Ensemble Seasonal Forecasts Research Articles

On the use of hindcast skill for merging NMME seasonal forecasts across the western U.S.

Abstract Multi-model ensemble forecasts have gained widespread use over the past decade. A yet unresolved issue is whether forecast skill benefits from the use of prior skill from each model in providing a weighted combination. Here we use the available seasonal ensemble forecasts of six models from the North American Multi-Model Ensemble (NMME) to study various aspects of prior skill-based weighting schemes and explore ways to merge multi-model forecasts. First, we post-process each NMME model through quantile mapping and a simple spread error adjustment. Then, using an equal weighted combination as the baseline forecast, we test merging the models together through skill-based weights by varying the prior skill metric and varying how the metrics are aggregated across the different subbasins and time of year. Results confirm prior work that the combined forecasts do outperform individual models. When evaluating prior skill, equal weighting generally performed as well or slightly better than all weighting schemes tried. The skill of the weighting scheme was not found to be strongly dependent on prior metric but did improve when aggregating all forecasted months and subbasins together to provide one overall weight to each model. Also, we found that including an offset to the prior metric that nudged the weights closer to equal weighting improves skill especially at longer leads where individual model skill is low. Results also show that the weighting schemes performed better than regression-based techniques including multiple linear regression and random forest.

Verification of Seasonal Ensemble Forecasts Based on the INM-CM5 Earth System Model

Verification of Seasonal Ensemble Forecasts Based on the INM-CM5 Earth System Model

Probabilistic seasonal precipitation forecasts using quantiles of ensemble forecasts

Seasonal precipitation forecasting is vital for weather-sensitive sectors. Global Circulation Models (GCM) routinely produce ensemble Seasonal Climate Forecasts (SCFs) but suffer from issues like low forecast resolution and skills. To address these issues in this study, we introduce a post-processing method, Quantile Ensemble Bayesian Model Averaging (QEBMA). It utilises quantiles from a GCM ensemble forecast to create a pseudo-ensemble forecast. Through their reasonable linear relationships with observations, each pseudo-member connects a hurdle distribution with a point mass at zero for dry months and a gamma distribution for wet months. These distributions are mixed to construct a forecast probability distribution with their weights, proportional to the quantiles’ historical forecast performance. QEBMA is applied to three GCMs, including GloSea5 from the United Kingdom, ECMWF from Europe and ACCESS-S1 from Australia, for monthly precipitation forecasts in 32 locations across four climate zones in Australia. Leave-one-month-out cross-validation results illustrate that QEBMA enhances forecast skills compared to raw GCMs and other post-processing techniques, including quantile mapping and Extended Copula Post-Processing (ECPP), for forecast lead time of 0 to 2 months, based on five metrics. The skill improvements achieved by QEBMA are often statistically significant, particularly when compared to raw GCM forecasts across the 32 study locations. Among these post-processing models, only QEBMA consistently outperforms the SCF benchmark climatology, offering a promising alternative for improving seasonal precipitation forecasts.

Predictive skill of North American Multi‐Model Ensemble seasonal forecasts for the climate rainfall over Central Africa

AbstractThis study evaluates the predictive performance of the North American Multi‐model Ensemble (NMME) over Central Africa (CA) using the historical rainfall data. The African Rainfall Climatology Version 2 (ARC2) is used as a substitute for reference observational data to examine the capability of 11 NMME and their NMME ensemble mean (MME) in simulating rainfall. Using the Kling‐Gupta efficiency (KGE), Taylor skill score (TSS), and Heidke skill score, the predictive evaluation of the models is performed from lead 0 to lead 5 of each season. The results show that the NMME models satisfactorily reproduce the bimodal and unimodal structure of rainfall in CA at the lead 0 level of different seasons: December–February (DJF), March–May (MAM), June–August (JJA), and September–November (SON). The pattern correlation coefficient (PCC) shows values of NMME and MME greater than ~0.69 and TSS > 0.60 for all four seasons. The MME presents a maximum in DJF between 0 and 1 month lead time. With the same time scale, just over of the NMME have a KGE between 0 and 0.42. It follows that as the forecast lead time increases, the PCC and TSS of each model become small, with PCC in JJA and DJF, TSS < 0.21 in JJA at lead 5. The NMME models exhibit an important rainfall bias and the calculated scores show the quality of the forecast decreases with increasing lead time; this may justify a constraint on the models to keep the good quality of the long‐term seasonal forecast in CA.

