Mapping Agriculture Areas Using Synthetic Aperture Radar in the Context of Flood-Related Disaster Events

Abstract:

Flooding in key agricultural areas can lead to widespread damage to crops and can have significant impacts on livelihoods and food availability in the affected areas. Access to actionable near real-time information on surface water extent, flood water depth, and impacted agricultural lands is often the main limitation of disaster management systems. This is especially true in many developing countries where in-situ observations for monitoring the impacts of rainfall and inundation on agriculture are often lacking.

Remote sensing is a promising mechanism to document agricultural activity on a regional scale and assess the impact of flooding events on crop health and crop productivity. However, cloud cover associated with many of these events limits monitoring of vegetation with optical systems. Data from Synthetic Aperture Radar (SAR) sensors such as Sentinel-1 or the upcoming NASA-ISRO SAR (NISAR) can supplement optical analyses by providing weather- and illumination-independent observations that are sensitive to vegetation structure and have proven capability in the mapping of surface water extent.

This paper introduces SAR-based algorithms and value-added data products that enable the monitoring of agricultural areas in the context of major flooding events. The work presented here is part of the HydroSAR project, which is building a cloud-based SAR data analysis service for the mapping of hydrological disasters and their impact on population and agriculture.

We will present the algorithms behind a SAR-based information product that is delineating regions of agricultural activity at the 1ha scale. It is derived from a statistical analysis of SAR time series data based on the coefficient of variation (CoV) metric. Changes in radar cross section throughout the crop cycle result in high CoV values, such that crop lands can be extracted from CoV via a thresholding algorithm. We will present first performance assessments of this algorithm for areas in the U.S. and in the Hindu-Kush Himalaya region.

We will also show how we intersect agriculture extent information with flood extent data to determine the extent and duration of flood impacts on crop lands. We will show examples of crop inundation maps for recent flooding events along the Nebraska River and for selected areas in eastern India and Bangladesh.