigated areas exceeded 300 mm.
The pre-processing capabilities of Google Earth Engine and continuous update of its satellite imagery catalogue, plus the straightforward implementation of CMRSET ETa, potentially on a continuous basis, provides an opportunity for monitoring irrigation dynamics and the assessment of structural and policy improvements in the IBIS.

Full text

Australia’s National Science Agency
Remotely sensed time-
series (2000‒2018)
estimation of
evapotranspiration in the
Indus Basin
Implementation, evaluation and analysis
Jorge L. Peña-Arancibia, Mobin-ud Din Ahmad, Mac Kirby and Muhammad J.M. Cheema
February 2020

ii | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
Citation
Peña-Arancibia JL, Ahmad MD, Kirby JM and Cheema MJM (2020) Remotely sensed time-series (2000‒2018)
estimation of evapotranspiration in the Indus Basin: Implementation, evaluation and analysis. Sustainable
Development Investment Portfolio (SDIP) project. CSIRO, Australia. 34 pp
Author affiliations
Jorge L. Peña-Arancibia and Mobin D. Ahmad are research scientists at CSIRO Land and Water, Canberra,
Australia.
Mac Kirby is a visiting research scientist at CSIRO Land and Water, Canberra.
Muhammad J.M. Cheema is an Associate Professor at University of Agriculture Faisalabad.
Copyright
With the exception of the Australian government crest, and the Australian Aid and CSIRO
logos, and where otherwise noted, all material in this publication is provided under a Creative Commons
Attribution 4.0 International License http://creativecommons.org/licenses/by/4.0/legalcode
The authors request attribution as ‘© Australian Government Department of Foreign Affairs and Trade
(DFAT) and CSIRO’.
Important disclaimer
CSIRO advises that the information contained in this publication comprises general statements based on scientific
research. The reader is advised and needs to be aware that such information may be incomplete or unable to be used
in any specific situation. No reliance or actions must therefore be made on that information without seeking prior
expert professional, scientific and technical advice. To the extent permitted by law, CSIRO (including its employees and
consultants) excludes all liability to any person for any consequences, including but not limited to all losses, damages,
costs, expenses and any other compensation, arising directly or indirectly from using this publication (in part or in
whole) and any information or material contained in it.
Ethics
The activities reported herein have been conducted in accordance with CSIRO Social Science Human Research Ethics
approval 011/17.
This report designed and implemented by CSIRO contributes to the South Asia Sustainable Development
Investment Portfolio and is supported by the Australian aid program. Further details on CSIRO SDIP projects
are available from http://research.csiro.au/sdip.
SDIP’s goal is increased water, food and energy security in South Asia to support climate resilient livelihoods and
economic growth, benefiting the poor and vulnerable, particularly women and girls
SDIP 2020 objective: Key actors are using and sharing evidence, and facilitating private sector engagement, to improve
the integrated management of water, energy and food across two or more countries - addressing gender and climate
change.
All CSIRO SDIP projects consider gender. In this report we have assumed that an improved, quantitative
understanding of spatial evapotranspiration is of benefit to all, regardless of gender and other social factors.
Excluding gender analysis, however, can lead to ‘gender blind’ tools, findings and decisions that reinforce
existing gender inequities. This gap should be borne in mind when interpreting this report, and any
application of its findings will need to integrate gender-specific and other social considerations to ensure
benefits are distributed equitably.

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | iii
Contents
Acknowledgments.......................................................................................................................................... v
Executive summary ....................................................................................................................................... vi
1 Introduction ..................................................................................................................................... 1
2 Study region .................................................................................................................................... 2
3 Methods and materials .................................................................................................................... 4
3.1 Long-term 10-day CMRSET ETa time-series estimates ...................................................... 4
3.2 Comparison of CMRSET ETa against SEBAL and ETLook ....................................................... 5
3.3 Comparison of CMRSET ETa against two locations with Bowen ratio ETa estimates .......... 6
3.4 ETa spatial and temporal dynamics for the IBIS and for canal commands ........................ 6
4 Results and discussion ..................................................................................................................... 7
4.1 Long-term 10-day CMRSET ETa time-series estimates ...................................................... 7
4.2 Comparison of CMRSET ETa against two RS products ...................................................... 10
4.3 Comparison of CMRSET ETa against two locations with Bowen ratio ETa estimates ........ 15
4.4 ETa spatial and temporal dynamics in the lower IBIS ....................................................... 15
5 Summary and conclusion ............................................................................................................... 19
A.1 Comparison of CMRSET ETa against ETLook ..................................................................................... 20
A.2 Comparison of CMRSET ETa against SEBAL .................................................................................... 22
References ........................................................................................................................................ 24

iv | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
Figures
Figure 2-1 Geographic characteristics of the Indus Basin Irrigation System including main rivers, canal commands
and location of two Bowen ratio stations ............................................................................................................................ 2
Figure 2-2 Canal commands and the corresponding agro-climatic zones (ACZ) in the IBIS. Canal commands are
labelled from north to south................................................................................................................................................. 3
Figure 4-1 Comparison of monthly CMRSET ETa time-series for the year 2013 aggregated to the canal command
scale for 56 canal commands. CMRSET using daily gridded 2.5 km ETp surfaces (blue lines) and CMRSET using
rescaled GLDAS ETp (red circles) ........................................................................................................................................... 7
Figure 4-2 Density plot of the pixel by pixel comparison for 56 canal commands ............................................................ 8
Figure 4-3 Spatial ETa time-series for each hydrologic year (April to March) from 2000‒2001 to 2017‒2018 ............... 9
Figure 4-4 Comparison of monthly CMRSET (solid blue line) and ETLook (dashed blue line) ETa time-series for the
calendar year 2007 aggregated to the canal command scale for 56 canal commands. The Enhanced Vegetation Index
(EVI) aggregated time-series is shown for reference (green line) .................................................................................... 11
Figure 4-5 Density scatterplots of monthly pixel by pixel CMRSET (X-axis) and ETLook (Y-axis) ETa for the calendar year
2007 and for 56 canal commands ...................................................................................................................................... 12
Figure 4-6 Comparison of monthly CMRSET (solid blue line) and SEBAL (dashed blue line) ETa time-series for the year
2004‒2005 (October to September) aggregated to the canal command scale for 40 canal commands. The Enhanced
Vegetation Index (EVI) aggregated time-series is show for reference (green line) ......................................................... 13
Figure 4-7 Density scatterplots of monthly pixel by pixel CMRSET (X-axis) and SEBAL (Y-axis) ETa for year 2004‒2005
(October to September) aggregated to the canal command scale for 40 canal commands .......................................... 14
Figure 4-8 Comparison between monthly CMRSET ETa for both MODIS (500 m spatial resolution green dashed line)
and Landsat (30 m spatial resolution, blue dashed line) and ETa from Bowen ratio measurements at two locations in
Punjab: (a) cotton filed at Faisalabad and (b) rice paddy at Pindi Bhattian ..................................................................... 15
Figure 4-9 Spatial ETa characteristics for the 2000‒2018 period in the lower IBIS including: (a) mean annual ETa, (b)
mean Kharif (April to September) ETa and (b) mean Rabi (May to October) ETa. Canal command boundaries are
shown in black. .................................................................................................................................................................... 16
Figure 4-10 Mean seasonal (2000‒2018 period) ETa aggregated for canal commands in the lower IBIS: (a) Kharif
(April to October) and Rabi (November to March) ............................................................................................................ 17
Figure 4-11 Kharif and Rabi ETa hydrologic year (April to March) time-series from 2000 to 2018, aggregated for canal
commands in the lower IBIS ............................................................................................................................................... 18
Tables
Table A-1 Canal command characteristics, corresponding Agro-Climatic Zone (ACZ) and statistics of the CMRSET and
ETLook comparison. Statistics include the Pearson’s correlation coefficient (r), percentage bias, root mean square
differences (RMSD) and the pixel-by-pixel spatial r........................................................................................................... 20
Table A-2 Canal command characteristics, corresponding Agro-Climatic Zone (ACZ) and statistics of the CMRSET and
SEBAL comparison. Statistics include the Pearson’s correlation coefficient (r), percentage bias, root mean square
differences (RMSD) and the pixel-by-pixel spatial r. ......................................................................................................... 22

