• The historical map of change should be interpreted in relation to the historical map of statistical confidence. We recommend that areas where there is low statistical confidence in historical change (grey areas on map of statistical confidence) should NOT be used for making climate-related decisions.

 

• The ClimateWizard uses a linear regression-style trend analysis to calculate the rate of climate change within every grid cell. This analysis produces maps of both the rate of change, as well as the statistical confidence in this change—more technically this is called statistical significance (p-value). Using the ClimateWizard, you can easily compare the map of change with the map of statistical confidence by toggling between them using the buttons above the maps. When presenting these maps graphically, download both the map of change and the map of statistical confidence and present them next to one another. Or when downloading GIS data, make sure to get both the map of change and map of statistical confidence, and use them together for running analyses and making decisions.

 

• Historical climate maps are developed from weather station observations that have been spatially interpolated to create a seamless map of climate information. We recommend that single grid cells NOT be used for making climate-related decisions, but rather decisions should be based on many grid cells showing region patterns of climate change with high statistical confidence.

 

• A variety of gridded climate data sets have been created that range in spatial resolution, geographic extent, time period, and climate variables [e.g. 1,5,6], and different data sets may contain different estimates of climate for the same location [7].  The ClimateWizard currently contain two such historic datasets that each have their own different caveats: the 1901-2002 CRU TS 2.1 global climate data at ~50km (0.5 degree, http://www.cru.uea.ac.uk/~timm/grid/CRU_TS_2_1.html); and the 1895-2006 PRISM climate data set for the contiguous United States (Lower 48 states, ~4km resolution, http://www.prism.oregonstate.edu/). Both of these datasets were initially developed for use in environmental models, and have been adapted for use in climate change analysis. Note that for some regions and time periods, climate-station data were insufficient for developing accurate climate estimates, particularly early in the twentieth century, and are represented as white grid cells having no data. 

Future Projected Analyses

 

• Ensembles of multiple models should be used to identify where models agree on climate change and where there is disagreement between models. We recommend that areas with severe disagreement between models NOT be used for making climate-related decisions—only areas with high model agreement should be used. This is particularly important when assessing future projected precipitation, as models often disagree on the direction of change in future precipitation.

 

• There are many future climate projections produced by different GCMs and a number of different types of uncertainty that accompany these projections [9]. Randall et al. [10] describe some of the strengths and weaknesses of GCM simulations.

 

• Even though the grid cell resolution of the data is ~50 km (0.5 degree) globally, and ~12 km (1/8th degree) for the contiguous United States (Lower 48 states), these data have been statistically “downscaled” from GCMs that were originally run at 2.5-3.5 degree resolutions. Although these downscaling techniques better estimate the actual projected temperature or amount of precipitation in a specific grid cell, they still only represent coarse scale global climate processes, and do not include regional or fine scale.

 

• Since the spatial resolution of GCMs (2.5-3.5 degrees) is often too coarse for many scientific, management, and planning questions, a variety of methods have been developed for downscaling climate data to create finer spatial resolution data sets (Fowler et al. 2007). However, increased spatial resolution, however, does not necessarily mean that data are more accurate. In some cases, the downscaling methods used to develop coarse-scale climate data are the same as those used to create fine-scale climate data, and these methods may ignore important processes that influence climate at regional and local scales (Daly 2006). It is important to remember that future climate simulations are projections of future climate, not accurate predictions of future climate change for any particular location or specific moment in time.  

Different Time Periods

 

• The time period analyzed can greatly influence degree climate change.

 

• The time period over which climate data are analyzed should be carefully selected to avoid drawing incorrect or biased conclusions [13]. The choice of the time period to analyze can influence the results of the analysis. For example, based on our analysis of the CRU TS 2.1 data, global mean temperature increased at 0.11 C/decade from 1941 to 2002, 0.16 C/decade from 1951 to 2002, 0.23 oC/decade from 1961 to 2002, and 0.33 oC/decade from 1971 to 2002 (See Girvetz et al. in review). Thus one’s conclusions about the pace of warming could vary threefold depending on the time interval sampled.

Absolute Change Versus Percent Change

 

• The ClimateWizard can also be used to produce absolute change and percent change analyses. Absolute change and percent change can both be useful for looking at temporal and spatial patterns of climate change, depending on the goals of the analysis being run, but each require an understanding of the climate data used in the analysis.

 

• For example, two grid cells may have the same future projected percent decrease in precipitation but a 10% decrease in precipitation for a grid cell in northern Africa that has relatively little annual precipitation may represent a much smaller precipitation amount than a 10% decrease for a grid cell in the Brazilian Amazon region that receives a large amount of annual precipitation. Likewise, a grid cell in the African Sahel region, for example, may have the same percent change in precipitation during January and August, but because it rains much more during August due to the monsoonal rains, the absolute precipitation change would be much greater during August. Thus, the ecological effect of the same simulated future percent decrease in precipitation may be very different depending on a number of factors, including the region’s total amount of annual precipitation, how the simulated precipitation decrease is distributed throughout the year, and the particular sensitivity to precipitation changes of the organism or system being studied. While we only presented percent change in precipitation, the ClimateWizard has the ability to calculate either percent or absolute change in precipitation.