Extreme precipitation events pose a significant threat to public safety, natural and managed resources, and the functioning of society. Changes in such high-impact, low-probability events have profound implications for decision-making, preparation and costs of mitigation and adaptation efforts. Understanding how extreme precipitation events will change in the future and enabling consistent and robust projections is therefore important for the public and policymakers as we prepare for consequences of climate change.

Projection of extreme precipitation events, however, particularly at the local scale, presents a critical challenge: the climate model-based simulations of precipitation that we currently rely on for such projections—general circulation models (GCMs)—are not very realistic, mainly due to the models’ coarse spatial resolution. This coarse resolution precludes adequate representation of highly influential, small-scale features such as moisture convection and topography. Regional circulation models (RCMs) provide much higher resolution and better representation of such features, and are thus often perceived as an optimum approach to producing more accurate heavy precipitation statistics than GCMs. However, they are much more computationally intensive, time-consuming and expensive to run.

In a previous paper, the researchers developed an algorithm that detects the occurrence of heavy precipitation events based on climate models’ well-resolved, large-scale atmospheric circulation conditions associated with those events—rather than relying on these models’ simulated precipitation. The algorithm’s results corresponded with observations with much greater precision than the model-simulated precipitation.

In this paper, the researchers show that using output from RCMs rather than GCMs for the new algorithm does not improve the precision of simulated extreme precipitation frequency. The algorithm thus presents a robust and economic way to assess extreme precipitation frequency across a broad range of GCMs and multiple climate change scenarios with minimal computational requirements.   

Assessments of climate change impacts on agriculture are increasingly relying on panel models to examine the relationship between agricultural outcomes and weather fluctuations. This article reviews the strengths and weaknesses of such models. We argue that panel models are ideal for assessing climate impacts on agriculture because they use group fixed effects to absorb all time-invariant variation and thus rely on weather deviations from the mean that are random and exogenous. Using this random and exogenous source of variation is crucial to identifying a causal relationship between agricultural outcomes and weather. In addition, the large number of observations offered by a panel data set allows the identification of a nonlinear response function, which is an important step in modeling the effects of climate change, as the response can be highly nonlinear. Despite these strengths of panel models, they may still suffer from omitted variable biases of time-varying variables, such as pollution shocks, which are correlated with the weather shocks. Moreover, because group fixed effects absorb a lot of the signal in the weather variables, the signal:noise ratio might decrease. Thus researchers should be careful when constructing the weather variables in order to avoid having noise in the data that causes downward biases in the coefficients.