Forecasting extreme events: making sense of noisy climate data in support of water resources planning

Abstract

Global climate model (GCM) projections are generally considered the best source of information for predicting future climate and hydrologic conditions in the face of a changing climate. Understanding and interpreting GCM projections is therefore critical for water resources planning. Unfortunately, this can be a challenging task as climate model data, particularly precipitation data, are notoriously noisy with large scatter and lacking in apparent patterns or trends. There is also usually large projection variability between models and model scenarios. This paper demonstrates a simple, practical method for synthesizing climate model data into more informative metrics using case studies of Atlanta, Georgia and Austin, Texas. Monthly and daily GCM projections, as well as historical observations, were translated into commonly used summary metrics for extreme event planning: peak 24-hour storm events and the Palmer Drought Severity Index (PDSI). Statistical trend analyses on these two metrics were used as a simple means to better understand the data. As expected, results identified significant, increasing, trends in projected 21st century temperatures for most GCM projections. Less expectedly, significant trends were also identified for projected future monthly and 24-hour maximum precipitation and drought severity. Implications of this work for water resources planning are discussed.

Review of the limitations and potential empirical improvements of the parametric group method of data handling for rainfall modelling

“…The modelling results illustrated a noticeable disparity between the accuracy of the temperature modelling when compared with the rainfall modelling using factors such as coefficients of determination. Cox et al ( 2019 ) also reported a similar outcome and detailed it as intrinsically noisy, featuring a large scatter with no obvious trend. However, due to the capabilities offered by such methods (e.g., Aghelpour and Varshavian 2020 ; Nguyen et al 2019 ), the authors committed to expanding their approach and investigating the potential for improvement of GMDH.…”