Framework for implementation of the Pitman-WR2012 model in seasonal hydrological forecasting: a case study of Kraai River, South Africa

Hydrological forecasting becomes an important tool in water resources management in forecasting the future state of the water resources in a catchment. The need for a reliable seasonal hydrologic forecast is significant and is becoming even more urgent under future climate conditions, as the assimilation of seasonal forecast information in decision making becomes part of the short and long-term climate change adaptation strategies in a range of contexts, such as energy supply, water supply and management, rural-urban, agriculture, infrastructure and disaster preparedness and relief. This work deals with the framework for implementation of the Pitman-WR2012 model in a hydrological forecasting mode. The Pitman-WR2012 model was forced with 10-member ensemble seasonal climate forecast from Climate Forecast Systems v.2 (CFSv2), which is downscaled using the principal components regression (PCR) approach. The generated seasonal hydrological forecast focused on the summer season, in particular on the Dec–Jan–Feb (DJF) period, which is the rainy season in the studied catchment (Kraai River catchment in the Eastern Cape Province of South Africa). The hydrological forecast issued at the end of November showed skill in December and February (assessed through Receiver Operating Characteristic (ROC) and Ranked Probability Skill Score (RPSS)), with poorer skill in January. Importantly, the skill of streamflow forecast was better than that of rainfall forecast, which likely results from the influence of the initial conditions of the hydrological model. In conclusion Pitman-WR2012 model performed realistically when implemented in seasonal hydrological forecasts mode, and it is important that in that mode the model is run with near-real-time rainfall data in order to maximize forecast skill arising from initial conditions.

Enhancing Ensemble Seasonal Streamflow Forecasts in the Upper Colorado River Basin Using Multi‐Model Climate Forecasts

AbstractIn the Colorado River Basin (CRB), ensemble streamflow prediction (ESP) forecasts drive operational planning models that project future reservoir system conditions. CRB operational seasonal streamflow forecasts are produced using ESP, which represents climate using an ensemble of meteorological sequences of historical temperature and precipitation, but do not typically leverage additional real‐time subseasonal‐to‐seasonal climate forecasts. Any improvements to streamflow forecasts would help stakeholders who depend on operational projections for decision making. We explore incorporating climate forecasts into ESP through variations on an ESP trace weighting approach, focusing on Colorado River unregulated inflows forecasts to Lake Powell. The k‐nearest neighbors (kNN) technique is employed using North American Multi‐Model Ensemble one‐ and three‐month temperature and precipitation forecasts, and preceding three‐month historical streamflow, as weighting factors. The benefit of disaggregated climate forecast information is assessed through the comparison of two kNN weighting strategies; a basin‐wide kNN uses the same ESP weights over the entire basin, and a disaggregated‐basin kNN applies ESP weights separately to four subbasins. We find in general that climate‐informed forecasts add greater marginal skill in late winter and early spring, and that more spatially granular disaggregated‐basin use of climate forecasts slightly improves skill over the basin‐wide method at most lead times.

Seasonal ensemble forecasts for soil moisture, evapotranspiration and runoff across Australia

