Sign of observed California temperature trends depends on data set homogenization: implications for weighting and downscaling
Abstract
Because downscaling methods can yield substantially different projections of future climate, it is imperative to constrain these projections with information from existing observations, while also recognizing observational uncertainty. California is a natural test case to develop observational constraints on future projections given prior studies that have purportedly found contrasting spatial patterns in late-20th-century trends of summertime daily-maximum temperature: cooling along the coast and warming inland. Revisiting this claim, we find that coastal cooling is largely confined to non-homogenized temperature records while homogenized observations show either non-significant cooling or warming trends throughout the state. This finding has implications for weighting LOCA-CMIP5 historical and future climate simulations. Failure to consider out-of-sample skill results in weighted and unweighted RCP 8.5 temperature trend estimates differing by 2 K/century or more in California. However, weighted mean estimates that properly account for trend uncertainty do not differ significantly from the unweighted mean.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.A53B..06C