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | v
Acknowledgments
We would like to acknowledge the support of the Government of Australia through its Department of
Foreign Affairs and Trade (DFAT) in financing this study. We also thank the Australian High Commission in
Islamabad, the Pakistan High Commission in Canberra and Dr John Dore (DFAT Sustainable Development
Investment Portfolio (SDIP) adviser) and Ms Paula Richardson (DFAT SDIP coordinator) for their active
engagement in the project. We would like to thank the Pakistan Ministry of Water Resources and provincial
irrigation departments for their interest, advice and support, including the provision of data and expertise,
hosting the project team on field visits, and provision of guidance throughout the project.
We would also like to acknowledge Dr Bakhshal Khan Lashari, Dr Arjumand Zehra Zaidi and Mr Nabeel Ali
Khan from USPCAS-W at Mehran University of Engineering & Technology, for discussions and their
contributions.
Dong Dong Kong (Sun-Yat Sen University) and Catherine Ticehurst (CSIRO Land and Water) are gratefully
acknowledged for their support with Google Earth Engine.
Dr Juan Pablo Guerschman and Susan Cuddy (CSIRO Land and Water) are acknowledged for revising the
report, which improved its quality and content.

vi | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
Executive summary
Evapotranspiration is the largest water balance component in semi-arid irrigated systems. The Indus Basin
Irrigation System (IBIS, ~140,000 km2) is the largest irrigated system in the world. Remote sensing can
provide consistent and robust spatial estimates of evapotranspiration at spatiotemporal scales (<1000 m and
monthly) that can be used to estimate the water balance and the performance of irrigated systems at the
canal command scale. This study evaluated the skill of the CMRSET (CSIRO MODIS ReScaled
EvapoTranspiration) algorithm to estimate actual evapotranspiration (ETa) in the 56 canal command areas of
the IBIS in Pakistan over the period 2000 to 2018 at 10-day temporal and 500 m spatial resolution. This
algorithm was selected as it only requires multi-temporal remote sensing imagery to derive a crop factor and
meteorological data for its implementation. To implementation was facilitated by pre-processing satellite
reflectance data using the geospatial analysis tool and remote sensing data repository Google Earth Engine.
Unlike previous studies of limited duration or spatial domain, these time-series provide the first long-term
(>15 years) consistent ETa time-series for the entire IBIS at spatial and temporal resolutions that are useful to
assess irrigation systems at the canal command scale.
To assess CMRSET’s accuracy and therefore its usefulness for water balance modelling and other
applications, its estimates were evaluated against existing estimates from two remotely sensed ETa products,
SEBAL and ETLook at the pixel and canal command scale, and against ground ETa measurements at the pixel
scale – to assess its accuracy. SEBAL and ETLook were implemented for the year October 2004 to September
2005 and for the calendar year 2007, respectively, and had a comparable spatial and temporal resolution as
CMRSET. SEBAL was implemented in the northern part (covering mainly Punjab canal commands) of the
study region, and ETLook for the whole Indus Basin (which includes the entire IBIS), therefore the assessment
was conducted in 40 canal commands for SEBAL, and 56 canals commands for ETLook. Generally, CMRSET
compared well against both datasets, both in terms of magnitude and temporal patterns. CMRSET agreed
better to SEBAL in terms of magnitude, with a mean Pearson’s correlation coefficient r of 0.85 (min of 0.54
and max of 0.97), a mean absolute percentage bias of 7.6% (min of -12.5% and max of 27.8%), and a mean
RMSD of 18.0 mm/mo (min of 9.4 mm/mo and max of 26.7 mm/mo), and no canal command had a bias
greater than ±30%. For ETLook, the mean Pearson’s correlation coefficient r was 0.93 (min of 0.68 and max of
1.00), the mean absolute percentage bias was 21.4% (min of -23% and max of 77%), and the mean RMSD
was 20.5 mm/mo (min of 7.3 mm/mo and max of 49.8 mm/mo). Whilst the temporal patterns were well
captured, the resulting magnitudes seemed to be mixed, with 14 canal commands having a percentage bias
larger than +30%. CMRSET ETa estimates were also assessed against two in situ Bowen ratio surface energy
balance ETa measurements, that were set up from July 2000 to March 2001 (9 months, during the growing
season) in two locations in the Punjab Province. In these locations, CMRSET was also implemented using
Landsat (30 m) bands and the same CMRSET model parameters to assess scale differences related to pixel
smearing and averaging in the coarser MODIS data (500 m). MODIS CMRSET showed reasonable agreement
both in magnitude and seasonality considering the scale differences. For both locations, the Pearson’s
correlation coefficient r was greater than 0.92, the percentage bias less than 20% and the RMSD less than 17
mm/mo. Results were markedly better for Landsat CMRSET estimates at both locations with bias in both
cases being lower than 5%, although with some seasonal compensation reducing bias and RMSD errors.
Considering the differences between CMRSET and the existing remotely sensed products, particular care is
required in the use of ETa for a quantitative water assessment or water balance analyses that uses absolute
values, such as the assessment of sustainable groundwater use. Ideally, the results of related applications
should be cross-checked for the presence of biases or inconsistencies in relation to the use of this or any
other remotely sensed ETa products. This report is companion to a report that assesses future scenarios

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | vii
impact on irrigated agriculture using the remotely sensed ETa products for scenario exploration in a way that
the differences do not exert an undue influence in the interpretation of results.
Monthly CMRSET ETa estimates were used to assess how ETa changed spatially and temporally in the lower
IBIS canal commands during the 2000‒2018 period. Temporal resolution was annual (April–March water
year) and seasonal – wet Kharif (April to September) and dry Rabi (October to March). The assessment
showed that ETa in most irrigated areas within the canal commands exceeded 600 mm/y, with some areas
that exceeded 1000 mm/y, particularly in rice canal commands in the Sindh Rice Wheat agro-climatic zone.
Lower mean annual and Kharif ETa occurred in the Sindh Cotton Wheat South agro-climatic zone and Sindh
Rice Wheat South agro-climatic zone canal commands, with around 400 mm to 500 mm (about 100 to 200
mm less) during Kharif. On the other hand, during Rabi, most irrigated areas exceeded 300 mm.
The pre-processing capabilities of Google Earth Engine and continuous update of its satellite imagery
catalogue, plus the straightforward implementation of CMRSET ETa, potentially on a continuous basis,
provides an opportunity for monitoring irrigation dynamics and the assessment of structural and policy
improvements in the IBIS.

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 1
1 Introduction
The Indus Basin Irrigation System (IBIS) provides food and economic security for about 207 million people in
Pakistan. The semi-arid climate and nearly-fully allocated surface and groundwater water resources
represent a continuous management challenge. The construction and improvement of hydraulic
infrastructure that regulates and conveys available water has played a pivotal role to increase the efficiency
of existing irrigation systems. Diagnosing the improvements in irrigation performance by implementing such
dedicated infrastructure has been hampered by lack of enough water supply and use data. Satellite Remote
Sensing (RS) may be the only means to provide robust and consistent estimates in large irrigated areas and
associated water use at a policy relevant scale (i.e. canal commands or <1 km spatial scale and monthly
temporal resolution). Multi-temporal RS data can potentially provide crop types, extent and associated
actual crop evapotranspiration (ETa), which can then be used to assess irrigation dynamics and performance
in large irrigation systems such as the IBIS, and to inform and improve water management (Ahmad et al.,
2009).
The purpose of this report is to present the results of the implementation and evaluation of a RS ETa
algorithm in the IBIS, the CMRSET (CSIRO MODIS ReScaled EvapoTranspiration, Guerschman et al., 2009)
algorithm, followed by a spatial and temporal analysis of the ETa dynamics in canal commands in the Sindh
Province. The implementation spans 2000 to 2018 at 10-day temporal and 500 m spatial resolution, making
it the first long-term time-series (>15 years) for the IBIS at spatial and temporal resolutions that are useful to
assess irrigation systems at the canal command scale. Previous estimates of ETa in the IBIS were performed
for a single year or a limited number of years, for a constrained spatial domain (e.g. Ahmad et al., 2008,
2009; Bastiaanssen et al., 2012).
CMRSET estimates were evaluated to assess their accuracy and therefore their usefulness for water balance
modelling and other applications. CMRSET was evaluated against (i) other previously implemented RS ETa
algorithms – SEBAL (Ahmad et al., 2008) and ETLook (Bastiaanssen et al., 2012; Cheema, 2012) - at the pixel
and canal command scale and (ii) in situ Bowen ratio ETa measurements at the pixel scale (Ahmad et al.,
2002).
This report is a companion to the report on assessments of climate change and dam sedimentation impacts
and urban water supply on irrigated agriculture in the Punjab and Sindh provinces (Ahmad et al., 2020a and
Ahmad et al. 2020b). The results described here are used as one of the inputs for the exploratory future
scenarios described in Ahmad et al. (2020a; 2020b).
Following this introduction, the remainder of this report is as follows: Section 2 presents the study region,
methods and materials follow in Section 3, Section 4 presents results followed by summary and conclusion in
Section 5.