The seasonal predictability of hydrological variables at high resolution across Australia was assessed using a new national seasonal ensemble forecasting system for soil moisture, evapotranspiration and runoff. The system consists of a gridded water balance model (AWRA-L) forced with downscaled and calibrated seasonal climate forecasts from a climate modelling system, ACCESS-S1. Here, we evaluate the hydrological forecasts at one- to six-months lead time relative to a historical reference simulation forced with observed climate inputs. Deterministic and probabilistic skill, accuracy and reliability of the forecast ensemble were assessed, with a specific focus on forecasts of hydrological extremes. Additionally, we assess the performance of the hindcast for selected use cases, particularly focusing on agriculture and water management.Nationally, our assessment of the hydrological forecasts indicated good skill for all variables at one-month lead time, and up to three-months lead time for soil moisture and actual evapotranspiration in many key regions in Australia. Regarding hydrological extremes, we find particularly high skill at longer lead times for low soil moisture and actual evapotranspiration events, highlighting the potential value for applications such as hydrological drought prediction or bushfire risk assessments. Bias of the hydrological forecasts based on calibrated seasonal climate forecasts was shown to be close to zero for all variables, regions and lead times and insignificant, with a minor exception for potential evapotranspiration.More detailed evaluation across economically important regions and river basins, in particular, the Australian Wheatbelt and the Murray-Darling Basin (the largest river basin in Australia and an economically and environmentally critical region), highlights promising potential for the soil moisture and evapotranspiration forecast to be used in applications in agriculture, such as for optimising irrigation or scheduling sowing and harvesting dates. For hydrologically important headwater regions across the Great Dividing Range and the alpine regions of Victoria, the forecasts of runoff show good skill at longer lead times indicating the value of the forecasts for reservoir management.Overall, we conclude that the forecasting system shows sufficient skill for a wide range of applications and regions up to a lead time of three months. We outline limitations of the presented system and highlight potential future research directions.

Downscaling of seasonal ensemble forecasts to the convection‐permitting scale over the Horn of Africa using the WRF model

AbstractThe new SEAS5 global ensemble forecast system was dynamically downscaled over the Horn of Africa for summer (June‐July‐August) 2018. For this purpose, a multi‐physics ensemble was designed with a grid increment of 3 km and without any intermediate nest based on the Weather Research and Forecasting model (WRF). The WRF and the SEAS5 model output were compared with each other and reference datasets to assess the biases in 4 different regions of Ethiopia. Also, the WRF ensemble variability was investigated in relation to model parameterization and lateral boundary conditions. Over the summer, the SEAS5 has a positive temperature bias of 0.17°C compared to ECMWF analysis average for the study domain, while the WRF bias is +1.14°C. Concerning precipitation, the WRF model had average accumulated values of 264 mm, compared to 248 mm for SEAS5 and 236 mm for the observations. Over south Ethiopia, however, the downscaling produced over 50% more precipitation than the other datasets. The maximum northward extension of the tropical rain belt was reduced by about 2° in both models when compared to observations. Downscaling increased reliability for precipitation, correcting the SEAS5 underdispersion: ensemble spread for precipitation was increased by about 70% in the WRF ensemble in three of the four Ethiopian sub‐regions, whereas the very dry Somali region remained unaffected. The WRF ensemble analysis revealed that the ensemble spread is mainly caused by the perturbed boundary conditions, as their effect is often 50% larger than the physics‐induced variability in the mountainous part of Ethiopia for precipitation and temperature.

The Canadian Seasonal to Interannual Prediction System Version 2 (CanSIPSv2)

AbstractThe second version of the Canadian Seasonal to Interannual Prediction System (CanSIPSv2) was implemented operationally at Environment and Climate Change Canada (ECCC) in July 2019. Like its predecessors, CanSIPSv2 applies a multimodel ensemble approach with two coupled atmosphere–ocean models, CanCM4i and GEM-NEMO. While CanCM4i is a climate model, which is upgraded from CanCM4 of the previous CanSIPSv1 with improved sea ice initialization, GEM-NEMO is a newly developed numerical weather prediction (NWP)-based global atmosphere–ocean coupled model. In this paper, CanSIPSv2 is introduced, and its performance is assessed based on the reforecast of 30 years from 1981 to 2010, with 10 ensemble members of 12-month integrations for each model. Ensemble seasonal forecast skill of 2-m air temperature, 500-hPa geopotential height, precipitation rate, sea surface temperature, and sea ice concentration is assessed. Verification is also performed for the Niño-3.4, the Pacific–North American pattern (PNA), the North Atlantic Oscillation (NAO), and the Madden–Julian oscillation (MJO) indices. It is found that CanSIPSv2 outperforms the previous CanSIPSv1 system in many aspects. Atmospheric teleconnections associated with the El Niño–Southern Oscillation (ENSO) are reasonably well captured by the two CanSIPSv2 models, and a large part of the seasonal forecast skill in boreal winter can be attributed to the ENSO impact. The two models are also able to simulate the Northern Hemisphere teleconnection associated with the tropical MJO, which likely provides another source of skill on the subseasonal to seasonal time scale.

