North American Multimodel Ensemble Models Research Articles

Performance‐based evaluation of NMME and C3S models in forecasting the June–August Central African rainfall under the influence of the South Atlantic Ocean Dipole

AbstractIn this study, hindcasts from eight Copernicus Climate Change Service (C3S) and three North American Multi‐Model Ensemble (NMME) operational seasonal forecast systems, based on dynamical climate models, are employed to investigate the influence of the South Atlantic Ocean Dipole (SAOD) on the predictive skill of Central Africa (CA) rainfall. The focus is primarily on the June–July–August season for 1993–2016. The findings reveal that, when regionally averaged, all models exhibit positive skill in predicting CA rainfall, except for the Geophysical Fluid Dynamics Laboratory (GFDL‐SPEAR) model. Notably, there are significant spatial variations in skill across different regions. Model performance is particularly low (high) in the Central African Republic and Congo Basin (Gabon and Chad) and tends to deteriorate with increasing lead‐time. Models that demonstrate a strong connection between SAOD and CA rainfall tend to exhibit better predictive skills in forecasting rainfall, in contrast to models with weaker connections. This leads to a significant in‐phase relationship between the predictive skills of rainfall and the strength of the SAOD–rainfall connection among the models. Furthermore, the atmospheric circulation responding to SST forcing associated with the El Niño–Southern Oscillation exerts a significant influence on the robust atmospheric circulation associated with the climatological mean of SST over the SAO. This suggests that mean state bias in the SAO/equatorial Pacific region plays a role in modulating the strength of the simulated SAOD–CA rainfall connection and, consequently, the prediction skill of CA rainfall. In general, both NMME and C3S models appear to be valuable tools capable of providing essential seasonal information several months in advance. These insights can aid decision‐makers in the region in making informed decisions regarding adaptation and mitigation measures.

Reconciling roles of the South China Sea summer monsoon and ENSO in prediction of the Indian Ocean dipole

The Indian Ocean dipole (IOD) is a remarkable interannual variability in the tropical Indian Ocean. The improved prediction of IOD is of a great value because of its large socioeconomic impacts. Previous studies reported that both El Niño-Southern Oscillation (ENSO) and South China Sea summer monsoon (SM) play a dominant role in the western and eastern pole of the IOD, respectively. They can be used as predictors of the IOD at 3 month lead beyond self-persistence. Here, we develop an empirical model of multi-factors in which the western pole is predicted by ENSO and persistence and the eastern pole is predicted by SM and persistence. This new empirical model outperforms largely the average level of the dynamical models from the North American multi-model ensemble (NMME) project in predicting the peak IOD in boreal autumn, with a correlation coefficient of ∼0.86 and a root mean square error of ∼0.24 °C. Furthermore, the hit rate of positive culminated IOD in this new empirical model is equivalent to that in current NMME models (above 65%), much higher than that for negative culminated IOD. This improvement of skill using the empirical model suggests a perspective for better understanding and predicting the IOD.

Unraveling the Formation Mechanism of Exceptionally Strong Marine Heatwaves in the Northeast Pacific in 2020

Abstract Throughout 2020, a persistent marine heatwave (MHW) event occurred in the northeast (NE) Pacific, which exhibited a record-breaking intensity and long duration. Three peaks of sea surface temperature (SST) anomaly are identified in April, July, and November 2020, respectively, all of which are caused by the surface latent heat flux (LH) term. A Taylor expansion analysis for the LH term reveals that positive LH anomaly (LHA) played the decisive role in increasing the latent heating effect, whereas the mixed layer depth anomaly was not important in the 2020 MHW. The positive LHA was primarily caused by the southerly wind anomalies, which advected more humid air from lower latitudes to the NE Pacific and thus reduced the difference between the saturated specific humidity at sea surface and actual surface specific humidity. This reduced sea–air humidity difference was conducive to less evaporation, leading to positive LHA. Using an atmospheric general circulation model forced by observed SSTs, we find that although the atmospheric circulation anomalies associated with the 2020 MHW can be partly constrained by both tropical Pacific SST and local SST over the North Pacific, the role of internal atmospheric variability in the 2020 MHW’s formation cannot be overlooked. The prediction skill for the 2020 MHW in the North American Multimodel Ensemble models is limited to one month, indicating the challenge of accurately predicting MHW in the NE Pacific. The finding about the contribution of sea–air humidity difference to the LHA provides a new insight into the formation mechanism of MHW. Significance Statement An exceptionally strong and persistent marine heatwave occurred in the northeast Pacific throughout 2020, with three peaks in April, July, and November, causing considerable damage to local fisheries and ecosystems. This study found that the evolution of this marine heatwave was due to the less-than-normal release of latent heat at the ocean surface. This was mainly caused by the reduced sea–air humidity difference, which was attributed to the increased near-surface moisture transported by southerly wind anomalies. Our model experiments demonstrated that the associated anomaly fields in the atmosphere partly stemmed from stochastic processes. The state-of-the-art dynamic models can only predict this marine heatwave one month in advance, indicating the necessity of improving prediction skill for marine heatwave’s evolution in current models.

