Efficacy of tendency and linear inverse models to predict southern Peru's rainy season precipitation

Abstract

Southern Peru receives over 60% of its annual climatological precipitation during the short period of January–March. This rainy season precipitation exhibits strong inter‐annual and decadal variability, including severe drought events that incur devastating societal impacts and cause agricultural communities and mining facilities to compete for limited water resources. Improving existing seasonal prediction models of summertime precipitation could aid in water resource planning and allocation across this water‐limited region. While various underlying mechanisms modulating inter‐annual variability have been proposed by past studies, operational forecasts continue to be largely based on rudimentary El Niño‐Southern Oscillation (ENSO)‐based indices, such as Niño3.4, justifying further exploration of predictive skill. To bridge the gap between understanding precipitation mechanisms and operational forecasts, we perform systematic studies on the predictability and prediction skill of southern Peru's rainy season precipitation by constructing statistical forecast models using best available weather station and reanalysis data sets. We construct a simple regression model, based on the principal component (PC) tendency of tropical Pacific sea surface temperatures (SST), and a more advanced linear inverse model (LIM), based on the empirical orthogonal functions of tropical Pacific SST and large‐scale atmospheric variables from reanalysis. Our results indicate that both the PC tendency and LIM models consistently outperform the ENSO‐only based regression models in predicting precipitation at both the regional scale and for individual station, with improvements for individual stations ranging from 10 to over 200%. These encouraging results are likely to foster further development of operational precipitation forecasts for southern Peru.

Bibliographic Information