Interannual variation of the East Asia Jet Stream and its impact on the horizontal distribution of aerosol in boreal spring

Interannual variation of the aerosol optical depth (AOD) in East Asia has been investigated using Moderate Resolution Imaging Spectroradiometer (MODIS) data and Modern Era Retrospective analysis for Research and Applications Version 2 (MERRA-2) data for 2000–2018. The data analysis focuses on boreal spring when Siberian biomass burning is at its seasonal maximum. The results indicate that the significant increase in organic and black carbon is primarily caused by emissions from biomass burning in East Asia, which leads to significant interannual variations in aerosol loading and pan-Pacific transport. The anomalous large-scale climate variability associated with the East Asia Jet Stream (EAJS) provides favorable conditions for increasing the AOD of organic and black carbon in Northeast Asia and may represent an underlying physical mechanism. When the EAJS shows greater weakening than normal, abnormal high-pressure anomalies are maintained in East Asia, which tend to drive warm advection over Northeast Asia. This warm advection expedites the melting of the Eurasian snow cover, which helps increase surface dryness in late spring and provides favorable conditions for biomass burning. The EAJS index can be predictable with statistical significance up to lead 1 month by the dynamical ensemble seasonal forecasts, suggesting a possible implementation of the empirical AOD forecasts using climate forecast models.

Seasonal Hydropower Planning for Data‐Scarce Regions Using Multimodel Ensemble Forecasts, Remote Sensing Data, and Stochastic Programming

AbstractIn data‐scarce regions, seasonal hydropower planning is hindered by the unavailability of reliable long‐term streamflow observations, which are required for the construction of inflow scenario trees. In this study, we develop a methodological framework to overcome the problem of streamflow data scarcity by combining precipitation forecasts from ensemble numerical weather prediction models, spatially distributed hydrologic models, and stochastic programming. We use evapotranspiration as a proxy for streamflow in generating reliable reservoir inflow forecasts. Using the framework, we compare three different formulations of inflow scenario structures and their applicability to data‐scarce regions: (1) a single deterministic forecast, (2) a scenario fan with the first stage deterministic, and (3) a scenario fan with all stages stochastic. We apply the framework to a cascade of two reservoirs in the Omo‐Gibe River basin in Ethiopia. Future reservoir inflows are generated using a 3‐model 30‐member ensemble seasonal precipitation forecast from the North American Multimodel Ensemble and the Noah‐MP hydrologic model. We then perform deterministic and stochastic optimization for hydropower operation and planning. Comparing the results from the three different inflow scenario structures, we observe that the uncertainty in reservoir inflows is significant only for the dry stages of the planning horizon. In addition, we find that the impact of model parameter uncertainty on hydropower production is significant (0.14–0.18×106 MWh).

On the skill of raw and post-processed ensemble seasonal meteorological forecasts in Denmark

Abstract. This study analyzes the quality of the raw and post-processed seasonal forecasts of the European Centre for Medium-Range Weather Forecasts (ECMWF) System 4. The focus is given to Denmark, located in a region where seasonal forecasting is of special difficulty. The extent to which there are improvements after post-processing is investigated. We make use of two techniques, namely linear scaling or delta change (LS) and quantile mapping (QM), to daily bias correct seasonal ensemble predictions of hydrologically relevant variables such as precipitation, temperature and reference evapotranspiration (ET0). Qualities of importance in this study are the reduction of bias and the improvement in accuracy and sharpness over ensemble climatology. Statistical consistency and its improvement is also examined. Raw forecasts exhibit biases in the mean that have a spatiotemporal variability more pronounced for precipitation and temperature. This variability is more stable for ET0 with a consistent positive bias. Accuracy is higher than ensemble climatology for some months at the first month lead time only and, in general, ECMWF System 4 forecasts tend to be sharper. ET0 also exhibits an underdispersion issue, i.e., forecasts are narrower than their true uncertainty level. After correction, reductions in the mean are seen. This, however, is not enough to ensure an overall higher level of skill in terms of accuracy, although modest improvements are seen for temperature and ET0, mainly at the first month lead time. QM is better suited to improve statistical consistency of forecasts that exhibit dispersion issues, i.e., when forecasts are consistently overconfident. Furthermore, it also enhances the accuracy of the monthly number of dry days to a higher extent than LS. Caution is advised when applying a multiplicative factor to bias correct variables such as precipitation. It may overestimate the ability that LS has in improving sharpness when a positive bias in the mean exists.

