High spatiotemporal resolution remotely sensed timeseries actual evapotranspiration estimates for irrigation management in salinity-affected areas of the southern Indus basin

Abstract

The Indus Basin Irrigation System (IBIS) in Pakistan is the largest contiguous irrigated system (~160,000 km 2 ) in the world. IBIS's crops sustain Pakistan's ~230 million people and provide a livelihood for ~90,000 farmers. Like many other irrigation systems around the world, IBIS is facing critical water management challenges, inter alia: (i) over-utilised surface and groundwater resources, (ii) areas with unsustainable groundwater use, (iii) areas with salinity and waterlogging reducing productivity, (iv) sediment loads reducing the capacity of reservoirs used to regulate flows and (v) climate change modifying rainfall patterns, river flow regime and evaporative demand. The system relies on two large reservoirs (16.5 km 3 combined storage capacity), 16 barrages and a vast interconnected network of channels to equitably provide water to 45 canal commands on a 10-day period. Timely and accurate estimates of water use are essential for adequate management of irrigation systems. In large and complex systems like the IBIS, distributed surface water canal or groundwater pumping volume data are often not available for water balance assessments and to monitor irrigation management practices. Remote sensing imagery and geostatistical techniques that increase data availability provide an unparalleled opportunity to estimate actual evapotranspiration (ETa) -a proxy for water use -from large areas at the required spatial scales and temporal frequencies to assess irrigation patterns beyond the canal command scale. This paper describes the development of the first 30 m resolution and 10day gap-free remote sensing-based ETa timeseries from 2010-20 for salinity-affected canal commands in the IBIS facing different irrigation management challenges: (i) Sidhnai and Lower Mailsi and Lower Pakpattan canal commands with deep and rapidly declining groundwater tables in the south of Punjab province (roughly (from north to south) in the middle section of the IBIS), and (ii) the Pinyari canal command with shallow groundwater table and waterlogged areas, located in the south of Sindh province at the IBIS's tailend. The ETa estimates were computed by scaling potential Priestley-Taylor evapotranspiration (E0) using the vegetation index (VI)-derived scaling coefficient obtained from the CMRSET (CSIRO MODIS ReScaled EvapoTranspiration) Landsat V2.2 model at the Landsat native spatial resolution of 30 m. The 10-day frequency was achieved by taking advantage of the daily frequency VIs from the daily Moderate Resolution Imaging Spectroradiometer (MODIS), which were subsequently blended with Landsat VIs using a method capable of detecting gradual (i.e., phenological) or abrupt (e.g., floods) land cover changes. The resulting CMRSET ETa compared reasonably well against monthly estimates from previously published ETa products in selected canal commands, SEBAL and ETLook, with monthly correlations >0.7 and percentage biases <20%. The ETa timeseries aggregated at the canal command scale, alongside observations of canal command withdrawals, showcase...