Detailed spatial information of changes in surface water extent is needed for water management and biodiversity conservation, particularly in drier parts of the globe where small, temporally-variant wetlands prevail. Although global surface water histories are now generated from 30 m Landsat data, for many locations they contain large temporal gaps particularly for longer periods (>10 years) due to revisit intervals and cloud cover. Daily Moderate Resolution Imaging Spectrometer (MODIS) imagery has potential to fill such gaps, but its relatively coarse spatial resolution may not detect small water bodies, which can be of great ecological importance. To address this problem, this study proposes and tests options for estimating the surface water fraction from MODIS 16-day 500 m Bidirectional Reflectance Distribution Function (BRDF) corrected surface reflectance image composites. The spatial extent of two Landsat tiles over Spain were selected as test areas. We obtained a 500 m reference dataset on surface water fraction by spatially aggregating 30 m binary water masks obtained from the Landsat-derived C-version of Function of Mask (CFmask), which themselves were evaluated against high-resolution Google Earth imagery. Twelve regression tree models were developed with two approaches, Random Forest and Cubist, using spectral metrics derived from MODIS data and topographic parameters generated from a 30 m spatial resolution digital elevation model. Results showed that accuracies were higher when we included annual summary statistics of the spectral metrics as predictor variables. Models trained on a single Landsat tile were ineffective in mapping surface water in the other tile, but global models trained with environmental conditions from both tiles can provide accurate results for both study areas. We achieved the highest accuracy with Cubist global model (R2 = 0.91, RMSE = 11.05%, MAE = 7.67%). Our method was not only effective for mapping permanent water fraction, but also in accurately capturing temporal fluctuations of surface water. Based on this good performance, we produced surface water fraction maps at 16-day interval for the 2000–2015 MODIS archive. Our approach is promising for monitoring surface water fraction at high frequency time intervals over much larger regions provided that training data are collected across the spatial domain for which the model will be applied.
|Number of pages||11|
|Journal||International Journal of Applied Earth Observation and Geoinformation|
|Publication status||Published - 1 Apr 2018|
- Machine learning
- Multi-temporal images
- Time series