Using daily data from seasonal forecasts in dynamic crop models for yield prediction: A case study for rice in Nepal’s Terai

Skillful seasonal climate predictions paired with a dynamic crop model can assist agricultural management and help farmers minimize risk. For crop yield predictions, the skill in generating realistic distributions of weather for the crop growing season matters more than the skill of forecasting the mean seasonal climate itself. In this regard, the ensemble of daily fields of the Seasonal Prediction Systems (SPSs) output could be a potential alternative to other methods that are available to generate daily weather from the monthly or seasonal mean forecasts. However, the SPSs are not expected to forecast individual weather events at a given grid point (deterministic forecast), but if the statistics of the predicted weather are correct, an ensemble of yield predictions using individual realizations of the ensemble seasonal forecast may produce a more skillful yield forecast. So far, the potential of this new approach has not been tested. The goal of this study was to determine the potential applicability of using daily data from SPSs to predict rice yield through a case study of Nepal’s Terai. The study used 28 years (1983–2010) daily hindcasts of the coupled forecast system model version 2 (CFSv2) SPS into a Cropping System Model (CSM)-CERES-Rice. The hindcasts of the CFSv2, initialized at different lead times, were used in various ways to simulate rice yield, which were then compared to the reference yield and to the simulated yield using climatology alone to examine the predictive skill at different lead times. The results from this study indicate that unlike the typical ensemble averaging approach commonly used in seasonal climate forecasting, averaging the simulated yield using individual member does not guarantee better yield prediction. Further analyses should be made, including alternative downscaling methods as well as a similar analysis for an area where quality meteorological and agricultural data are available and where the seasonal forecasts exhibit better skill.

Prediction Skill for the East Asian Winter Monsoon Based on APCC Multi-Models

The prediction skill for the East Asian winter monsoon (EAWM) has been analyzed, using the observations and different climate models that participate in the APEC Climate Center (APCC) multi-model ensemble (MME) seasonal forecast. The authors first examined the characteristics of the existing EAWM indices to find a suitable index for the APCC seasonal forecast system. This examination revealed that the selected index shows reasonable prediction skill of EAWM intensity and well-represents the characteristics of wintertime temperature anomalies associated with the EAWM, especially for the extreme cold winters. Although most models capture the main characteristics of the seasonal mean circulation over East Asia reasonably well, they still suffer from difficulty in predicting the interannual variability (IAV) of the EAWM. Fortunately, the POAMA has reasonable skill in capturing the timing and strength of the EAWM IAV and reproduces the EAWM-related circulation anomalies well. The better performance of the POAMA may be attributed to the better skill in simulating the high-latitude forcing including the Siberian High (SH) and Artic Oscillation (AO) and the strong links of the ENSO to the EAWM, compared to other models.

Drought Prediction for Areas with Sparse Monitoring Networks: A Case Study for Fiji

Hybrid drought prediction models were developed for areas with limited monitoring gauges using the APEC Climate Center Multi-Model Ensemble seasonal climate forecast and machine learning models of Extra-Trees and Adaboost. The models provide spatially distributed detailed drought prediction data of the 6-month Standardized Precipitation Index for the case study area, Fiji. In order to overcome the limitation of a sparse monitoring network, both in-situ data and bias-corrected dynamic downscaling of historical climate data from the Weather Research Forecasting (WRF) model were used as reference data. Performance measures of the mean absolute error as well as classification accuracy were used. The WRF outputs reflect the topography of the area. Hybrid models showed better performance than simply bias corrected forecasts in most cases. Especially, the model based on Extra-Trees trained using the WRF model outputs performed the best in most cases.

