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

53Citations
Citations of this article
90Readers
Mendeley users who have this article in their library.

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.

Figures

  • Figure 1. Map of Australian climate zones overlaid by gauging locations plotted as red triangles and labelled with catchment ID. Details of the catchments, including size, are presented in Table 1.
  • Table 1. Catchment ID, catchment name (river and gauging location), gauge ID, region and catchment size.
  • Figure 2. Example RPP-S rainfall forecast for the Burdekin River at Sellheim, initialised on the 1 October 2001 and forecasting 100 days ahead. Forecasts of daily amounts are shown in (a) and forecasts of accumulated totals are shown in (b). Dark blue is the forecast [0.25, 0.75] quantile range; medium blue is the forecast [0.10, 0.90] quantile range; and light blue is the forecast [0.05, 0.95] quantile range. Grey lines are the climatological reference forecast [0.05, 0.95] quantile range. The black line is the climatological reference forecast median. The red line is the observation.
  • Figure 3. Bias in daily rainfall forecasts for raw, QM and RPP forecasts (rows) and selected days ahead/lead times (columns). The scatterplots are of forecast mean over all events (x axis) versus bias (y axis). There is one blue circle for each catchment and forecast initialisation time. The average absolute bias (AB) is printed in the top left corner.
  • Figure 4. As for Fig. 3, except for accumulated totals.
  • Figure 5. α-index of reliability for forecast daily rainfall amounts (a) and forecast accumulated rainfall totals (b). Results for four types of forecasts are presented: raw, QM, RPP-S before Schaake Shuffle (pre SS) and RPP-S forecasts after the Schaake Shuffle. Higher α-index indicates better reliability. The box plots display the median as a black line. The box spans the interquartile range and the whiskers span the [0.1, 0.9] quantile range.
  • Figure 6. Scatterplots of CRPS skill scores for daily amounts (a) and accumulated totals (b). Results for QM are on the horizontal axis and results for RPP-S are on the vertical axis. Higher CRPS skills scores reflect better forecast performance. Red text preceded by a “+” symbol indicates the number of points plotted outside the axis limits in the quadrant nearest the text.
  • Figure 7. Scatterplots of CRPS skill scores for accumulated totals for each catchment. The scatterplots plot QM skill scores (horizontal axis) against RPP-S skill scores (vertical axis). Each scatterplot combines results for all initialisation dates and all lead times.

Register to see more suggestions

Mendeley helps you to discover research relevant for your work.

Already have an account?

Cite

CITATION STYLE

APA

Schepen, A., Zhao, T., Wang, Q. J., & Robertson, D. E. (2018). A Bayesian modelling method for post-processing daily sub-seasonal to seasonal rainfall forecasts from global climate models and evaluation for 12 Australian catchments. Hydrology and Earth System Sciences, 22(2), 1615–1628. https://doi.org/10.5194/hess-22-1615-2018

Readers over time

‘17‘18‘19‘20‘21‘22‘23‘24‘2507142128

Readers' Seniority

Tooltip

PhD / Post grad / Masters / Doc 33

56%

Researcher 18

31%

Professor / Associate Prof. 5

8%

Lecturer / Post doc 3

5%

Readers' Discipline

Tooltip

Earth and Planetary Sciences 26

48%

Engineering 20

37%

Environmental Science 6

11%

Computer Science 2

4%

Save time finding and organizing research with Mendeley

Sign up for free
0