2 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
2 Study region
The study region overlays the IBIS (~140,000 km2, ) within the Indus Basin (1,125,000 km2). The IBIS is the
largest contiguous irrigation system in the world (Condon et al., 2014). The region, located south of the
Himalayan mountains, is comprised of mostly flat topography with rich soils resulting from the erosion of the
Himalayan mountains and deposition in large alluvial valleys. The region’s climate is influenced by the South
Asian Monsoon, with most of the annual precipitation occurring from June to September. The monsoon is
the annual reversal of wind direction caused by excess heating over the South Asian land mass. It draws
moisture from the Arabian Sea and Bay of Bengal into South Asia, across Pakistan and into the southern
upper Indus Basin (where most precipitation occurs) during June to September (Charles et al., 2016).
Precipitation has a strong northeast-to-southwest gradient, with about 1800 mm per year precipitating in
the Himalayas to about 200 mm per year in the south of the IBIS (Hutchinson and Xu, 2013, Stewart et al.,
2018).
Figure 2-1 Geographic characteristics of the Indus Basin Irrigation System including main rivers, canal commands and
location of two Bowen ratio stations
The main rivers supplying the IBIS are the Chenab, Jhelum, Indus and Kabul, whereas the Sutlej and Ravi have
most of their flows diverted in India before entering Pakistan (Figure 2-1). Within Pakistan, approximately
75% (131 km3) of the mean annual flow in the Indus river (175 km3) is diverted to agriculture (mainly in the
IBIS) producing 90% of the food for Pakistan (Stewart et al., 2018).
There are 60 main canal commands areas (existing, plus those that are planned or under construction) along
the main rivers in the IBIS (numbered from north to south, Figure 2-2.). Irrigated agriculture accounts for

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 3
about 85 % of cereal production, all sugar production and nearly all cotton production (Archer et al., 2011).
Surface water is diverted from a river’s main course through an extensive network of barrages and canals to
canal commands so it can be used for irrigation for a wide range (50+) crops and horticulture through the
growing season. Horticulture and sugarcane are annual crops. The dominant Kharif (April to September)
crops are cotton, rice, maize and fodder. Wheat and fodder are the dominant crops during the Rabi (October
to March) season (Kirby et al., 2017, Ahmad et al., 2019).
Figure 2-2 Canal commands and the corresponding agro-climatic zones (ACZ) in the IBIS. Canal commands are labelled
from north to south.Source: Indus River System Authority (IRSA). A summary of canal commands and identification
number can be found in Appendix A. BRW=Balochistan Rice Wheat, KPKS=Khyber Pakhtunkhwa Sugarcane, KPMW=
Khyber Pakhtunkhwa Mixed Crops West, PCWE=Punjab Cotton Wheat East, PCWW=Punjab Cotton Wheat West,
PMW=Punjab Mixed Crops West, PRW=Punjab Rice West, PSW=Punjab Sugarcane West, SCWN=Sindh Cotton Wheat
North, SCWS= Sindh Cotton Wheat South, SRWN=Sindh Rice Wheat North, SRWN=Sindh Rice Wheat South
In this report, provinces and crop types define 12 different Agro-Climatic Zones (ACZ) for the canal
commands (Figure 2-2). Punjab and Sindh Provinces are the main irrigated areas, and cotton and/or rice
(Kharif) and wheat (Rabi) are the main crops. For example, PRW corresponds to Punjab Rice Wheat, whereas
PCWW corresponds to Punjab Cotton Wheat West, PMW corresponds to Punjab Mixed Wheat and PSW
corresponds to Punjab Sugarcane West. Accordingly, SCWN corresponds to Sindh Cotton Wheat North, and
so on. Other provinces included are Balochistan (B) and Khyber Pakhtunkhwa (KP), which follow the same
rationale as explained above. A summary of canal commands and their identification numbers (as in Figure
2-2) can be found in Appendix A .

4 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
3 Methods and materials
3.1 Long-term 10-day CMRSET ETa time-series estimates
The CMRSET (CSIRO MODIS ReScaled EvapoTranspiration, Guerschman et al., 2009) algorithm was used to
estimate ETa at a temporal resolution of 10 days and spatial resolution of 500 m for the 2000‒2018 period.
Monthly ETa is estimated by scaling Hargreaves potential evapotranspiration (ETp) via a remote sensing-
based crop factor (Kc), which is obtained from two indices – the Enhanced Vegetation Index (EVI, Huete et
al., 2002) and the Global Vegetation Moisture Index (GVMI, Ceccato et al., 2002). EVI and GVMI can be used
to discriminate open water when EVI is low and GVMI is high, and to detect vegetation water content when
EVI is high. The main advantages of CMRSET are that it uses a single set of parameters (i.e. does not need an
auxiliary land cover map) and it does not require manual calibration to detect wet or dry pixels, as in some
energy balance algorithms. Also, CMRSET can estimate ETa in areas with significant direct evaporation,
including lakes and floodplains. CMRSET has been evaluated in several studies, in different climates and for
different applications, including: water assessment modelling (Paca et al., 2019; Peña-Arancibia et al., 2016,
2019; van Dijk et al., 2011), ecosystem mapping (Barron et al., 2014; Peña-Arancibia et al., 2014) and
recharge studies (Crosbie et al., 2014; Silberstein et al., 2013). Three steps are required to obtain time-series
of 500 m, 10-day Kc, ETp and ETa, respectively:
1) 10-day EVI and GVMI composites at 500 m spatial resolution for the entire IBIS were obtained via
Google Earth Engine (Gorelick et al., 2017) from the daily Moderate Resolution Imaging
Spectroradiometer (MODIS) surface spectral reflectance product (MOD09GA, collection 6). The
average pixel value was selected within the 10-day composite, while minimising cloud cover and
nulls. The 10-day EVI and GVMI were extracted from February 2000 to December 2018. Any gaps
prevailing in the vegetation indices time-series were filled using Harmonic ANalysis of Time Series
(HANTS, Zhou et al., 2015).
2) Daily gridded surfaces (2.5 km resolution, Hutchinson and Xu, 2013) minimum and maximum
temperature were used to calculate ETp following Hargreaves (1974) for the 2000‒2013 period. To
update the daily ETp until 2018, daily Hargreaves ETp (250 km resolution) was computed via Google
Earth Engine from the Global Land Data Assimilation System (GLDAS) Version 2.1 (Rodell et al.,
2004). Further, both gridded datasets were accumulated to 10-day to match the temporal resolution
of the EVI and GVMI VIs for further upscaling. Finally, the GLDAS ETp were oversampled to 2.5 km
spatial resolution, and bias-corrected and evaluated for consistency using 10-day scaling factors
obtained from the 2000‒2013 period, in which both gridded ETp datasets overlapped.
3) 10-day ETa for 2000‒2018 was estimated with the following procedure: First, a residual Moisture
Index (RMI) was developed by combining EVI and GVMI. EVI and GVMI respond positively to
increases in near-infrared reflectance, therefore the two indices have a relatively high correlation
(Guerschman et al., 2009). If the correlation between the two indices is removed, the residuals could
be used as an indicator of vegetation moisture (Guerschman et al., 2009). The Residual Moisture
Index (RMI) was calculated for each pixel as the vertical distance between its corresponding GVMI
and a baseline as follows:
RMI=max(0, GVMI-(KRMI×EVI+CRMI), Eq. 1
where CRMI and KRMI are calibrated parameters which describe the exact position of the baseline, in
this way most of the correlation between the two indices was eliminated. The exact position of the
baseline was calculated using MODIS EVI and GVMI data for all the pixels in Australia for the months

