Agreement and uncertainty among climate change impact models: A synthesis of sagebrush steppe vegetation predictions

Ecologists have built numerous models to predict how climate change will impact vegetation, but these predictions are difficult to validate, making their utility for land management planning unclear. In the absence of direct validation, researchers can ask whether predictions from varying models are consistent. Here, we analyzed 43 models of climate change impacts on sagebrush (Artemisia tridentata Nutt.), cheatgrass (Bromus tectorum L.), pinyon-juniper (Pinus spp. and Juniperus spp.), and forage production on Bureau of Land Management (BLM) lands in the United States Intermountain West. These models consistently projected pinyon-juniper declines, forage production increases, and the potential for sagebrush increases in some regions of the Intermountain West. In contrast, models of cheatgrass did not predict consistent changes, making cheatgrass projections uncertain. While differences in emission scenarios had little influence on model projections, predictions from different modeling approaches were inconsistent in some cases. This model-choice uncertainty emphasizes the importance of comparisons such as this.

The projected vegetation changes have important management implications for agencies such as the BLM. Pinyon-juniper declines would reduce the BLM’s need to control pinyon-juniper encroachment, and increases in forage production could benefit livestock and wildlife populations in some regions. Sagebrush habitat may benefit where sagebrush is predicted to increase, but sagebrush conservation and restoration projects will be challenged in areas where climate may not remain hospitable. Projected vegetation changes may also interact with increasing future wildfire risk, potentially impacting vegetation and increasing management challenges related to fire.

Included in this page are the data and code used to complete this analysis and visualize results. This includes the original images of model results used in our analysis, and the code used to process and analyze these images to produce our final results.