Precipitation downscaling and its use in the assessment of hydrologic effects of climate variability and change

Abstract

The combined use of physically based meteorological and hydrologic modeling can enhance the accuracy and reliability of hydrologic assessments. However, because of the different characteristic spatial and temporal scales of atmospheric and hydrologic processes, and their large spatial and temporal variability, a methodology must be developed for an appropriate coupling of these two kinds of models. In this thesis, a new methodology is developed that couples GCM macro-scale climate model output and distributed watershed rainfall-runoff models through a precipitation downscaling algorithm based on random cascades. The downscaling model is a composite of a Stochastic Space-Time Disaggregation Model (SSTDM) that preserves the spatial and temporal dependency characteristics and an Intermittency multiplicative Random Cascade Model (IRCM) that incorporates the statistical self-similarity, spatial intermittency, and the spatial storm cluster formation of precipitation. Each of the above sub-models is applied over a specific range of scales. The corresponding scale ranges are non-overlapping and the reference level at which the models are switched is a function of the correlation structure of the observations. High-resolution (2 x 2km) NEXRAD observations were used for the characterization of the spatial and temporal distribution, the correlation structure, and the statistical self-similarity of precipitation. The scaling analysis demonstrated that a mono-fractal structure of self-similarity of small-scale field less than 32 km is well developed. Based on the results of this analysis process, GCM precipitation was downscaled to the 2 km-scale. In order to assess the hydrologic impacts of climate variability, the output from the Canadian Climate Center Global Circulation Model was downscaled using the new downscaling scheme described above. The downscaled fields were then used to drive the HEC HMS, a physically based, semi-distributed hydrologic model. The South Platte Headwaters basin (2,465km2) was selected as the study area for the assessment. Climate output corresponding to the IPCC SRES (Intergovernmental Panel on Climate Change Special Report on Emission Scenarios) "B2" scenario was used for the analysis. Applying the downscaling model to produce precipitation fields at different spatial scales that effectively smoothed the spatial variability led to a reduction of streamflow, so directly applying the GCM precipitation with no downscaling can give rise to gross underestimation of streamflows. Hydrologic response of the South Platte (i.e., streamflows, evapotranspiration, etc.) was simulated for 2 temporal windows (2011-2020 and 2081-2090) corresponding to the above GCM scenario. Results indicate that the JJA runoff volume is decreased 15.37% and the JJA peak flow rate is decreased 18.01%. Much of these reductions are the result of the decrease of JJA precipitation and increase of JJA evaporation.