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 5
of January, May and September 2001. These are considered representative for this study region,
given the range of EVI-GVMI combinations represented in Australia and similar climatic
characteristics corresponding to semi-arid areas. The values for KRMI and CRMI, after calibration, were
1.778 and -0.350, respectively.
4) Secondly, Kc was estimated as follows:
Kc=Kc_max×(1-exp(-a×EVIα-b×RMIβ), Eq. 2
where Kc_max is multiplied by a sigmoidal function of EVI and RMI. The model was calibrated for
MODIS data using eddy-covariance ETa from seven flux towers in Australia and validated with water
balance data for 227 unimpaired catchments across Australia (Guerschman et al., 2009). The
parameters used herein had the following values: Kc_max=1.00, a=14.42, α=2.701, b=2.086, β=0.953.
Finally, ETa was obtained by scaling ETp with the estimated Kc as:
ETa= Kc× ETp. Eq. 3
3.2 Comparison of CMRSET ETa against SEBAL and ETLook
CMRSET ETa was compared against two remote sensing ETa datasets previously implemented in the study
region (or part thereof): ETLook (Bastiaanssen et al., 2012; Cheema, 2012) and SEBAL (Ahmad et al., 2008).
ETLook uses soil moisture derived from passive microwave sensors as the driving force for calculation of a
surface energy balance that includes a soil moisture component. The advantage of this method is that is not
hampered by cloud cover and it allows estimation of ETa at finer than 10-day temporal resolution. ETLook uses
simple downscaling methods to oversample the 25 km spatial resolution passive microwave data from the
Advanced Microwave Scanning Radiometer (AMSR‐E) on the Aqua satellite to 1 km spatial resolution. ETLook
requires ancillary data (other than soil moisture) common to other radiation and energy balance methods:
spectral vegetation index, surface albedo, atmospheric optical depth, land use and land cover data, soil
physical properties, and meteorological data. Bastiaanssen et al. (2012) estimated 8-day ETa using the ETLook
method for the entire Indus Basin for the calendar year 2007.
SEBAL is a well-tested algorithm and has been evaluated in several countries against in situ data (see Ahmad
et al., 2009 and references therein). SEBAL resolves the radiation and energy balance, it requires manual
calibration to pick dry and wet pixels used to estimate the sensible heat flux of the energy balance. The dry
and wet pixels are manually selected, based on vegetation index, surface temperature, albedo and some
basic knowledge of the study area (Ahmad et al., 2009). Ahmad et al. (2008) implemented the monthly
SEBAL algorithm using MODIS (MOD021KM, collection 5) calibrated Radiances at 1 km spatial resolution for
the year 2004‒2005 (October to September) for the Punjab province, which corresponds to the northern
parts of the IBIS.
Both RS ETa datasets (ETLook and SEBAL) were oversampled to 500 m resolution for comparison with CMRSET.
The evaluation was performed for 56 (or 40 for SEBAL) out of the 60 canal command areas provided by the
Indus River System Authority (IRSA). Four canal commands were excluded because they were under
construction or planned for construction in the future or did not have any discernible irrigated areas during
the comparison time period. The presence/absence of irrigation was assessed by estimating mean 10-day
maximum and minimum EVI values aggregated at the canal command and further verified using Google
Earth.
Temporal and spatial comparisons between CMRSET and the two RS datasets were conducted:
• For the temporal comparison, canal command monthly aggregated time-series ETa values for the
corresponding years were compared visually and the following goodness-of-fit-statistics were computed:
Pearson’s correlation coefficient (r), percentage bias and root mean square difference (RMSD).

6 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
• For the spatial comparison, canal command pixels for all months for the corresponding years were
compared one-to-one visually and the following goodness-of-fit- statistics were computed: Pearson’s
correlation coefficient (r), percentage bias and root mean square difference (RMSD).
3.3 Comparison of CMRSET ETa against two locations with Bowen
ratio ETa estimates
Bowen ratio surface energy balance measurement systems to directly measure ETa were set up in a cotton-
wheat field (co-ordinates: 73°2´49.8´´E, 31°23´26.2´´N) at Faisalabad and a rice-wheat field (co-ordinates:
73°20´50.2´´E, 31°52´34.2´´N) at Pindi Bhattian in 2000 (Ahmad 2002, Ahmad et al., 2002). Bowen ETa
measurements were aggregated to monthly for comparison with CMRSET and were available from July 2000
to March 2001. The following goodness-of-fit-statistics were computed: Pearson’s correlation coefficient,
percentage bias and root mean square difference (RMSD).
3.4 ETa spatial and temporal dynamics for the IBIS and for canal
commands
The ETa monthly time-series were used to assess the similarities and differences between canal commands
and associated ACZs. The ETa dynamics assessment was conducted both spatially and temporally for the
lower IBIS canal commands, located in the Sindh Province (Balochistan canal commands were also included
for completeness, as the streamflow that enters the irrigation systems are shared for the 2000‒2018
period), including:
• Description of ETa spatial annual mean maps for a hydrological year (April to March for the 2000‒
2018 period), and seasonal mean maps for the wet (Kharif, April to September) and dry (Rabi,
October to March) seasons, highlighting differences between ACZs.
• Description of seasonal time-series for canal commands, aggregated by canal deliveries surface
water source.

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 7
4 Results and discussion
4.1 Long-term 10-day CMRSET ETa time-series estimates
CMRSET ETa was estimated at 10-day temporal resolution and 500 m spatial resolution for the 2000‒2018
period. The two ETp datasets used were: (i) the 10-day Hargreaves ETp obtained from daily gridded surfaces
(2.5 km resolution) minimum and maximum temperature for the 2000‒2013 period and, (ii) 10-day
Hargreaves ETp (250 km resolution) from the Global Land Data Assimilation System (GLDAS) Version 2.1. The
GLDAS ETp were oversampled to 2.5 km spatial resolution, and bias-corrected using: (i) 10-day scaling factors
obtained from the period 2000‒2013 and (ii) using cumulative distribution function (CDF) matching for each
grid cell (Reichle and Koster, 2004; Yin and Zhan, 2018) for the overlapping period, and evaluated for
consistency. Both resulting monthly CMRSET ETa time-series for the year 2013 aggregated to the canal
command scale were compared in the 56 canal commands (Figure 4-1).
Figure 4-1 Comparison of monthly CMRSET ETa time-series for the year 2013 aggregated to the canal command scale
for 56 canal commands. CMRSET using daily gridded 2.5 km ETp surfaces (blue lines) and CMRSET using rescaled GLDAS
ETp (red circles). Canal commands are reported from north to south across rows from left to right

8 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
It is readily apparent that both datasets are very similar both in terms of magnitude and temporal patterns.
The mean Pearson’s correlation coefficient r is 0.99, mean absolute percentage bias is 2.7%, and the RMSD is
4.4 mm/mo.
The pixel by pixel comparison for all months for 2013 were compared one-to-one in the 56 canal commands
with results showing no major differences in either dataset (Figure 4-2).
Figure 4-2 Density plot of the pixel by pixel comparison for 56 canal commands. Canal commands are reported from
north to south across rows from left to right. The colour ramp shows the data density, with blue through yellow to red
denoting lower to higher density. Note that the maximum and minimum values to determine the ramp are per panel
Figure 4-3 shows the mean hydrologic year (April to March) spatial distribution of ETa in the IBIS canal
commands. The mean ETa for the 2000‒2018 period is 485 mm/y, with a minimum ETa of 404 and maximum
of 530 mm/y, in 2002 and 2013 respectively. There is large spatial variability within canal commands due to
the presence of arid and intensively cropped areas, with mean minimum ETa of 35 mm/y and mean
maximum ETa of 1938 mm/y. ETa remains within 20% of the mean in each hydrologic year, controlled by
relatively stable areas of high ETa in canal commands.

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 9
Figure 4-3 Spatial ETa time-series for each hydrologic year (April to March) from 2000‒2001 to 2017‒2018

