Predictive Uncertainty Estimation on a Precipitation and Temperature Reanalysis Ensemble for Shigar Basin, Central Karakoram

Water

Biswajit Mukhopadhyay¹  Gabriele Coccia²  Paolo Reggiani³ 

[1] DHI (India) Water and Environment, New Delhi 110020, India;Department of Biological, Geological, and Environmental Sciences, University of Bologna, Bologna 40126, Italy;Department of Civil Engineering, University of Siegen, Siegen 57068, Germany;

关键词: Karakoram;    Shigar;    Upper Indus;    predictive uncertainty;    reanalysis;    ensemble;    precipitation;    temperature;    Bayesian paradigm;    uncertainty post-processor;    poorly-gauged basin;   

DOI  :  10.3390/w8060263

来源: DOAJ

【 摘 要 】

The Upper Indus Basin (UIB) and the Karakoram Range are the subject of ongoing hydro-glaciological studies to investigate possible glacier mass balance shifts due to climatic change. Because of the high altitude and remote location, the Karakoram Range is difficult to access and, therefore, remains scarcely monitored. In situ precipitation and temperature measurements are only available at valley locations. High-altitude observations exist only for very limited periods. Gridded precipitation and temperature data generated from the spatial interpolation of in situ observations are unreliable for this region because of the extreme topography. Besides satellite measurements, which offer spatial coverage, but underestimate precipitation in this area, atmospheric reanalyses remain one of the few alternatives. Here, we apply a proven approach to quantify the uncertainty associated with an ensemble of monthly precipitation and temperature reanalysis data for 1979–2009 in Shigar Basin, Central Karakoram. A Model-Conditional Processor (MCP) of uncertainty is calibrated on precipitation and temperature in situ data measured in the proximity of the study region. An ensemble of independent reanalyses is processed to determine the predictive uncertainty of monthly observations. As to be expected, the informative gain achieved by post-processing temperature reanalyses is considerable, whereas significantly less gain is achieved for precipitation post-processing. The proposed approach indicates a systematic assessment procedure for predictive uncertainty through probabilistic weighting of multiple re-forecasts, which are bias-corrected on ground observations. The approach also supports an educated reconstruction of gap-filling for missing in situ observations.

【 授权许可】

Unknown