Sentinel-1 application to rainfall response and geomorphic mapping

Rainfall is one of the primary geomorphic drivers on Earth’s surface. How a surface responds to rainfall directly impacts erosional, geomorphic, and natural hazard processes. In the absence of vegetation, whether a land surface retains rainfall as soil moisture or whether rainfall is quickly infiltrated or run off is largely a function of geomorphologic and geologic conditions. In this study, we combine a time series of synthetic aperture radar (SAR) backscatter with daily precipitation to analyze the response of arid and semi-arid land surfaces to rainfall from the event to seasonal scale. The study focuses on northwestern Argentina, where we have extensive field knowledge of local geomorphic features, and is implemented using the cloud computing capacities of Google Earth Engine (GEE).

Th Sentinel-1 satellites provide high spatial (10 m) and temporal resolution images of Earth’s surface, irrespective of cloud cover. We created a 3 year time series from 2018 through 2020 of Sentinel-1 sigma-naught (σ₀) backscatter from Ground Range Detected (GRD) products available on GEE. Combining the ascending and descending orbits of the Sentinel-1A and -1B satellites into a single time series provides 3 to 6 day temporal resolution in our area of interest. The Global Precipitation Measurement Mission (GPM) was aggregated to daily and monthly precipitation measurements to identify single rainfall events and the seasonal rainfall signal.

The response and recovery of SAR backscatter to individual rainfall events across different land surfaces was calculated over 4 to 6 week periods centered on and following a specific rainfall date, respectively. The temporal trend of the backscatter data in these time windows is calculated for every pixel in the backscatter stack to create a map how the surface responds to a large rainfall event. The location of standing water, increased soil moisture, and high infiltration surfaces are detectable in the response maps. The recovery maps provide a proxy for the rate of drying following the rainfall event.

In the monsoon-dominated region of northwestern Argentina, both precipitation and SAR backscatter show a clear, periodic seasonal signal over our three-year time series. By aggregating all data to monthly resolution, we can calculate pixel-wise linear regressions and correlation coefficients between precipitation and SAR backscatter. Regressions and correlation analysis are done at the resolution of the Sentinel-1 data and are used to identify whether a surface retains soil moisture, has high infiltration, or experiences seasonal standing water or snow cover. Areas dominated by highly weathered granites and sandstones that can retain soil moisture, for example, have strong positive correlation between rainfall and backscatter due to the increased dielectric constant of wet sediment. In contrast, gravel terraces where rainfall can easily infiltrate the surface show little correlation between backscatter and precipitation. The result is a high resolution map characterizing the propensity for soil moisture retention, high infiltration, and standing water and snow cover. Future work will focus on using these relationships to classify geomorphic surfaces across the arid and semi-arid central Andes.