10 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
4.2 Comparison of CMRSET ETa against two RS products
4.2.1 Comparison of CMRSET ETa against ETLook
Figure 4-4 shows the comparison of monthly CMRSET and ETLook ETa time-series for the calendar year 2007
aggregated to the canal command scale for 56 canal commands. The results described herein are tabulated
in Appendix A.1. Both datasets agree reasonably well both in terms of magnitude and temporal patterns. The
mean Pearson’s correlation coefficient r is 0.93 (min of 0.68 and max of 1.00), the mean absolute percentage
bias is 21.4% (min of -23% and max of 77%), and the mean RMSD is 20.5 mm/mo (min of 7.3 mm/mo and
max of 49.8 mm/mo). Whilst the temporal patterns are well captured as shown in Figure 4-4, the resulting
magnitudes seem to be mixed, with CMRSET being higher in some areas and lower in some others. CMRSET
is higher than ETLook in the canal commands in KPKS and KPMW (Figure 2-2) during most months, with a
mean bias of 29.6%. CMRSET is somewhat higher or slightly lower with a bias generally lower than 25% in
the canal commands north-east of the Indus River, including canal commands in PSW, PRW and PCWE
(except for Sadiqia, Bahawal and Forwah). In canal commands in PCWW (except for Sidnhai, east of the Indus
River), CMRSET is higher, more so in the canal commands that have large non-vegetated areas (see EVI time-
series in Figure 4-4) with a mean bias of 26.9%. CMRSET is lower than ETLook in canal commands in BRW
(mean bias of -10,7%) and in canal commands in SRWN the mean bias is -11.5%. Results are mixed in canal
commands in SCWN, with CMRSET being in close agreement with ETLook in Khairpur West (bias of 10.7%) and
Rohri (bias of 10.5%), but higher for Khairpur East (bias of 39.4%) and Ghotki (bias of 27.3%). CMRSET is in
close agreement with ETLook in the canal commands in SRWN and SRWS, with most canals having a bias lower
than 20%, except for Kalri (SRWS, bias of 28.5%) and North W (bias of -23.8). CMRSET is higher in canal
commands in SCWS with a mean bias of 54%, note that Upper Nara has a bias of 77% and Nara of 32%.
Figure 4-5 shows density scatterplots of monthly pixel by pixel CMRSET and ETLook ETa for the calendar year
2007 for 56 canal commands. These plots can be used to interpret both the agreement between pixels of
CMRSET and the RS ETa products. A perfect agreement would show all pixels sitting on the 1:1 line. The
density colour map – from blue (less dense) to yellow (denser) ‒ shows the distribution of ETa within the
canal command. Overall, CMRSET and ETLook compare well, the mean Pearson’s correlation coefficient r is
0.74 (min of -0.05 and max of 0.98). For the canal commands in KPKS, the mean r is 0.55 (min of -0.05 and
max of 0.94) and the density plots show that CMRSET is generally higher than ETLook, with most pixels falling
to the right of the 1:1 line, except for Upper Swat and Kabul River. For the canal commands in KPMW, the
mean r is 0.78 (min of 0.74 and max of 0.83) and the density plots show that CMRSET is higher than ETLook,
with most pixels falling to the right of the 1:1 line. For the canal commands in PSW, the mean r is 0.82 (min
of 0.70 and max of 0.95) and the density plots show that CMRSET agrees well with ETLook, with most pixels
falling around the 1:1 line. For the canal commands in PRW, the mean r is 0.46 (min of 0.03 and max of 0.71)
and the density plots show that CMRSET agrees well with ETLook, with most pixels falling around the 1:1 line,
except for Raya and Marala Ravi, where CMRSET is higher than ETLook (note that the low r value in Marala
Ravi seems to be an statistical artefact due to the high concentration of pixels in one location). For the canal
commands in PCWW, the mean r is 0.82 (min of 0.70 and max of 0.89) and the density plots show that
CMRSET is higher than ETLook, with most pixels falling to the right of the 1:1 line, except for Panjad that is
more in agreement. For the canal commands in PCWE, the mean r is 0.73 (min of 0.53 and max of 0.86) and
the density plots show that CMRSET agrees well with ETLook, with most pixels falling around the 1:1 line,
except for Sadiqia and Bahawal, where CMRSET is higher than ETLook. For the canal commands in BRW and
SRW, the mean r is 0.92 (min of 0.89 and max of 0.98) and the density plots show that CMRSET agrees well
with ETLook, with most pixels falling around the 1:1 line, although CMRSET shows less variability than ETLook.
For the canal commands in SCWN, SCWS and SRWS, the mean r is 0.84 (min of 0.65 and max of 0.98) and the
density plots show that CMRSET agrees well with ETLook, with most pixels falling around the 1:1 line, except
for Khairpur E and Upper Nara (both SCWN canals), where CMRSET is higher than ETLook.

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 11
Figure 4-4 Comparison of monthly CMRSET (solid blue line) and ETLook (dashed blue line) ETa time-series for the calendar
year 2007 aggregated to the canal command scale for 56 canal commands. The Enhanced Vegetation Index (EVI)
aggregated time-series is shown for reference (green line). Canal commands are reported from north to south across
rows from left to right

12 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
Figure 4-5 Density scatterplots of monthly pixel by pixel CMRSET (X-axis) and ETLook (Y-axis) ETa for the calendar year
2007 and for 56 canal commands. Canal commands are reported from north to south across rows from left to right. The
colour ramp shows the data density, with blue through yellow to red denoting lower to higher density. Note that the
maximum and minimum values to determine the ramp are per panel
4.2.2 Comparison of CMRSET ETa against SEBAL
Figure 4-6 shows the comparison of monthly CMRSET and SEBAL ETa time-series for the year 2004‒2005
aggregated to the canal command scale for 40 canal commands. Both datasets agree reasonably well both in
terms of magnitude and temporal patterns. The mean Pearson’s correlation coefficient r is 0.85 (min of 0.54
and max of 0.97), the mean absolute percentage bias is 7.6% (min of -12.5% and max of 27.8%), and the
mean RMSD is 18.0 mm/mo (min of 9.4 mm/mo and max of 26.7 mm/mo). Both the temporal patterns and
magnitudes compare well as shown in Figure 4-6. CMRSET agrees well with SEBAL in most canal commands,
although June-July ETa peaks are lower in CMRSET than in SEBAL for Marala Ravi, Raya and Upper Chenab
(PRW), and Lower Dipalpur (PCWE). Figure 4-7 shows density scatterplots of monthly pixel by pixel CMRSET
and SEBAL ETa for the year 2004‒2005 aggregated to the canal command scale for 40 canal commands.
Overall, CMRSET and SEBAL compare well, the mean Pearson’s correlation coefficient r is 0.58 (min of -0.25
and max of 0.94). The density plots show that CMRSET agrees well with SEBAL, with most pixels falling
around the 1:1 line, except for Bahawal, Abbasia and Panjad where the density plots show that CMRSET is
higher than SEBAL, with most pixels falling to the right of the 1:1 line.

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 13
Figure 4-6 Comparison of
monthly CMRSET (solid blue
line) and SEBAL (dashed blue
line) ETa time-series for the
year 2004‒2005 (October to
September) aggregated to the
canal command scale for 40
canal commands. The
Enhanced Vegetation Index
(EVI) aggregated time-series is
show for reference (green
line). Canal commands are
reported from north to south
across rows from left to right

14 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
Figure 4-7 Density scatterplots of monthly pixel by pixel CMRSET (X-axis) and SEBAL (Y-axis) ETa for year 2004‒2005
(October to September) aggregated to the canal command scale for 40 canal commands. Canal commands are reported
from north to south across rows from left to right. The colour ramp shows the data density, with blue through yellow to
red denoting lower to higher density. Note that the maximum and minimum values to determine the ramp are per
panel
4.2.3 Differences between the ETa products
The comparison between CMRSET ETa against SEBAL and ETLook at the pixel and canal command scale
showed that although temporal patterns are largely similar, there are differences in magnitudes. This is
somewhat expected given the differences in the models, from a somewhat simple vegetation index
approach in CMRSET, increasing complexity in the land surface temperature (LST) approach in SEBAL, to the
combined passive microwave-LST approach in ETLook further constraining soil moisture availability. The time
series density scatterplots, particularly in the canal commands in the southern IBIS (see Figure 2-2 and
comparisons between CMRSET and ETLook in these canals), show that CMRSET tends to overestimate ETa in
dry pixels/areas when compared to ETLook. Conversely, this overestimation does not seem to occur when
CMRSET ETa is compared to SEBAL. Some irrigation landscapes that are highly spatially heterogenous at the
500 m pixel scale (for example narrow irrigated strips along the Upper Nara Canal) may present a challenge
for ETLook which relies on downscaled 25 km spatial resolution passive microwave data. This may be more
pronounced at the interface between irrigated and non-irrigated areas. On the other hand, some
overestimation in CMRSET may be related to CMRSET’s lack of mechanism to induce stomatal closure when
there is strong stress associated to temperature or water availability. Again, the agreement in the time series
(particularly in canal commands intensively irrigated) and density scatterplots suggest this is not expected to
be an issue in wetter and irrigated areas.

