A multi-model ensemble of empirical and process-based models improves the predictive skill of near-term lake forecasts

Abstract

Water temperature forecasting in lakes and reservoirs is a valuable tool to manage crucial freshwater resources in a changing and more variable climate, but previous efforts have yet to identify an optimal modelling approach. Here, we demonstrate the first multi-model ensemble (MME) reservoir water temperature forecast, a forecasting method that combines individual model strengths in a single forecasting framework. We developed two MMEs: a three-model process-based MME and a five-model MME that includes process-based and empirical models to forecast water temperature profiles at a temperate drinking water reservoir. Our results showed that the five-model MME improved forecast performance by 8-30% relative to individual models and the process-based MME, as quantified using an aggregated probabilistic skill score. This increase in performance was due to large improvements in forecast bias in the five-model MME, despite increases in forecast uncertainty. High correlation among the process-based models resulted in little improvement in forecast performance in the process-based MME relative to the individual process-based models. The utility of MMEs is highlighted by two results: 1) no individual model performed best at every depth and horizon (days in the future), and 2) MMEs avoided poor performances by rarely producing the worst forecast for any single forecasted period (<6% of the worst ranked forecasts over time). This work presents an example of how existing models can be combined to improve water temperature forecasting in lakes and reservoirs and discusses the value of utilising MMEs, rather than individual models, in operational forecasts.

A Multi‐Model Ensemble of Baseline and Process‐Based Models Improves the Predictive Skill of Near‐Term Lake Forecasts

“…Instructions on reproducing the individual model forecasts as well as the MME are available in Olsson et al. (2023b). In addition, the forecasts and scores can be accessed here reproduce the manuscript figures (Olsson et al., 2023a).…”

“…(2023b). In addition, the forecasts and scores can be accessed here reproduce the manuscript figures (Olsson et al., 2023a).…”

“…FCR is monitored by a series of high‐frequency sensors deployed at fixed depths in the water column at its deepest site near the dam. Water temperature data were collected using T‐Node FR thermistors (NexSens, Fairborn, OH, USA) from March 2019 to March 2023 (Carey et al., 2023; Olsson et al., 2023a), with minor data gaps due to sensor maintenance (see metadata in Carey et al., 2023), across 10 depths in the water column (0.1, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0 m). Additional temperature data were collected at 1.6 m with a YSI EXO2 sonde (Xylem Inc., Yellow Springs, OH, USA), and at 5.0 and 9.0 m using RDO PRO‐X Dissolved Oxygen Sensors (In‐Situ Inc., Fort Collins, CO, USA).…”

“…All data and code are archived and available in the Zenodo repository (Olsson et al, 2023a(Olsson et al, , 2023b or the Environmental Data Initiative repository . Instructions on reproducing the individual model forecasts as well as the MME are available in Olsson et al (2023b). In addition, the forecasts and scores can be accessed here reproduce the manuscript figures (Olsson et al, 2023a).…”

“…All data and code to produce the forecasts, scores, and figures presented in this manuscript are available in the Zenodo repositories (Olsson et al, 2023a, 2023b) or the Environmental Data Initiative repositories (Carey & Breef‐Pilz, 2023; Carey et al., 2023).…”