Premium
Real‐time estimation of snow water equivalent in the U pper C olorado R iver B asin using MODIS ‐based SWE Reconstructions and SNOTEL data
Author(s) -
Schneider Dominik,
Molotch Noah P.
Publication year - 2016
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1002/2016wr019067
Subject(s) - snow , context (archaeology) , regression , regression analysis , linear regression , environmental science , snowpack , terrain , mean squared error , statistics , hydrology (agriculture) , mathematics , meteorology , cartography , geography , geology , archaeology , geotechnical engineering
Abstract Changes in climate necessitate improved snowpack information to better represent anomalous distributions of snow water equivalent (SWE) and improve water resource management. We estimate the spatial distribution of SWE for the Upper Colorado River basin weekly from January to June 2001–2012 in quasireal‐time by two regression techniques: a baseline regression of in situ operationally measured point SWE using only physiographic information and regression of these in situ points combining both physiographic information and historical SWE patterns from a remote sensing‐based SWE reconstruction model. We compare the baseline regression approach to our new regression in the context of spatial snow surveys and operational snow measuring stations. When compared to independent distributed snow surveys, the new regression reduces the bias of SWE estimates from −5.5% to 0.8%, and RMSE of the SWE estimates by 8% from 0.25 m to 0.23 m. Notable improvements were observed in alpine terrain with bias declining from −38% to only 3.4%, and RMSE was reduced by 13%, from 0.47 to 0.41 m. The mean increase in cross‐validated r 2 for the new regression compared to the baseline regression is from 0.22 to 0.33. The largest increase in r 2 in any one year is 0.19, an 83% improvement. The new regression estimates, on average, 31% greater SWE depth than the baseline regression in areas above 3000 m elevation, which contributes up to 66% of annual SWE volume in the driest year. This indicates that the historical SWE patterns from the reconstruction adds information to the interpolation beyond the physiographic conditions represented by the SNOTEL network. Given that previous works using SWE reconstructions were limited to retrospective analyses by necessity, the work presented here represents an important contribution in that it extends SWE reconstructions to real‐time applications and illustrates that doing so significantly improves the accuracy of SWE estimates.