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 15
4.3 Comparison of CMRSET ETa against two locations with Bowen
ratio ETa estimates
Bowen ratio surface energy balance ETa measurements were set up from July 2000 to March 2001 in two
locations in the Punjab Province (Figure 4-8), one in a cotton field at Faisalabad and the other in a rice paddy
at Pindi Bhattian. Monthly CMRSET ETa for both MODIS (500 m spatial resolution) and Landsat (30 m spatial
resolution, USGS Landsat 5 Surface Reflectance Tier 1) were compared to monthly ETa Bowen ratio
measurements. Landsat data were used to asses scale differences related to pixel smearing and averaging in
the coarser MODIS data. MODIS CMRSET shows reasonable agreement both in magnitude and seasonality
considering the scale differences (Figure 4-8). In Faisalabad, the Pearson’s correlation coefficient r is 0.95,
the percentage bias is 16.3%, and the mean RMSD is 17.0 mm/mo. In Pindi Bhattian, the Pearson’s
correlation coefficient r is 0.92, the percentage bias is -3.9%, and the mean RMSD is 16.4mm/mo. As
expected, results were better for Landsat CMRSET estimates at both locations with bias in both cases being
lower than 5%.
Figure 4-8 Comparison between monthly CMRSET ETa for both MODIS (500 m spatial resolution green dashed line) and
Landsat (30 m spatial resolution, blue dashed line) and ETa from Bowen ratio measurements at two locations in Punjab:
(a) cotton filed at Faisalabad and (b) rice paddy at Pindi Bhattian
4.4 ETa spatial and temporal dynamics in the lower IBIS
The ETa monthly time-series were used to assess the spatial and temporal ETa dynamics for the Sindh canal
commands for the 2000‒2018 period. Figure 4-9 shows the spatial characteristics of mean annual, mean
Kharif (April to September) and mean Rabi (October to March) ETa, respectively. All maps are shown in a
similar linear scale colour map for comparative purposes. In terms of annual ETa means, large extents of
irrigated areas within canal commands exceed 600 mm/y, and there are areas that exceed 1000 mm/y,
particularly in canal commands in SRWN and SCWN. In part of these areas, ETa exceeds 600 mm during
Kharif. Lower mean annual and Kharif ETa occur in canal commands in SCWS and SRWS, where annual means
are generally around 900 mm/y and Kharif ETa is generally 100 to 200 mm lower (around 400 mm to 600
mm) than in canal commands in SRWN and SCWN. On the other hand, there are areas in canal commands in
both SRWN and SCWN with ETa values exceeding 400 mm during Rabi, whereas most irrigated areas in canal
commands in SCWS and SRWS are between 300 mm and 400 mm.
For further analysis, canal commands were grouped into larger canal command areas that are supplied from
barrages in Sindh (see Figure 4-9a) and also by ACZs. Figure 4-10 shows the mean seasonal averages for the
individual canal commands grouped by ACZs. The canal commands in BRW and SRWN (in blue, Figure 4-10)

16 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
have higher Kharif (492 mm/y) and Rabi (362 mm/y) ETa, respectively, than the SCWN and SRWS (in orange,
474 and 340 mm/y for Kharif and Rabi, respectively), and SCWN and SCWS (in green, 456 and 331 mm/y for
Kharif and Rabi, respectively). In BRW and SRWN (in blue), the Rice canal command (51 in Figure 4-9) had the
highest mean ETa in both seasons, with 572 mm and 495 mm, for Kharif and Rabi respectively. Conversely,
the Pat Feeder canal command (44 in Figure 4-9) had the lowest ETa in Kharif (407 mm) and Desert canal
command (45 in Figure 4-9) in Rabi (397 mm). In canal commands SCWN and SRWS (in orange), the Khairpur
West (50 in Figure 4-9) ) canal command had the highest mean ETa in both seasons, with 572 mm and 495
mm, for Kharif and Rabi respectively. In canal commands SCWN and SCWS (in green), the Nara canal
command (56 in Figure 4-9) had the highest mean ETa in both seasons, with 468 mm and 342 mm, for Kharif
and Rabi respectively.
Figure 4-9 Spatial ETa characteristics for the 2000‒2018 period in the lower IBIS including: (a) mean annual ETa, (b)
mean Kharif (April to September) ETa and (b) mean Rabi (May to October) ETa. Canal command boundaries are shown in
black. Note: the ID numbers in (a) depict a canal order ascending from north to south. A summary of canal commands
and identification number can be found in Appendix A

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 17
Figure 4-10 Mean seasonal (2000‒2018 period) ETa aggregated for canal commands in the lower IBIS: (a) Kharif (April
to October) and Rabi (November to March). Note: the ID numbers in (a) depict a canal order ascending from north to
south, canal commands are colour coded to denote canals sharing the same surface water supply source
The grouped canal commands per ACZs exhibit seasonal variability on a year-to-year basis (Figure 4-11).
Canal commands in SCWN and SCWS ACZs (green) had consistently less ETa (about 10% on average) that the
other two groups before the hydrologic year 2011‒2012, particularly during Kharif. The situation changed
after 2011‒2012, where Kharif values in canal commands in SCWN and SCWS have come closer (about 3%
less on average) to ETa in canal commands BRW and SRWN and SCWN and SRWS, potentially indicating
improvement in cropping conditions in canal commands in SCWN and SCWS. It is also noted that the input
ET0 data used for after 2013 came from a global model, rather than the locally interpolated ET0 data used
until 2013. Although great care has been taken in providing robust ET0 time-series after 2013 using a per-
pixel bias correction and ensuring consistency, it was not ascertained what the contributions of input ET0 or
VIs are in driving the changes after 2013.

18 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
Figure 4-11 Kharif and Rabi
ETa hydrologic year (April to March) time-series from 2000 to 2018, aggregated for canal
commands in the lower IBIS. Note: canal commands are grouped by colour to denote canals sharing the same surface
water supply source

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 19
5 Summary and conclusion
The CMRSET (CSIRO MODIS ReScaled EvapoTranspiration) ETa algorithm was implemented from 2000 to
2018 at 10-day temporal and 500 m spatial resolution in the Indus Basin Irrigation System (IBIS). Composites
of daily MODIS data were aggregated to 10-day means to estimate two vegetation indices, the Enhanced
Vegetation Index (EVI) and Global Vegetation Moisture Index (GVMI). Both indices are complementary: when
EVI is low and GVMI is high, CMRSET can detect and estimating open water ETa. This makes the algorithm
useful for estimating ETa under flood irrigation conditions. The indices were pre-processed and downloaded
from Google Earth Engine, which facilitates pre-processing by minimising the effects of cloud cover and nulls.
CMRSET estimates were evaluated against the remotely sensed algorithms SEBAL and ETLook at pixel and
canal command scales, and against in situ Bowen ratio ETa measurements at pixel scale. Generally, CMRSET
compared well against both datasets, both in terms of magnitude and temporal patterns. CMRSET agreed
better with SEBAL. For ETLook, whilst the temporal patterns were well captured, the resulting magnitudes
seem to be mixed, with 14 out of 56 canal commands having a positive percentage bias larger than 30%.
CMRSET ETa estimates were also assessed against two in situ Bowen ratio surface energy balance ETa
measurements. For both locations, CMRSET had good agreement for both temporal patterns and magnitude.
The similarity of spatial and temporal patterns of actual evapotranspiration between CMRSET and the
existing remotely sensed products gives confidence in its use for some applications. A companion report that
assesses future scenarios impact on irrigated agriculture uses the remotely sensed ETa products for scenario
exploration and shows that they can be used with confidence for assessing changes in the water balance.
However, the differences in actual values of the various methods suggests that care is required in the use of
ETa for quantitative water assessment or water balance analyses that uses absolute values, such as the
assessment of volumes of groundwater that may be extracted sustainably.
The ETa monthly time-series were used to assess spatial and temporal ETa dynamics for the Sindh Province
and for 16 canal commands therein for the 2000‒2018 period. The assessment was performed annually
(hydrologic year starting in April and finishing in March the next year) for the wet summer Kharif (April to
September) season and for the dry winter Rabi (October to March). The assessment showed that ETa in most
irrigated areas within canal commands exceeded 600 mm/y, with areas that exceed 1000 mm/y, particularly
in rice canal command areas. Lower mean annual and Kharif ETa occur in cotton and rice canal commands in
the south, with around 400 mm to 500 mm (about 100 to 200 mm less than in the canal commands more to
the north). On the other hand, during Rabi, ETa in most irrigated areas exceeded 300 mm.
These ETa time-series provide the first long-term (>15 years) consistent ETa time-series for the IBIS at spatial
and temporal resolutions that are useful to assess irrigation systems at the canal command scale. CMRSET
can be implemented straightforwardly using freely available data and processing through Google Earth
Engine. CMRSET requires only multi-temporal satellite optical data to estimate a crop factor and climate data
to estimate potential evapotranspiration and the product of these provides ETa, both continuously updated
and available through Google Earth Engine. This can enhance opportunities for monitoring irrigation
dynamics and the assessment of structural and policy improvements in the IBIS.

