A Bayesian modelling method for post-processing daily sub-seasonal to seasonal rainfall forecasts from global climate models and evaluation for 12 Australian catchments

Abstract

Abstract. Rainfall forecasts are an integral part of hydrological forecasting systems at sub-seasonal to seasonal timescales. In seasonal forecasting, global climate models (GCMs) are now the go-to source for rainfall forecasts. For hydrological applications however, GCM forecasts are often biased and unreliable in uncertainty spread, and calibration is therefore required before use. There are sophisticated statistical techniques for calibrating monthly and seasonal aggregations of the forecasts. However, calibration of seasonal forecasts at the daily time step typically uses very simple statistical methods or climate analogue methods. These methods generally lack the sophistication to achieve unbiased, reliable and coherent forecasts of daily amounts and seasonal accumulated totals. In this study, we propose and evaluate a Rainfall Post-Processing method for Seasonal forecasts (RPP-S), which is based on the Bayesian joint probability modelling approach for calibrating daily forecasts and the Schaake Shuffle for connecting the daily ensemble members of different lead times. We apply the method to post-process ACCESS-S forecasts for 12 perennial and ephemeral catchments across Australia and for 12 initialisation dates. RPP-S significantly reduces bias in raw forecasts and improves both skill and reliability. RPP-S forecasts are also more skilful and reliable than forecasts derived from ACCESS-S forecasts that have been post-processed using quantile mapping, especially for monthly and seasonal accumulations. Several opportunities to improve the robustness and skill of RPP-S are identified. The new RPP-S post-processed forecasts will be used in ensemble sub-seasonal to seasonal streamflow applications.

Enhancing the Predictability of Seasonal Streamflow With a Statistical‐Dynamical Approach

“…Model inputs can be improved by using the entire GCM forecast ensemble instead of simply the mean value and applying forecast ensembling approaches such as Bayesian updating (Bradley et al, 2015), optimal weights (Wanders & Wood, 2016), Bayesian joint probability (Schepen et al, 2018), or forecast monetary value . Model inputs can be improved by using the entire GCM forecast ensemble instead of simply the mean value and applying forecast ensembling approaches such as Bayesian updating (Bradley et al, 2015), optimal weights (Wanders & Wood, 2016), Bayesian joint probability (Schepen et al, 2018), or forecast monetary value .…”

“…In future work, statistical-dynamical seasonal flow forecasting systems can be further enhanced by improving the quality of model inputs, outputs, and the model formulation. Model inputs can be improved by using the entire GCM forecast ensemble instead of simply the mean value and applying forecast ensembling approaches such as Bayesian updating (Bradley et al, 2015), optimal weights (Wanders & Wood, 2016), Bayesian joint probability (Schepen et al, 2018), or forecast monetary value . The types of model inputs can also be expanded to include information about the initial land surface conditions or large-scale climate precursors such as the North Atlantic Oscillation or El Niño-Southern Oscillation (Emerton et al, 2017;Yuan, Wood, & Ma, 2015).…”

Robust Model Predictive Control of Irrigation Systems With Active Uncertainty Learning and Data Analytics

“…By deliberately designing loss functions in RMPC, one can pursue goals of minimizing water usage by penalizing large values of {u t }, and/or maintaining soil moisture around a prespecified level according to other economic criteria. Meanwhile, the control horizon could be several days or several weeks depending on the forecast model [21], [22], [23]. When the forecast horizon does not match the decision horizon, the terminal loss function provides more flexibility to drive the system towards a preferable state beyond the horizon.…”

Lựa chọn thành phần dự báo tổ hợp cho hệ thống dự báo hạn mùa

“…Phương pháp BJP xây dựng mối quan hệ giữa số liệu dự báo của mô hình và số liệu quan trắc dựa trên phân bố xác suất kết hợp. Phân bố kết hợp được mô hình hóa bằng một hàm phân bố chuẩn song biến (bivariate normal distribution) [10,12,15]. Sau đó mối quan hệ này sẽ được áp dụng cho các kết quả dự báo từ mô hình trong tương lai.…”