Can remote sensing identify successful agricultural water management interventions in smallholder farms?

Food security in sub-Saharan African (SSA) countries hinges on smallholder farms, most of which are currently characterized by low productivity. Increasing the productivity of smallholder systems in SSA is therefore critical to meet the region's rapidly growing food demand. Africa's yield gap – the difference between an agricultural system's potential yield and actual farm yield – is much greater than in other continents.

Inadequate water availability remains one of the key limiting factors in increasing the productivity of SSA's smallholder systems. Without implementing accelerated water management interventions targeting smallholder farmers at scale, SSA will struggle to meet rising food demand and will remain food insecure.

Efficient water management for smallholders

Between 2013 and 2018, IWMI and partners piloted wetting front detectors (WFDs) in Ethiopia's Koga irrigation scheme. Simultaneously, ICRISAT and partners piloted WFDs in Silalatshani and Mkoba, two smallholder irrigation schemes in Zimbabwe.

Developed by the Virtual Irrigation Academy, the WFD is a simple soil moisture tool that informs farmers when to irrigate and when to stop applying water to their fields. The results from the IWMI and ICRISAT irrigation schemes, as well as other pilots in different countries, have shown potential to reduce overall water use in the field while significantly improving crop yield. Efforts are now underway to scale up implementation to other smallholder irrigation schemes.

The IWMI and ICRISAT field pilots in Koga, Silalatshani and Mkoba mapped out the exact locations of farms in which WFD interventions were successful. This has provided a timely opportunity to test the efficacy of freely-available remote sensing datasets in these irrigation schemes to identify the locations of successful on-farm interventions

Satellites can show the way

Freely-available satellite data, covering a wide geographical area with repeated measurements over time, offers an alternative way to monitor and assess intervention impacts in fragmented smallholder landscapes. Although it is has become common to use satellite data to assess the impacts of land restoration measures, its application to determine how on-farm water management practices impact agricultural productivity improvements is still in its infancy.

A wide range of freely-available satellite data sources exists, capturing the Earth's surface in various levels of detail. To assess the impact of water management interventions in SSA, a recently-completed IWMI study, supported by WLE, mainly used Landsat 8 satellite data. This data provided the best available resolution for the period during which on-farm water management interventions took place in the Koga, Silalatshani and Mkoba irrigation schemes.

Implementing a Before and After Control Impact design

The study adopted a Before and After Control Impact (BACI) design. Already popular in environmental impact assessments, BACI designs are now increasingly used for land restoration impact assessments using remote sensing data. A BACI design compares a biophysical variable in two sites: the "control", where a business-as-usual scenario prevails, and the "impact", where analyses are conducted before and after the intervention.

The two variables of interest used in the study's BACI design were crop water use and the enhanced vegetation index (EVI), a proxy for yield derived directly from Landsat 8 data. Crop water use was estimated using an energy balance model drawing from a multitude of satellite-based inputs, including Landsat 8. Satellite data allowed the on-farm interventions to be assessed at a resolution of 30 square metres.

Comparing before and after interventions

The field data collected by IWMI in Koga, and by ICRISAT in Silalatshani and Mkoba, were used to determine increases in water and land productivity in both the controls and the pilot locations where WFDs were used for irrigation scheduling. To carry out the BACI design, the remote sensing indicators (crop water use and EVI) were extracted for periods before and during or after the intervention for the farms designated as successful.

The study, based on this BACI design, indicated that crop water use improved in Koga and Silalatshani, identifying 50% and 55% of farms with intervention impacts, respectively. EVI performed better in Mkoba, where the study identified 62% of intervention farms with impacts. Of all the pilot farms with successful WFD implementation, 80% of the locations were captured by either crop water use or EVI, although fewer than 20% of impact sites had overlapping crop water use and EVI indicators. However, it is quite likely that some farms experienced a significant reduction in crop water use but a marginal increase in yield, or vice versa. Because the Landsat 8 data used in this study is of moderate resolution, the remote sensing indicators were not sensitive enough to capture marginal increases observed in farms.

Limitations exist but opportunities galore

The results highlight that, with moderate resolution satellite data like Landsat, the identification of intervention impact locations at farm scale is possible, albeit with some limitations. One of the major constraints of the approach is its inability to resolve marginal improvements in water use or yield gain at the farm level. In addition, pixels from Landsat representing farms often encompass adjacent land parcels due to the small and irregular sizes of farms in SSA.

When analyzing past on-farm water management interventions implemented at scale, some of the limitations associated with moderate resolution remote sensing data can be resolved by analyzing impacts at irrigation block or water user association level instead of farm level. However, for interventions after 2018, impact assessments can be successfully carried out at the farm level in smallholder irrigation schemes thanks to the recent availability of high-resolution datasets like Sentinel 2 and Planet-NICFI.

The report shows that the use of remote sensing indicators for impact assessment in smallholder systems can help decision makers identify where on-farm water management interventions are successful and where they do not produce the desired impacts. By enabling decision makers to select interventions that successfully increase the productivity of SSA's smallholder farms, satellite-based inputs can ultimately help close the region's yield gap, strengthening food security for millions who live there.