The main objective in this project was to determine the homogeneity of extreme rainfall trends within commonly used regions to inform climate change at the national scale. In other words, how well is the regional trend representing the individual station trends that are aggregated within the region?

There exist different ways to partition the continental U.S. based on historical climate anomalies, climatatologic similarity, etc. One most commonly used to report trends at the national scale is the NCDC Climate Regions, sometimes adapted into slightly different regions (for example, those used in the National Climate Assessment).

Because of changing climate conditions, these regions might no longer be appropriate to study trends at the regional scale. We used another set of climatologically consistent regions named Bukovsky regions to assess the differences in the estimated regional trend when using a different set of regions. To look at whether the trends within the region are spatially homogeneous, we used the standard deviation of the aggregated stations trends.

  • Commonly used regions to aggregate rainfall extreme trends exhibit increasing trends in spatial variability suggesting that parts within the region are changing at different paces
  • Using a different set of regions leads to differences in the aggregated trend, and this uncertainty should be communicated
  • There exist potential for defining a new set of regions to study specifically historical and future changes in rainfall extremes through methods such as regional frequency analysis, or self-organizing maps.

Python scripts used in this project are hosted here.

The station records used in this study are part of the Global Historical Climate Network, and the data can be found here.

We are grateful to the National Centers for Environmenatal Information, Climate Data Online, for making the data publicly accessible.