To characterize climate change impacts in long range simulations using economic and Integrated Assessment (IA) models, we are also developing statistical emulators (i.e. response functions) that map from inputs (e.g. irrigation and fertilizer application) and atmospheric conditions (e.g. atmospheric CO2 and regional temperature and precipitation measures) to agricultural productivity measures at a variety of scales. We are developing a number of aggregation and scaling methodologies to translate high-resolution gridded products to decision-relevant environmental and political scales (e.g. county, watershed, or national boundaries). These methodologies must be versatile enough to map vulnerability, impact, and adaptation (VIA) measures to arbitrary spatial scales, while consistently tracking ensemble uncertainty information throughout.
The emulation tool provides statistical representations of the behaviors of a number of general circulation models (GCMs) from different research groups. The emulator is designed to mimic how the larger models would have responded, and can provide projections of regional temperatures for any user-chosen scenario of future CO2 concentrations.
The emulator is ‘trained’ on publicly-available model simulations of scenarios specified for the IPCC 5th Assessment Report. Users can explore emulated model responses to three IPCC scenarios (RCPs 2.6, 4.5, and 8.5) and one scenario of continuous exponential growth in CO2. Users can also upload their own CO2 scenarios.
SCREENSHOTS OF THE CLIMATE EMULATOR TOOL
Statistical emulation of climate model output involves using a ‘training set’ of simulations for each model to fit the parameters of a simple statistical model that describes the climate response to changing CO2 concentrations. In the tool here, each region is fit separately.
The training set used is archived GCM simulations made as part of the Coupled Model Intercomparison Project 5 (CMIP5), and archived at the Earth Systems Grid database (available from various sites, including here or here). The simulations use CO2 scenarios developed for the Intergovernmental Panel on Climate Change 5th Assessment Report. These ‘Representative Concentration Pathways’, or RCPs, describe the CO2 concentrations associated with various potential emissions scenarios (see graphic on the left). Where available, we use RCP 4.5, RCP 6.0, and RCP 8.5 as the training set. For some models that did not archive all scenarios, we use only RCP 4.5 and RCP 8.5.
The statistical model used for emulation is described in detail in Castruccio et al 2013