Gaining insight into Andean snowpack climatology and change using a snow reanalysis approach applied over the Landsat satellite record.

摘要

This dissertation presents the results of a snowpack estimation system based on the integration of remotely sensed data with snow modeling over the extratropical Andes domain (27°S to 37°S). The framework is based on the Bayesian principles of data assimilation: by assimilating Landsat fractional snow cover imagery in an ensemble snow model, the snow model estimates are conditioned by the observed depletion as sensed by the remote sensing platform during 1984 to present. The snow model is forced using the MERRA atmospheric reanalysis data set. Uncertainty of MERRA was characterized using in-situ precipitation and temperature observations. The results of the framework are daily ensemble of estimates of snowpack states with a spatial resolution of 180m, distributed throughout the domain. Verification of the estimates was performed using in-situ snow survey data taken over several headwaters of the domain during the 2009-2014 winter and spring months. Results of the in-situ verification showed that the posterior estimates of the framework are in general much more accurate than the prior estimates, with significant reductions in mean error, root mean square error and increases in correlation. The error metrics were invariant to the different physiographic characteristics of each the snow survey sites and the fSCA imagery availability over each of the surveyed data points. An analysis of runoff and the SWE estimates over a large headwater basin of the Aconcagua river showed a strong correlation between surface streamflow and peak SWE estimates. The framework was implemented for the regional domain and the SWE volume for each of the watersheds was analyzed. This constituted the first analysis of the impact of El Nino in the water stored as snow for the extratropical Andes region. The effect of El Nino is particularly important for the northern watersheds (north of 32°S) of the domain. El Nino was related to the wettest year observed, which presented SWE volumes an order of magnitude larger than normal years. For Southern watersheds, the effect of El Nino is diminished, with SWE volumes showed a marked reduction in interannual variability with respect to the northern domain. Longitudinal SWE transects were analyzed in order to quantify the effect of the Andes barrier in orographic enhancement and suppression, which affects SWE accumulation directly. The analysis showed significant variability of the effect of the barrier in SWE, with varying enhancing and suppression rates throughout the domain. The strongest magnitude of the effect were seen over the southern extent of the domain (35°S), with reductions in the long-term SWE average of an order of magnitude between windward and leeward slopes.