A Novel Reanalysis Dataset for Improving Seasonal Snowpack Characterization: Application to the Sierra Nevada (USA) Over the Landsat 5 Record (1985-2011)

Highly variable spatial and temporal patterns in snow properties are the norm in complex terrain, yet most mountain regions remain significantly under-sampled with respect to in situ data. We argue that significant strides in our understanding of snow processes in montane regions can be made using a regional reanalysis approach that merges information from uncertain model estimates and globally available remote sensing measurements. While many large-scale reanalysis datasets (e.g. NARR, NCEP/NCAR, MERRA, etc.) exist, their coarse spatial resolution and modeling/assimilation approaches are generally insufficient and/or inaccurate for direct application to analysis of snow processes in remote complex mountainous terrain. The newly developed state-of-the-art snow water equivalent (SWE) reanalysis approach presented in this paper is focused on snow and is based on the leveraging of long-term Landsat fractional snow-covered area (fSCA) datasets together with existing meteorological reanalysis products and snow modeling, merging all information streams within a data assimilation framework. The method is applied over the Sierra Nevada (USA) for the Landsat 5 record (1985-2011). The resolution (daily, 90-meter), temporal extent (27 years), and accuracy (mean and root-mean-squared errors less than 3 and 14 cm respectively and correlation greater than 0.94 compared with in situ SWE observations) provide a unique dataset for investigating snow processes. Preliminary analysis has focused on the peak SWE climatology over the reanalysis period. The pixel-wise peak SWE volume over the domain was found to be 19.9 km3 on average with a range of 9.8-35.8 km3. The 27-year climatology for range-wide peak SWE was found to be 16.4 km3 and occurred on 28 March. The results show that the assumption that 01 April SWE is representative of the peak range-wide SWE can lead to significant underestimation both on average (16%) and on an inter-annual basis (up to 47%). The reanalysis is being extended to present using Landsat 7 and 8 data, which will allow for characterizing recent drought years compared to the 30-year climatology. The methods are general and therefore designed to be extended to other important mountain regions with significant seasonal snowpack.