A new dense 18-year time series of surface water fraction estimates from MODIS for the Mediterranean region

Abstract

Abstract. Detailed knowledge on surface water distribution and its changes is of high importance for water management and biodiversity conservation. Landsat-based assessments of surface water, such as the Global Surface Water (GSW) dataset developed by the European Commission Joint Research Centre (JRC), may not capture important changes in surface water during months with considerable cloud cover. This results in large temporal gaps in the Landsat record that prevent the accurate assessment of surface water dynamics. Here we show that the frequent global acquisitions by the Moderate Resolution Imaging Spectrometer (MODIS) sensors can compensate for this shortcoming, and in addition allow for the examination of surface water changes at fine temporal resolution. To account for water bodies smaller than a MODIS cell, we developed a global rule-based regression model for estimating the surface water fraction from a 500 m nadir reflectance product from MODIS (MCD43A4). The model was trained and evaluated with the GSW monthly water history dataset. A high estimation accuracy (R2=0.91, RMSE =11.41 %, and MAE =6.39 %) was achieved. We then applied the algorithm to 18 years of MODIS data (2000–2017) to generate a time series of surface water fraction maps at an 8 d interval for the Mediterranean. From these maps we derived metrics including the mean annual maximum, the standard deviation, and the seasonality of surface water. The dynamic surface water extent estimates from MODIS were compared with the results from GSW and water level data measured in situ or by satellite altimetry, yielding similar temporal patterns. Our dataset complements surface water products at a fine spatial resolution by adding more temporal detail, which permits the effective monitoring and assessment of the seasonal, inter-annual, and long-term variability of water resources, inclusive of small water bodies.

Development of the global dataset of Wetland Area and Dynamics for Methane Modeling (WAD2M)

“…Characterizing the seasonal and interannual variation in wetland extent is critical to improving global-scale wetland CH 4 modeling. Contemporary evidence from remote sensing (Alsdorf et al, 2000(Alsdorf et al, , 2007Hu et al, 2018;Lunt et al, 2019;Melack et al, 2004;Pandey et al, 2021;Prigent et al, 2007Prigent et al, , 2012Rodell et al, 2018) and field monitoring (Dunne and Aalto, 2013) suggests that global wetlands, especially tropical floodplains, have a significant seasonal cycle and interannual variability in spatial extent that depend on changes in water balance (i.e., precipitation, runoff, and evapotranspiration) and local topography. Despite the critical importance of spatial and temporal changes in wetland area, there are large discrepancies among the estimates of global wetland extent (Aires et al, 2018;Melton et al, 2013;Pham-Duc et al, 2017;Wania et al, 2013) and only a limited number of available global products characterize temporal dynamics in wetland extent (Gallant, 2015;Huang et al, 2014;Prigent et al, 2007.…”

Combining Multi-Sensor Satellite Imagery to Improve Long-Term Monitoring of Temporary Surface Water Bodies in the Senegal River Floodplain

“…The results demonstrate the clear limitations before 2013 of GSW datasets built on one image every 30 days, to monitor flood dynamics in floodplains. These are notably insufficient to reproduce surface water dynamics in the smallest (<1 km 2 ), temporary water bodies with flash floods [7,82,83] and are also insufficient in floodplains with slow flood dynamics. Using the full Landsat archive provides observations every 24 days on average over 1999-2019, but the benefits of integrating additional data sources remain indisputable, e.g., to capture flood peaks in 2003 and 2004 missed here by Landsat.…”

Determining Temporal Uncertainty of a Global Inland Surface Water Time Series

“…For example, compared to MODIS (daily revisit times of two sensors), Landsat sensor imagery exhibits a significantly lower temporal resolution (~16-day revisit period coupled with cloud contamination). Consequently, derived temporal uncertainty has to be interpreted in a different context, as important dynamics might be missed [39,40]. If information is generated with a detail level that is not sustained by data availability, uncertainty increases due to sampling errors.…”