Forecast skill of Bangladesh summer monsoon rainfall in C3S and NMME models after calibration

This study assesses the dynamical seasonal predictions initialized in April and May for forecasting Bangladesh summer (June to September: JJAS) monsoon rainfall (BSMR) over the 1993–2016 period. The BSMR from nine models, sourced from the North American Multi-Model Ensemble (NMME) and the Copernicus Climate Change Service (C3S), undergoes a calibration process. This calibration employs the Canonical Correlation Analysis (CCA) technique to rectify biases within each individual model's ensemble mean BSMR data (referred as predictor or X variable). These corrections are made in comparison to observed BSMR (referred as predictand or Y variable), acquired from the Climate Hazards Group InfraRed Precipitation with Station data. Subsequently, the models that undergo calibration are amalgamated to construct a calibrated multi-model ensemble (CMME), which, in turn, facilitates the generation of a forecast for BSMR. The CCA correction brings about a significant improvement in root-mean-square error, underscoring the presence of correctable systematic biases in the raw model forecasts. However, these CCA corrections weakly enhance the skill (anomaly correlation) across the region. The scores that assess discrimination (the two-alternative forced-choice: 2AFC and the area under the relative operating characteristic curve: ROC for above/below normal BSMR) for tercile-based forecasts exceeded 50 % across a substantial portion of the region. This indicates a superior level of discrimination compared to what one would anticipate based on climatology. Using the CMME approach, a probabilistic forecast for the 2022 BSMR was generated and proved quite effective in capturing the observed 2022 BSMR tercile, which includes below-normal and above-normal categories of rainfall in the central-southern and northern regions of Bangladesh respectively. Furthermore, the absence of substantial skill improvements may be attributed to inaccuracies in the teleconnection patterns of the simulated first leading principal component (PC) time series of BSMR with the El Niño Southern Oscillation. In contrast, the second PC time series exhibits a similar connection to observations. These findings emphasize the importance and utility of statistical post-processing in producing reliable seasonal climate outlooks for the region.

Statistical refinement of the North American Multi-Model Ensemble precipitation forecasts over Karoon basin, Iran

Abstract An effective postprocessing approach has been examined to improve the skill of North American Multi-Model Ensemble (NMME) precipitation forecasts in the Karoon basin, Iran. The Copula–Bayesian approach was used along with the Normal Kernel Density marginal distribution and the Kernel Copula function. This process creates more than one postprocessing precipitation value as results candidates (first pass). A similar process is used for a second pass to obtain preprocessed values based on the candidate inputs, which helps identify the most suitable postprocessed value. The application of the technique for order preference by similarity to the ideal solution method based on conditional probability distribution functions of the first and second passes leads to achieving final improved forecast data among the existing candidates. To validate the results, data from 1982–2010 and 2011–2018 were used for the calibration and forecast periods. The results show that while the GFDL and CFS2 models tend to overestimate precipitation, most other NMME models underestimate it. Postprocessing improves the accuracy of forecasts for most models by 20%–40%. Overall, the proposed Copula–Bayesian postprocessing approach could provide more reliable forecasts with higher spatial and temporal consistency, better detection of extreme precipitation values, and a significant reduction in uncertainties.