Probabilistic maize yield prediction over East Africa using dynamic ensemble seasonal climate forecasts

We tested the usefulness of seasonal climate predictions for impacts prediction in eastern Africa. In regions where these seasonal predictions showed skill we tested if the skill also translated into maize yield forecasting skills. Using European Centre for Medium-Range Weather Forecasts (ECMWF) system-4 ensemble seasonal climate hindcasts for the period 1981–2010 at different initialization dates before sowing, we generated a 15-member ensemble of yield predictions using the World Food Studies (WOFOST) crop model implemented for water-limited maize production and single season simulation. Maize yield predictions are validated against reference yield simulations using the WATCH Forcing Data ERA-Interim (WFDEI), focussing on the dominant sowing dates in the northern region (July), equatorial region (March-April) and in the southern region (December). These reference yields show good anomaly correlations compared to the official FAO and national reported statistics, but the average reference yield values are lower than those reported in Kenya and Ethiopia, but slightly higher in Tanzania.We use the ensemble mean, interannual variability, mean errors, Ranked Probability Skill Score (RPSS) and Relative Operating Curve skill Score (ROCSS) to assess regions of useful probabilistic prediction. Annual yield anomalies are predictable 2-months before sowing in most of the regions. Difference in interannual variability between the reference and predicted yields range from ±40%, but higher interannual variability in predicted yield dominates. Anomaly correlations between the reference and predicted yields are largely positive and range from +0.3 to +0.6. The ROCSS illustrate good pre-season probabilistic prediction of above-normal and below-normal yields with at least 2-months lead time. From the sample sowing dates considered, we concluded that, there is potential to use dynamical seasonal climate forecasts with a process based crop simulation model WOFOST to predict anomalous water-limited maize yields.

Seasonal forecasting of green water components and crop yields of winter wheat in Serbia and Austria.

A probabilistic crop forecast based on ensembles of crop model output (CMO) estimates offers a myriad of possible realizations and probabilistic forecasts of green water components (precipitation and evapotranspiration), crop yields and green water footprints (GWFs) on monthly or seasonal scales. The present paper presents part of the results of an ongoing study related to the application of ensemble forecasting concepts for agricultural production. The methodology used to produce the ensemble CMO using the ensemble seasonal weather forecasts as the crop model input meteorological data without the perturbation of initial soil or crop conditions is presented and tested for accuracy, as are its results. The selected case study is for winter wheat growth in Austria and Serbia during the 2006-2014 period modelled with the SIRIUS crop model. The historical seasonal forecasts for a 6-month period (1 March-31 August) were collected for the period 2006-2014 and were assimilated from the European Centre for Medium-range Weather Forecast and the Meteorological Archival and Retrieval System. The seasonal ensemble forecasting results obtained for winter wheat phenology dynamics, yield and GWF showed a narrow range of estimates. These results indicate that the use of seasonal weather forecasting in agriculture and its applications for probabilistic crop forecasting can optimize field operations (e.g., soil cultivation, plant protection, fertilizing, irrigation) and takes advantage of the predictions of crop development and yield a few weeks or months in advance.

Complex relationship between seasonal streamflow forecast skill and value in reservoir operations

Abstract. Considerable research effort has recently been directed at improving and operationalising ensemble seasonal streamflow forecasts. Whilst this creates new opportunities for improving the performance of water resources systems, there may also be associated risks. Here, we explore these potential risks by examining the sensitivity of forecast value (improvement in system performance brought about by adopting forecasts) to changes in the forecast skill for a range of hypothetical reservoir designs with contrasting operating objectives. Forecast-informed operations are simulated using rolling horizon, adaptive control and then benchmarked against optimised control rules to assess performance improvements. Results show that there exists a strong relationship between forecast skill and value for systems operated to maintain a target water level. But this relationship breaks down when the reservoir is operated to satisfy a target demand for water; good forecast accuracy does not necessarily translate into performance improvement. We show that the primary cause of this behaviour is the buffering role played by storage in water supply reservoirs, which renders the forecast superfluous for long periods of the operation. System performance depends primarily on forecast accuracy when critical decisions are made – namely during severe drought. As it is not possible to know in advance if a forecast will perform well at such moments, we advocate measuring the consistency of forecast performance, through bootstrap resampling, to indicate potential usefulness in storage operations. Our results highlight the need for sensitivity assessment in value-of-forecast studies involving reservoirs with supply objectives.

