Evaluation of a new 18-year MODIS-derived surface water fraction dataset for constructing Mediterranean wetland open surface water dynamics

Abstract

Wetlands are among the most biodiverse ecosystems in the world, due largely to their dynamic hydrology. Frequent observations by satellite sensors such as the Moderate Resolution Imaging Spectrometer (MODIS) allow for monitoring the seasonal, inter-annual and long-term dynamics of surface water extent. However, existing MODIS-based studies have only demonstrated this for large water bodies (>100 km2) despite the ecological importance of smaller-sized wetland systems. In this paper, we assessed if the temporal dynamics of surface water extent can be effectively captured with MODIS for a wide range of water body sizes, specifically 0.01 km2 and larger. We constructed the temporal dynamics of surface water extent for 340 individual water bodies in the Mediterranean region between 2000 and 2017, using a previously developed 8-day 500 m MODIS surface water fraction (SWF) dataset. We then evaluated how MODIS SWF time series corresponded to water extent derived from a Landsat-based dataset and to satellite-altimetry derived water level data. Results showed that although correlations between MODIS- and Landsat-derived SWF were poor for relative static water bodies (r = 0.17), they increased for more dynamic water bodies (r = 0.81). For dynamic water bodies>100 km2, time series of water extent derived from MODIS SWF showed good correlation with both Landsat (r ≥ 0.76) and water level data (ρ ≥ 0.63). Our MODIS SWF dataset can also effectively monitor the variability of very small water bodies (<1 km2) when comparing with Landsat data as long as the temporal variability in their surface water area was high. We conclude that MODIS SWF is a useful product to help understand hydrological dynamics for both small and larger-sized water bodies, and to monitor their seasonal, intermittent, inter-annual and long-term changes.