Evolution of the NMME Rainfall Seasonal Forecasting over Central Africa

The North American Multi-Model Ensemble (NMME) has grown into a fully developed scientific database for seasonal and sub-seasonal climate forecasts, progressing prediction from global to regional scales. The NMME has continuously developed, with new models replacing old ones; it is hypothesized that this development will generate more accurate forecasts over time. However, to date, this hypothesis has not been verified in Central Africa (CA). This study investigates the hypothesis that the skill of NMME models will increase as the forecasting system advances, focusing on rainfall in CA. The study is conducted for the four configuration (phases) of NMME models, from the oldest to the most recent. The analyses are performed with Short Lead (SL) time and Long Lead (LL) time hindcasts very coherent with the perspectives of the CA. The results show from configuration 1 (phase 1) to configurations 4 (phase 4), the NMME models reasonably replicate the spatial structures in the seasonal rainfall climatology of the observations with a remarkable bias at LL. The mean absolute error and root mean square difference reveal small but incremental improvements in the prediction skills of NMME models from phase 1 to phase 4. The Pearson coefficient (r) increased in SL by about 1%, i.e., from 0.94 to 0.95 during June–August (JJA) season and about 4% during the September–November (SON), i.e., from [Formula: see text] in phase 1 to [Formula: see text] in phase 4, about 3% from phase 1 to phase 4 during the March–May (MAM). The categorical scores show that the Probability of Detection (POD) and False Alarm (FAR) increased very slightly from phase 1 to phase 4, but is it noted that the different combinations of the NMME forecasting system present difficulties in predicting rainy and dry events. It should be added that by introducing newer models into a multi-model ensemble as they are developed, and by eliminating older models, small skill gains are observed in the NMME forecasting system in CA.

Inter-annual variability of pre-monsoon surface air temperatures over India using the North American Multi-Model Ensemble models during the global warming era

Inter-annual variability of pre-monsoon surface air temperatures over India using the North American Multi-Model Ensemble models during the global warming era

Improved Monthly and Seasonal Multi-Model Ensemble Precipitation Forecasts in Southwest Asia Using Machine Learning Algorithms

Southwest Asia has different climate types including arid, semiarid, Mediterranean, and temperate regions. Due to the complex interactions among components of the Earth system, forecasting precipitation is a difficult task in such large regions. The aim of this paper is to propose a learning approach, based on artificial neural network (ANN) and random forest (RF) algorithms for post-processing the output of forecasting models, in order to provide a multi-model ensemble forecasting of monthly precipitation in southwest Asia. For this purpose, four forecasting models, including GEM-NEMO, NASA-GEOSS2S, CanCM4i, and COLA-RSMAS-CCSM4, included in the North American multi-model ensemble (NMME) project, are considered for the ensemble algorithms. Since each model has nine different lead times, a total of 108 different ANN and RF models are trained for each month of the year. To train the proposed ANN an RF models, the ERA5 reanalysis dataset is employed. To compare the performance of the proposed algorithms, four performance evaluation criteria are calculated for each model. The results indicate that the performance of the ANN and RF post-processing is better than that of the individual NMME models. Moreover, RF outperformed ANN for all lead times and months of the year.

Extended seasonal prediction of spring precipitation over the Upper Colorado River Basin

This study provides extended seasonal predictions for the Upper Colorado River Basin (UCRB) precipitation in boreal spring using an artificial neural network (ANN) model and a stepwise linear regression model, respectively. Sea surface temperature (SST) predictors are developed taking advantage of the correlation between the precipitation and SST over three ocean basins. The extratropical North Pacific has a higher correlation with the UCRB spring precipitation than the tropical Pacific and North Atlantic. For the ANN model, the Pearson correlation coefficient between the observed and predicted precipitation exceeds 0.45 (p-value < 0.01) for a lead time of 12 months. The mean absolute percentage error (MAPE) is below 20% and the Heidke skill score (HSS) is above 50%. Such long-lead prediction skill is probably due to the UCRB soil moisture bridging the SST and precipitation. The stepwise linear regression model shows similar prediction skills to those of ANN. Both models show prediction skills superior to those of an autoregression model (correlation < 0.10) that represents the baseline prediction skill and those of three of the North American Multi-Model Ensemble (NMME) forecast models. The three NMME models exhibit different skills in predicting the precipitation, with the best skills of the correlation ~ 0.40, MAPE < 25%, and HSS > 40% for lead times less than 8 months. This study highlights the advantage of oceanic climate signals in extended seasonal predictions for the UCRB spring precipitation and supports the improvement of the UCRB streamflow prediction and related water resource decisions.

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 &gt; 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 &lt; 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.

