Let’s scale up remote sensing technology to better evaluate projects

Let’s scale up remote sensing technology to better evaluate projects

Development professionals are using innovative practices in the area of agriculture.

By Garrett Kilroy, Andrew Brubaker, Maya Vijayaraghavan

Remote sensing technology can be a cost-effective tool to help international development practitioners monitor project outcomes

Technology can be very handy to monitor progress over time and assess results of projects, especially where resources are limited and investments should be prioritized. Remote sensing technologies offer exciting opportunities for more effective monitoring.

ADB’s Independent Evaluation Department has been piloting remote sensing in recent evaluations, with impressive results.

Remote sensing is the process of detecting the physical characteristics of the terrain using sensors or cameras in satellites, aircraft, and even with sonar systems on ships. Large images help us to see and understand much more of what is happening on the ground. If we use historical remotely sensed images, we can examine changes over time such as for land use or deforestation.

We began by using remotely sensed land cover data to help identify comparison towns for an impact evaluation of a water supply project in Nepal, which aimed to improve household welfare in small towns through better water service delivery. More recently, we also deployed this technology for a sector-wide evaluation of ADB’s support for agriculture, natural resources and rural development by measuring outcomes in two case studies on natural resources and irrigation.

The results from both pilots offer strong evidence to justify wider uptake and application of remote sensing for outcome monitoring and evaluation.

  The opportunities and challenges of using remote sensing technology for project evaluation

Projects in sectors like agriculture and natural resources can be particularly difficult to monitor and evaluate, as sites are typically distributed widely across landscapes and often in remote locations. This poses challenges to evaluators, both in terms of logistics and time to ensure that sufficient sites are visited to allow for a representative assessment of results.

However, projects from these sectors are often designed with spatially related outcome targets—such as an increase in the area of restored wetland, increased forest cover, or greater wheat tonnage per hectare of farmed area—which can be readily measured over time using remote sensing technologies.

Remotely sensed land cover data captured by sensors in satellites like NASA’s Landsat are ideal to track the achievement of such outcome targets. We piloted this approach in a wetland restoration project in the Peoples’ Republic of China and a water resources project in Bangladesh.

The majestic wetlands of the Sanjiang Plain, in the northwest of the Peoples’ Republic of China, is one of the richest areas of globally significant biodiversity, particularly for water birds. However, agriculture has created a major pressure on these wetlands.

ADB, with cofinancing from the Global Environment Facility, implemented an integrated conservation and development model with restoration activities, reforestation, and provision of alternative livelihoods for affected peoples. To measure the impact of wetland restoration activities completed between 2006 and 2008, we decided to use freely available earth observation Landsat data to assess changes in land cover in the Qixinghe Nature Reserve, one of six protected areas covered by the project.

We used remotely-sensed data over four time slices in 2000, 2005, 2011, and 2017 to track land cover change before and after the project. The remotely sensed data provided two findings that would have been difficult to get without such technology.

On the positive side, the data showed that since the project began in 2005, the increase in agriculture use within the protected area has halted. Today, there is no trace of agriculture activities in the core zone of the wetland. Unfortunately, ongoing agriculture activities can still be seen in the buffer and experimental zones, which requires continued attention to restore the integrity of the wetland.

The second case study was in a 300-hectare irrigation command area in Bangladesh, a subproject that had benefited from two sequential ADB investments over 15 years. These had combined irrigation infrastructure improvements with the introduction of high-yield varieties of rice and were cofinanced by the International Fund for Agricultural Development.

Using the normalized difference vegetation index (NDVI, a measure of crop “greenness”) from remotely sensed data as a surrogate for crop yield, we tracked the index levels pre- and post-project just before the harvest time of the rice crops. The remotely sensed data used to track increases in crop yields from 2000 to 2017 validated the yields reported by the farmers, and confirmed that the relative increases in rice yield was of a similar order of magnitude to officially reported values.

Again, such independent confirmation would not be readily available through other means.

  Remote sensing has huge potential in the area of development project evaluation

Looking to the future and other sectors, there are growing opportunities to apply remote sensing technologies, like in transport where nighttime luminosity data have proven to be a good indicator of economic activity. For example, a recent USAID geospatial impact evaluation analyzed the timing and location of road improvements and remotely sensed nighttime lights outcome data to determine a program’s impact on economic development.

There are some considerations to be borne in mind when using remote sensing for monitoring project outcomes.

Cloud-free data may not always be available for the time series of interest and, while remotely sensed data is readily accessible and often free of charge, it does require expertise to process. Collecting ground truth data to verify the remotely sensed data will also bear a cost.

Nonetheless, remote sensing can be a cost-effective tool to help practitioners monitor outcomes for relatively straightforward indicators like land cover change and the normalized difference vegetation index, adding value to traditional monitoring approaches, helping establish baselines and better track the achievement of spatially related outcomes.