20 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
Appendix A Assessed canal commands in the IBIS
A.1 Comparison of CMRSET ETa against ETLook
Table A-1 presents the summary for results for the comparison between CMRSET and ETLook ETa for the
calendar year 2007 for each of the 56 canal commands assessed in this report.
Table A-1 Canal command characteristics, corresponding Agro-Climatic Zone (ACZ) and statistics of the CMRSET and
ETLook comparison. Statistics include the Pearson’s correlation coefficient (r), percentage bias, root mean square
differences (RMSD) and the pixel-by-pixel spatial r
ID Name ACZ Area
(km2)
ETLook 2007
(mm/y)
CMRSET ‘07
(mm/y)
r % Bias RMSD r spatial
1 Abazai KPKS 374 776 1116 0.99 30.4 31.7 0.50
2 Upper Swat KPKS 1474 881 1095 0.99 19.6 20.2 0.43
3 Warsak LB KPKS 59 823 947 0.97 13.0 14.7 0.82
4 Doada LB KPKS 156 821 1188 1.00 30.9 33.2 -0.05
5 L Swat KPKS 822 772 1151 0.99 33.0 34.9 0.27
6 Kabul River KPKS 581 896 1087 0.98 17.6 18.6 0.57
7 Pehur KPKS 229 856 1062 0.99 19.4 19.2 0.73
8 Warsak KPKS 549 574 710 0.96 19.1 13.8 0.94
9 Terbela KPKS 275 665 874 0.95 23.9 20.3 0.77
10 Shah Joe KPMW 63 568 885 0.94 35.7 30.2 0.76
11 Landi Dak KPMW 180 486 843 0.93 42.3 34.0 0.74
12 Central Bannu KPMW 535 425 774 0.91 45.0 32.6 0.83
14 U Jehlum PSW 2856 1119 1092 0.95 -2.5 15.8 0.70
15 Marala Ravi PRW 868 965 1105 0.97 12.7 15.4 0.03
16 L Jehlum PSW 7341 897 1008 0.99 11.0 11.2 0.72
17 Raya PRW 2190 917 1109 0.96 17.4 19.2 0.49
18 U Chenab PRW 4649 1013 1112 0.97 8.9 13.0 0.40
19 CR BC-KP KPMW 1607 453 1012 0.79 55.2 49.8 0.70
20 Thal PMW 12026 359 680 0.92 47.2 27.9 0.89
21 Jhang PRW 7432 985 1019 0.97 3.3 10.1 0.70
23 Gugera PRW 8721 1088 1100 0.98 1.0 11.5 0.71
24 Bari Doab PRW 3532 985 1080 0.97 8.8 12.0 0.68
25 CR BC-P PCWW 1178 293 764 0.78 61.7 41.1 0.82
26 U Dipalpur PRW 1634 1023 1130 0.98 9.5 12.7 0.85
27 U Rangpur PCWW 1071 715 1039 0.96 31.2 28.5 0.86
28 Haveli PSW 780 933 1054 0.98 11.4 11.7 0.91
29 Karanga PSW 176 748 906 0.98 17.5 14.5 0.95

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 21
ID Name ACZ Area
(km2)
ETLook 2007
(mm/y)
CMRSET ‘07
(mm/y)
r % Bias RMSD r spatial
30 L Dipalpur PCWE 2770 1105 1157 0.98 4.5 11.5 0.73
31 L Bari Doab PCWE 7889 1144 1125 0.99 -1.7 8.3 0.76
32 U Pakpattan PCWE 4397 1113 1160 0.99 4.0 7.5 0.53
33 Sidhnai PCWW 3464 1137 1181 0.99 3.8 7.3 0.73
34 Forwah PCWE 2004 881 1123 0.97 21.6 21.4 0.66
35 L Rangpur PCWW 509 859 1073 0.97 19.9 19.1 0.88
36 Muzafffargarh PCWW 4166 855 1078 0.94 20.7 21.2 0.81
37 L Mailsi Pakpattan PCWE 4662 1143 1163 0.97 1.8 12.1 0.73
38 Sadiqia PCWE 5068 543 938 0.96 42.1 34.3 0.81
39 Dera GK PCWW 5437 711 1042 0.86 31.7 32.0 0.89
40 Bahawal PCWE 4356 646 918 0.94 29.6 24.2 0.86
41 Abbasia PCWW 1416 521 780 0.95 33.2 22.6 0.85
42 Panjnad PCWW 6082 1039 1200 0.95 13.4 18.6 0.73
44 Pat Feeder BRW 3067 1026 882 0.89 -16.3 24.3 0.93
45 Desert BRW 1798 945 898 0.94 -5.2 14.9 0.93
46 Begari SRWN 4533 1057 891 0.94 -18.6 20.7 0.91
47 Ghotki SCWN 4818 639 879 0.90 27.3 23.3 0.81
48 North W SRWN 4028 938 758 0.94 -23.8 21.3 0.98
50 Khairpur W SCWN 1169 1060 1187 0.98 10.7 12.4 0.65
51 Rice SRWN 2244 1209 1085 0.92 -11.5 24.2 0.94
52 Khairpur E SCWN 1940 579 955 0.90 39.4 33.0 0.82
53 Upper Nara SCWS 1355 117 507 0.82 77.0 34.1 0.76
54 Dadu SRWN 2550 738 801 0.90 7.9 11.8 0.93
55 Rohri SCWN 11621 972 1086 0.90 10.5 13.3 0.82
56 Nara SCWS 9641 574 844 0.68 32.1 25.4 0.78
57 Fuleli SRWS 4227 797 719 0.83 -10.9 12.9 0.91
58 Lined Canal SRWS 2143 482 582 0.68 17.1 13.4 0.89
59 Kalri SRWS 2799 361 506 0.75 28.5 13.2 0.90
60 Pinyari SRWS 3904 584 558 0.79 -4.6 9.0 0.90

22 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
A.2 Comparison of CMRSET ETa against SEBAL
Summary of results for the comparison between CMRSET and SEBAL ETa for the year (October to September)
2004‒2005 for each of the 40 canal commands for which SEBAL was available (Table A-2).
Table A-2 Canal command characteristics, corresponding Agro-Climatic Zone (ACZ) and statistics of the CMRSET and
SEBAL comparison. Statistics include the Pearson’s correlation coefficient (r), percentage bias, root mean square
differences (RMSD) and the pixel-by-pixel spatial r.
Order Name ACZ Area (km2) SEBAL 0405
(mm/y)
CMRSET 0405
(mm/y)
r % Bias RMSD r spatial
1 Abazai KPKS 374 806 871 0.82 7.5 21.3 0.71
2 Upper Swat KPKS 1474 812 848 0.91 4.3 13.8 0.60
3 Warsak LB KPKS 59 793 767 0.88 -3.4 16.1 0.47
4 Doada LB KPKS 156 918 948 0.78 3.1 26.0 0.11
5 L Swat KPKS 822 885 928 0.80 4.7 23.2 0.36
6 Kabul River KPKS 581 909 901 0.88 -0.9 16.9 0.32
7 Pehur KPKS 229 847 823 0.94 -3.0 10.8 -0.25
8 Warsak KPKS 549 731 695 0.94 -5.2 9.4 0.90
9 Terbela KPKS 275 743 694 0.82 -7.0 16.7 0.53
10 Shah Joe KPMW 63 834 742 0.97 -12.5 14.6 0.29
11 Landi Dak KPMW 180 696 684 0.92 -1.8 11.9 0.79
12 Central Bannu KPMW 535 696 692 0.95 -0.5 9.8 0.94
14 U Jehlum PSW 2856 866 889 0.87 2.5 19.9 0.20
15 Marala Ravi PRW 868 924 864 0.90 -7.0 22.4 0.10
16 L Jehlum PSW 7341 802 837 0.89 4.1 14.9 0.60
17 Raya PRW 2190 948 853 0.88 -11.0 21.2 0.18
18 U Chenab PRW 4649 921 877 0.87 -4.9 20.5 0.17
19 CR BC-KP KPMW 1607 774 765 0.94 -1.2 10.6 0.79
20 Thal PMW 12026 420 576 0.85 27.1 15.6 0.87
21 Jhang PRW 7432 760 826 0.92 8.0 12.6 0.70
23 Gugera PRW 8721 831 873 0.91 4.8 13.9 0.38
24 Bari Doab PRW 3532 863 851 0.88 -1.4 18.1 0.49
25 CR BC-P PCWW 1178 566 585 0.78 3.2 12.8 0.71
26 U Dipalpur PRW 1634 911 897 0.84 -1.6 21.9 0.71
27 U Rangpur PCWW 1071 836 871 0.91 4.0 13.6 0.65
28 Haveli PSW 780 884 893 0.90 1.1 16.4 0.70
29 Karanga PSW 176 846 818 0.91 -3.5 16.7 0.90
30 L Dipalpur PCWE 2770 921 902 0.85 -2.1 21.4 0.70
31 L Bari Doab PCWE 7889 844 905 0.86 6.8 17.6 0.64
32 U Pakpattan PCWE 4397 861 908 0.75 5.2 23.5 0.42
33 Sidhnai PCWW 3464 875 979 0.86 10.6 18.6 0.54

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 23
Order Name ACZ Area (km2) SEBAL 0405
(mm/y)
CMRSET 0405
(mm/y)
r % Bias RMSD r spatial
34 Forwah PCWE 2004 802 845 0.72 5.1 22.7 0.33
35 L Rangpur PCWW 509 907 927 0.92 2.3 14.7 0.73
36 Muzafffargarh PCWW 4166 777 886 0.86 12.3 17.4 0.84
37 L Mailsi Pakpattan PCWE 4662 798 902 0.66 11.5 26.7 0.70
38 Sadiqia PCWE 5068 508 704 0.63 27.8 23.8 0.78
39 Dera GK PCWW 5437 754 812 0.72 7.2 21.3 0.85
40 Bahawal PCWE 4356 541 718 0.54 24.7 25.3 0.87
41 Abbasia PCWW 1416 454 621 0.76 26.8 18.6 0.94
42 Panjnad PCWW 6082 749 942 0.81 20.5 25.3 0.82

