Hernandez, Manuel (2015-08). Assessment of Statistically Downscaled CMIP5 Simulations of the North American Monsoon System. Master's Thesis. Thesis uri icon


  • The North American Monsoon System (NAMS) has been projected to undergo a redistribution of precipitation due to enhanced spring convective barriers. Previous studies have utilized coarse-scaled global climate models (GCMs) from the Coupled Model Intercomparison Project (CMIP) Phase 5 to assess the impacts climate change over the southwest United States and northern Mexico. However, GCMs' spatial resolution limits the representation of regional to local scale processes, especially over complex terrain. Thus, the focus of this research is to determine whether statistically downscaled GCMs are viable tools for assessments of the NAMS. First, two reanalysis systems are compared to station observations over the NAMS region using a suite of goodness of fit measures. This evaluation seeks to illustrate the suitability of each product in quantifying the climate characteristics associated with the NAMS over the monsoon-affected region. Second, simulated temperature and precipitation characteristics of downscaled CMIP5 GCMs are assessed against the reanalysis product identified in our previous analysis. The use of downscaled CMIP5 model output has been introduced as a means for evaluating the climate system and their impacts on a regional to local scale. This second objective will demonstrate the improvements, or limitations, in using statistical downscaling to assess the NAMS. Two highly resolved, gridded reanalyses, the North American Regional Reanalysis (NARR) and the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Reanalysis (ERA-I) are assessed in their ability to simulate temperature and precipitation in comparison to in-situ Global Historical Climatology Network (GHCN) observations. Results indicate similar temperature agreement for both reanalyses, while simulated precipitation is better captured in NARR. Precipitation demonstrated varying degrees of agreement, signifying difficulties when simulating precipitation. However, seasonal errors suggest better performance with NARR when compared to ERA-I, thus chosen as the superior reanalysis. Statistically downscaled CMIP5 output was then compared to NARR to explore the improvements, or limitations, in using downscaled climate model output. Various downscaled CMIP5 outputs, including from CNRM-CM5, HadGEM2-CC, and HadGEM2-ES, demonstrated improved representation of temperature and precipitation when compared to their coarsely resolved counterparts. Furthermore, the downscaled temperature and precipitation climatologies reveal better portrayal of the monsoon seasonality, capturing the onset and decay phase of the NAMS. However, all three downscaled models exhibit a warm, wet bias over complex terrain, indicating continued difficulties in reproducing observations. These results confirm enhancements achieved via statistical downscaling, and suggest this statistical approach to improve future climate projections. Thus, statistically downscaled model output serves a viable tool for improved climate projections for water resources and future adaptation.

publication date

  • August 2015