Developing a Snowmelt Forecast Model in the Absence of Field Data

In data poor regions predicting water availability is a considerable challenge for water resource managers. In snow-dominated watersheds with minimal in situ measurements, satellite imagery can supplement sparse data networks to predict future water availability. This technical note presents the first phase of an operational forecast model in the data poor Elqui River watershed located in northern Central Chile (30°S). The approach applies remotely-sensed snow cover products from the Moderate Resolution Imaging Spectrometer (MODIS) instrument as the first order hydrologic input for a modified Snowmelt Runoff Model. In the semi-arid Elqui River, snow and glacier melt are the dominant hydrologic inputs but precipitation is limited to up to six winter events annually. Unfortunately winter access to the Andean Cordillera where snow accumulates is incredibly challenging, and thus measurements of snowpack are extremely sparse. While a high elevation snow monitoring network is under development, management decisions regarding water resources cannot wait as the region is in its eighth consecutive year of drought. Our model applies a Monte Carlo approach on monthly data to determine relationships between lagged changes in snow covered area and previous streamflow to predict subsequent streamflow. Despite the limited data inputs the model performs well with a Nash-Sutcliffe Efficiency and R2 of 0.830 and 0.833 respectively. This model is not watershed specific and is applicable in other regions where snow dominates hydrologic inputs, but measurements are minimal.