24 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
References
Ahmad MUD, Bastiaanssen WGM, Feddes RA. 2002. Sustainable use of groundwater for irrigation: A numerical analysis
of the subsoil water fluxes. Irrigation and Drainage 51: 227-41
Ahmad MD, Gamage N, Vithanage J, Masih I, Muthuwatte L, Gunasinghe S, Hussain A. 2008. Assessing the impact of
physical and management interventions through RS/GIS tools on irrigation system performance in Punjab
Pakistan (Final Report 2008) Colombo, Sri Lanka: International Water Management Institute.
Ahmad MD, Stewart JP, Peña-Arancibia JL, Kirby JM. 2020a. Sindh water outlook: Impacts of climate change, dam
sedimentation and urban water supply on irrigated agriculture. Technical report. Sustainable Development
Investment Portfolio project. CSIRO, Australia.
Ahmad MD, Stewart JP, Peña-Arancibia JL, Kirby JM. 2020b. Sindh Water Outlook: Impacts of Climate Change, Dam
Sedimentation and Urban Water Supply on Irrigated Agriculture. Sustainable Development Investment Portfolio
(SDIP) project. CSIRO, Australia.
Ahmad MD, Turral H, Nazeer A. 2009. Diagnosing irrigation performance and water productivity through satellite
remote sensing and secondary data in a large irrigation system of Pakistan. Agricultural Water Management 96:
551-64
Archer DR Forsythe N, Fowler HJ, Shah SM. 2010. Sustainability of water resources management in the Indus Basin
under changing climatic and socio-economic conditions. Hydrology and Earth System Sciences, 14, 1669-1680
Barron OV, Emelyanova I, Van Niel TG, Pollock D, Hodgson G. 2014. Mapping groundwater-dependent ecosystems using
remote sensing measures of vegetation and moisture dynamics. Hydrological Processes 28: 372-85
Bastiaanssen WGM, Cheema MJM, Immerzeel WW, Miltenburg IJ, Pelgrum H. 2012. Surface energy balance and actual
evapotranspiration of the transboundary Indus Basin estimated from satellite measurements and the ETLook
model. Water Resources Research 48
Ceccato P, Flasse S, Gregoire JM. 2002. Designing a spectral index to estimate vegetation water content from remote
sensing data-Part 2. Validation and applications. Remote Sensing of Environment 82: 198-207
Charles SP. 2016. Hydroclimate of the Indus ‐ synthesis of the literature relevant to Indus Basin hydroclimate processes,
trends, seasonal forecasting and climate change. CSIRO Sustainable Development Investment Portfolio project.
CSIRO Land and Water, Australia. 48pp
Cheema MJM. 2012. Understanding water resources conditions in data scarce river basins using intelligent pixel
information: Transboundary Indus Basin (Doctoral Thesis). Technische Universiteit, Delft.
Condon M, Kriens D, Lohani A, Sattar E. 2014. Challenge and response in the Indus Basin. Water Policy 16: 58-86
FAO and IHE Delft. 2019. WaPOR quality assessment. Technical report on the data quality of the WaPOR FAO database
version 1.0. Rome. 134 pp. http://www.fao.org/3/ca4895en/ca4895en.pdf
Gorelick N et al. 2017. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of
Environment, 202: 18-27. DOI:10.1016/j.rse.2017.06.031
Guerschman JP, Van Dijk A, Mattersdorf G, Beringer J, Hutley LB, et al. 2009. Scaling of potential evapotranspiration
with MODIS data reproduces flux observations and catchment water balance observations across Australia.
Journal of Hydrology 369: 107-19
Hargreaves GH. 1974. Estimation of Potential and Crop Evapotranspiration. Transactions of the ASAE 17: 701-4
Huete A, Didan K, Miura T, Rodriguez EP, Gao X, Ferreira LG. 2002. Overview of the radiometric and biophysical
performance of the MODIS vegetation indices. Remote Sensing of Environment 83: 195-213
Hutchinson MF, Xu T. 2013. ANUSPLIN Version 4.4 User Guide. Fenner School of Environment and Society, Australian
National University, Canberra. http://fennerschool.anu.edu.au/files/anusplin44.pdf

Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin | 25
Kirby M, Ahmad MUD, Mainuddin M, Khaliq T, Cheema MJM. 2017. Agricultural production, water use and food
availability in Pakistan: Historical trends, and projections to 2050. Agricultural Water Management 179: 34-46
Paca VHD, Espinoza-Davalos GE, Hessels TM, Moreira DM, Comair GF, Bastiaanssen WGM. 2019. The spatial variability
of actual evapotranspiration across the Amazon River Basin based on remote sensing products validated with
flux towers. Ecological Processes 8
Peña-Arancibia JL, Mainuddin M, Kirby JM, Chiew FHS, McVicar TR, Vaze J. 2016. Assessing irrigated agriculture's surface
water and groundwater consumption by combining satellite remote sensing and hydrologic modelling. Science
of the Total Environment 542: 372-82
Peña-Arancibia JL, Mainuddin M, Ahmad M, Hodgson G, Khandakar F, Ticehurst C, Maniruzzaman M, Golam M, Kirby
JM. 2019. Groundwater use and rapid irrigation expansion in a changing climate: hydrological drivers in one of
the world's food bowls. Journal of Hydrology. https://doi.org/10.1016/j.jhydrol.2019.124300.
Peña-Arancibia JL, McVicar TR, Paydar Z, Li L, Guerschman JP, et al. 2014. Dynamic identification of summer cropping
irrigated areas in a large basin experiencing extreme climatic variability. Remote Sensing of Environment 154:
139-52
Reichle RH, Koster RD. 2004. Bias reduction in short records of satellite soil moisture. Geophysical Research Letters 31
http://dx.doi.org/Artn L1950110.1029/2004gl020938.
Rodell M, Houser PR, Jambor U, Gottschalck J, Mitchell K, et al. 2004. The Global Land Data Assimilation System. Bulletin
of the American Meteorological Society 85. https://journals.ametsoc.org/doi/10.1175/BAMS-85-3-381
Silberstein RP, Dawes WR, Bastow TP, Byrne J, Smart NF. 2013. Evaluation of changes in post-fire recharge under native
woodland using hydrological measurements, modelling and remote sensing. Journal of Hydrology 489: 1-15
Stewart JP, Podger GM, Ahmad MD, Shah MA, Bodla H, Khero Z, Rana MKI. 2018. Indus River System Model (IRSM) – a
planning tool to explore water management options in Pakistan: model conceptualisation, configuration and
calibration. August 2018. Technical report. South Asia Sustainable Development Investment Portfolio (SDIP)
Project. CSIRO, Australia. 155p
Van Dijk AIJM, Bacon D, Barratt D, Crosbie R, Daamen C, et al. 2011. Design and development of the Australian Water
Resources Assessment system. Presented at Proceedings, Water Information Research and Development Alliance
Science Symposium, August 2011. Melbourne. Available from http://www.csiro.au/WIRADA-Science-
Sympsosium-Proceedings
Yin JF, Zhan XW. 2018. Impact of Bias-Correction Methods on Effectiveness of Assimilating SMAP Soil Moisture Data into
NCEP Global Forecast System Using the Ensemble Kalman Filter. Ieee Geoscience and Remote Sensing Letters 15:
659-63
Zhou J, Jia L, Menenti M. 2015. Reconstruction of global MODIS NDVI time series: Performance of Harmonic ANalysis of
Time Series (HANTS). Remote Sensing of Environment 163: 217-28

26 | Remotely sensed time-series (2000‒2018) estimation of evapotranspiration in the Indus Basin
As Australia’s national science agency and
innovation catalyst, CSIRO is solving the
greatest challenges through innovative science
and technology.
CSIRO. UNLOCKING A BETTER FUTURE FOR EVERYONE
FOR FURTHER INFORMATION
CSIRO Land and Water
Dr Jorge Luis Pena Arancibia
+61 2 6246 5711
jorge.penaarancibia@csiro.au
www.csiro.au/
Dr Mobin-ud-Din Ahmad
Project Leader – Indus SDIP Pakistan
+61 2 6246 5936
mobin.ahmad@csiro.au
research.csiro.au/sdip/projects/indus/
1300 363 400
+61 3 9545 2176
CSIROEnquiries@csiro.au
www.csiro.au