Tue, Aug 03, 2021:On Demand
Background/Question/Methods
Ecologists studying the effects of climate change often use fine-scale climate projections for their research. However, general circulation models (GCMs) that produce climate projections and their associated downscaled estimates relevant to ecological systems are of varied quality. Currently, there are no simple tools for selecting the most appropriate downscaled projections for ecological problems. Major model deficiencies resulting from unrealistic or missing earth system processes cannot be corrected during the downscaling process. Identification of model deficiencies is necessary to select robust climate models prior to the application of downscaling techniques. Here, we develop an approach to selecting a robust suite of climate model projections that can be used for ecological problems. We show the utility of this approach in identifying climate projections for predicting future changes in the geographic distribution of Vesicular Stomatitis (VS), a livestock disease that occurs throughout the western US. Our model selection method involves examining GCM simulations of 20th century seasonal temperature and precipitation using 10 performance metrics relevant to our VS application. We assessed simulations of seasonal means, trend, and variability as compared to observations and identified the subset of GCMs that performed well. We then used this information to choose appropriate downscaled projections.
Results/Conclusions Results show a wide range in model skill even for long-term mean seasonal climate where the highest skill was expected. Biases in long-term mean winter and summer climate were found to be on the order of interannual variability in seasonal climate. Many models actually reversed the sign of the observed 100-yr trend in winter and summer temperature and precipitation over the western US. Results identified a suite of 8 models out of 37 from the CMIP5 archives that performed relatively well according to our performance metrics. Interestingly, many of the 8 best performing models were not included in the Multivariate Adaptive Constructed Analogs (MACA) downscaled dataset, a product that is often used by ecologists. This type of analysis is essential to assist ecologists in an objective selection of climate projections to be used in their research. Selection of the most appropriate model projections is a critical step in studying the effects of a changing climate on ecological systems. Our approach improves confidence in predictions of future ecosystem dynamics by reducing the impact of model errors in climate change projections.
Results/Conclusions Results show a wide range in model skill even for long-term mean seasonal climate where the highest skill was expected. Biases in long-term mean winter and summer climate were found to be on the order of interannual variability in seasonal climate. Many models actually reversed the sign of the observed 100-yr trend in winter and summer temperature and precipitation over the western US. Results identified a suite of 8 models out of 37 from the CMIP5 archives that performed relatively well according to our performance metrics. Interestingly, many of the 8 best performing models were not included in the Multivariate Adaptive Constructed Analogs (MACA) downscaled dataset, a product that is often used by ecologists. This type of analysis is essential to assist ecologists in an objective selection of climate projections to be used in their research. Selection of the most appropriate model projections is a critical step in studying the effects of a changing climate on ecological systems. Our approach improves confidence in predictions of future ecosystem dynamics by reducing the impact of model errors in climate change projections.