Challenges encountered while assembling data sets for the analysis of climate and snowpack variability in the Rocky Mountain region

We have used data from multiple sources (i) as input for snowmelt runoff modeling in the Upper Rio Grande and (ii) to explore relationships between El Niño Southern Oscillation (ENSO) events, climate and snowpack in the Southern and Northern Rockies. Here, we present some of the challenges we experienced in using climate and remotely sensed data in our research. We obtained climate data from the National Oceanographic and Atmospheric Administration's National Climatic Data Center (NOAA NCDC), the Remote Automated Weather Station (RAWS) archive and the Natural Resources Conservation Service National Water and Climate Center (NRCS NWCC). The first challenge we encountered was identifying all functioning climate stations in an area. At the time of searching, there was no single source that listed all climate stations from all agencies. In downloading and pre-processing data, a minor inconvenience was the units used by different providers for temperature and precipitation data. For example, the NCDC delivers temperature and precipitation data in degrees Fahrenheit and inches respectively while the NWCC uses degrees Celsius and inches. A second challenge in pre-processing the climate data is identifying erroneous values. Brief analysis of the temperature and precipitation data is often sufficient to identify obvious problems in the data series but more subtle errors are harder to detect. A third challenge in pre-processing the climate data is how to replace erroneous or missing values. To address the challenge of missing data we have developed our own set of rules and methods for repairing data sets and for deciding which climate records should be discarded. For input into the Snowmelt Runoff Model, we used two sources of remotely sensed data for estimating snow covered area (SCA). Moderate Resolution Imaging Spectrometer (MODIS) snow cover data sets were acquired from the National Snow and Ice Data Center. Landsat Thematic Mapper (TM) imagery was obtained from the USGS Earth Resources Observation and Science Center and was classified to binary SCA. In comparing snow depletion curves modeled from the different sources of snow cover data, we found that MODIS snow cover products can be problematic in our study areas. Comparison of SCA estimated from TM data and MODIS snow products indicate that the MODIS binary product underestimates SCA under conditions of discontinuous snow cover. The MODIS fractional product underestimates SCA throughout the melt season because it does not account for sub-canopy snow cover. These results indicate the need for greater validation of snow cover products in mountainous, forested areas. In summary, the main advantage of all the climate and remotely sensed data sets we have used in our research is that they are freely and easily available through the internet. But in both cases, the data must be used with a degree of caution and knowledge of the shortcomings. We propose that the use of these data would be greatly facilitated by universal standards for pre-processing and validation.