Spatio-temporal rainfall interception loss at the catchment scale from earth observation in a data-scarce area, Northern Ethiopia

Abstract

Rainfall interception loss (EI) is an important spatio-temporal variable in hydrological studies. The EI is required as input of integrated hydrological models, which nowadays are increasingly applied in water resources management. However, the EI is typically not considered or arbitrarily simplified, which causes wrong model parameterization and typical overestimation of water resources. Such problem can be particularly consequential in water-limited environments, where EI represents a substantial percentage of rainfall. The EI is rarely quantified experimentally because field data acquisition for its direct estimation is cumbersome and costly, especially at large catchment/basin scales. Hence, Earth Observation Satellites (EOS) can be used as an alternative source of data acquisition to quantify EI, especially at large spatial scales. The main aim of this study was to estimate the daily variability of rainfall interception loss in the data-scarce Zamra catchment (1588 km2) in Ethiopia, applying the EOS solution of the revised Gash model (RGM). Next to EOS, this study utilized four automated weather station data. To test the RGM assumption permitting the use of rainfall as one storm per rainy day (instead of the event-based rainfall), high temporal resolution (15 min) automated weather-station data was used. The test confirmed the validity of that assumption in the Zamra catchment (ZC), so the daily bias-corrected by station data rainfall was used as input. The weather station data was also used to define the event-based ratio of the mean wet-canopy evaporation rate to the mean rainfall rate (Ec¯/R¯) ranging from ∼0.037 to 0.041 at the low-elevated stations to ∼0.063 at the high-elevated station. As the variability of the Ec¯/R¯ of the three low-elevated stations was pretty uniform, the mean 0.04 was assumed for areas below 2000 m a.s.l., but for the areas above, the Ec¯/R¯ was assigned as altitude-dependent applying geographical weighted regression method. Other biophysical parameters, the fractional canopy cover and the canopy storage capacity obtained from leaf area index, were derived from Sentinel 2. The final EI was highly spatially and temporally variable, ranging from zero at bare lands, to 30% of annual rainfall in forested areas and with ∼4% mean of annual rainfall for the whole ZC. While applying EOS solution of RGM, facilitated with limited amount of ground-based data, the presented study offers an efficient and realistic way of spatio-temporal quantification of EI, particularly suitable for large data-scarce areas such as the ZC.