Development Of A Hydrological Model Driven And Calibrated By Multisource Remote Sensing Data In Poorly Gauged Basins

Since the 20 th century,global temperatures have been rising and human activities have intensified.As a result,changes in water storage and the water cycle have influenced water supply,ecosystems,and the environment.In China,glaciers and perennial snowpack on most areas of the Tibetan Plateau known as Asian’s water towers have been melting due to rising temperatures.Therefore,it is critical to understand whether there will be sufficient ice and snow meltwater to recharge the rivers and potential impacts on the downstream water supply,irrigation,and ecology.Predicting changes in glaciers and other hydrological variables can be studied by using hydrological models.Applying a conceptual hydrological model requires sufficient gauged data to calibrate the model parameters.However,due to the harsh climate,complex terrain,relatively backward infrastructure and lack of observational networks,the Tibetan Plateau is a typical poorly gauged region.It is therefore difficult to determine hydrological model parameters in such areas.This study focuses on how to perform hydrological modeling and effectively estimate model parameters in poorly gauged basins on the Tibetan Plateau.We have further improved the CREST-snow hydrological model that was developed for use in alpine mountainous areas in three aspects.First,we added an elevation-band model to CREST-snow with a relatively low spatial resolution,which can achieve the balance between the computation efficiency and simulation accuracy.Second,a groundwater module was added to CREST-snow based on the free water storage of the Xinanjiang model,and the formulas of calculating total water storage in the model were revised.The problem of the time lag and lower amplitude of the simulated total water storage relative to GRACE satellite observations that exist in CREST-snow has been resolved.Third,the NSGA-II multi-objective optimization algorithm was applied to obtain optimal solution sets instead of the unique optimal solution,and then the uncertainty of the simulation can be analyzed.Based on the enhanced hydrological model,a new model CREST-RS with a calibration scheme using multisource remote sensing data was developed.In particular,satellite altimetry water levels and other multisource remote sensing data were jointly used in combination with a few in situ flow measurements(e.g.,4-day observations).The model was applied to the Yarlung Tsangpo River and the middle reaches of the Brahmaputra River.Results show that if the satellite altimetry water level and a few flow data were used for model calibration,CREST-RS can effectively simulate total runoff with the Nash-Sutcliffe efficiency coefficient(NSE>0.9)and water storage(correlation coefficient >0.9).When only satellite altimetry water level data were used,the runoff was slightly underestimated during the low-flow period,but can be well simulated during the high-flow period(NSE>0.87).These results show that the satellite altimetry water level data can be used as an alternative of in situ flow data for model calibration.This study provides an effective approach for estimating hydrological model parameters and performing hydrological modeling to understand cryospheric hydrological processes in poorly gauged high-mountain basins on the Tibetan Plateau and potentially similar regions globally.