Skill of ECMWF system‐4 ensemble seasonal climate forecasts for East Africa

ABSTRACTThis study evaluates the potential use of the ECMWF System‐4 seasonal forecasts (S4) for impact analysis over East Africa. For use, these forecasts should have skill and small biases. We used the 15‐member ensemble of 7‐month forecasts initiated every month, and tested forecast skill of precipitation (tp), near‐surface air temperature (tas) and surface downwelling shortwave radiation (rsds). We validated the 30‐year (1981–2010) hindcast version of S4 against the WFDEI reanalysis (WATCH Forcing Data ERA‐Interim) and to independent relevant observational data sets. Probabilistic skill is assessed using anomaly correlation, ranked probability skill score (RPSS) and the relative operating curve skill score (ROCSS) at both grid cell and over six distinct homogeneous rainfall regions for the three growing seasons of East Africa (i.e. MAM, JJA and OND).S4 exhibits a wet bias in OND, a dry bias in MAM and a mix of both in JJA. Temperature biases are similar in all seasons, constant with lead‐time and correlate with elevation. Biases in rsds correlate with cloud/rain patterns. Bias correction clears biases but does not affect probabilistic skills.Predictability of the three variables varies with season, location and lead‐time. The choice of validating dataset plays little role in the regional patterns and magnitudes of probabilistic skill scores. The OND tp forecasts show skill over a larger area up to 3 months lead‐time compared to MAM and JJA. Upper‐ and lower‐tercile tp forecasts are 20–80% better than climatology. Temperature forecasts are skillful for at least 3 months lead‐time and they are 40–100% better than climatology. The rsds is less skillful than tp and tas in all seasons when verified against WFDEI but higher in all lead months against the alternative datasets. The forecast system captures El‐Niño Southern Oscillation (ENSO)‐related anomalous years with region‐dependent skill.

Reliable long-range ensemble streamflow forecasts: Combining calibrated climate forecasts with a conceptual runoff model and a staged error model

We present a new streamflow forecasting system called forecast guided stochastic scenarios (FoGSS). FoGSS makes use of ensemble seasonal precipitation forecasts from a coupled ocean-atmosphere general circulation model (CGCM). The CGCM forecasts are post-processed with the method of calibration, bridging and merging (CBaM) to produce ensemble precipitation forecasts over river catchments. CBaM corrects biases and removes noise from the CGCM forecasts, and produces highly reliable ensemble precipitation forecasts. The post-processed CGCM forecasts are used to force the Wapaba monthly rainfall-runoff model. Uncertainty in the hydrological modelling is accounted for with a 3-stage error model. Stage 1 applies the log-sinh transformation to normalize residuals and homogenize their variance; Stage 2 applies a conditional bias-correction to correct biases and help remove negative forecast skill; Stage 3 applies an autoregressive model to improve forecast accuracy at short lead-times and propagate uncertainty through the forecast. FoGSS generates ensemble forecasts in the form of time series for the coming 12-months. In a case study of two catchments, FoGSS produces reliable forecasts at all lead-times. Forecast skill with respect to climatology is evident to lead-times of about 3 months. At longer lead-times, forecast skill approximates that of climatology forecasts; that is, forecasts become like stochastic scenarios. Because forecast skill is virtually never negative at long lead-times, forecasts of accumulated volumes can be skillful. Forecasts of accumulated 12-month streamflow volumes are significantly skillful in several instances, and ensembles of accumulated volumes are reliable. We conclude that FoGSS forecasts could be highly useful to water managers. This article is protected by copyright. All rights reserved.