Skill of the Saudi-KAU CGCM in Forecasting ENSO and its Comparison with NMME and C3S Models

This paper assesses the skill of the Saudi-King Abdulaziz University coupled ocean–atmosphere Global Climate Model, namely Saudi-KAU CGCM, in forecasting the El Niño-Southern Oscillation (ENSO)-related sea surface temperature. The model performance is evaluated based on a reforecast of 38 years from 1982 to 2019, with 20 ensemble members of 12-month integrations. The analysis is executed on ensemble mean data separately for boreal winter (December to February: DJF), spring (March to May: MAM), summer (June to August: JJA), and autumn (September to November: SON) seasons. It is found that the Saudi-KAU model mimics the observed climatological pattern and variability of the SST in the tropical Pacific region. A cold bias of about 0.5–1.0 °C is noted in the ENSO region during all seasons at 1-month lead times. A statistically significant positive correlation coefficient is observed for the predicted SST anomalies in the tropical Pacific Ocean that lasts out to 6 months. Across varying times of the year and lead times, the model shows higher skill for autumn and winter target seasons than for spring or summer ones. The skill of the Saudi-KAU model in predicting Niño 3.4 index is comparable to that of state-of-the-art models available in the Copernicus Climate Change Service (C3S) and North American Multi-Model Ensemble (NMME) projects. The ENSO skill demonstrated in this study is potentially useful for regional climate services providing early warning for precipitation and temperature variations on sub-seasonal to seasonal time scales.

Developing a framework for an early warning system of seasonal temperature and rainfall tailored to aquaculture in Bangladesh

The occurrence of high temperature and heavy rain events during the monsoon season are a major climate risk affecting aquaculture production in Bangladesh. Despite the advances in the seasonal forecasting, the development of operational tools remains a challenge. This work presents the development of a seasonal forecasting approach to predict the number of warm days (NWD) and number of heavy rain days (NHRD) tailored to aquaculture in two locations of Bangladesh (Sylhet and Khulna). The approach is based on the use of meteorological and pond temperature data to generate linear models of the relationship between three-monthly temperature and rainfall statistics and NWD and NHRD, and on the evaluation of the skill of three operational dynamical models from the North American Multi-Model Ensemble (NMME) project. The linear models were used to evaluate the forecasts for two seasons and 1-month lead time: May to July (MJJ), forecast generated in April, and August to October (ASO), forecast generated in July. Differences were observed in the skill of the models predicting maximum temperature and rainfall (Spearman correlation, Root Mean Square Error, Bias statistics, and Willmott’s Index of Agreement,), in addition to NWD and NHRD from linear models, which also vary for the target seasons and location. In general, the models show higher predictive skill for NWD than NHRD, and for Sylhet than in Khulna. Among the three evaluated NMME models, CanSIPSv2 and GFDL-SPEAR exhibit the best performance, they show similar features in terms of error metrics, but CanSIPSv2 presents a lower interannual standard deviation.

Rainy season precipitation forecasts in coastal Peru from the North American Multi‐Model Ensemble

AbstractWe assessed operational forecasts of rainy season (January–March [JFM]) precipitation over coastal Peru from seven global climate models (GCMs) in the North American Multi‐Model Ensemble (NMME) from 1982 to 2020. Similar to previous representations by GCMs, regional climate models, and atmospheric reanalyses, the NMME models produce excessive precipitation (more than twice the observed values) over all of coastal Peru. Based on a set of commonly used metrics (correlation, mean error, RMSE, and Heidke skill score), there is a wide range in accuracy among the NMME models. As expected, forecast accuracy degrades with increasing lead time from December to July, but the rate of decline varies greatly by region and among models. The cause(s) of differences in model performance remains unclear, but resolution does not appear to be a significant factor. Highest forecast skill is found for southwestern Peru (correlations with observations are significant at all lead times), probably because of its stronger relationship between rainfall and El Niño–Southern Oscillation (ENSO), and lowest skill occurs for west‐central Peru (insignificant correlations at all lead times). The distribution of interannual rainy season precipitation in northwestern Peru is extremely positively skewed (skewness &gt;2), making evaluation of forecasts difficult from only a few years that deviate strongly from the mean. Improved correlations with observations across regions are generally found by subsetting multi‐model mean forecasts using the better‐performing models. Over northwest Peru at short lead times, NMME rainy season precipitation forecasts outperform a simple statistical correlation between JFM rainfall and ENSO, but not at longer leads, while NMME's forecasts for west‐central Peru underperform statistical correlations at all lead times. Over southwest Peru, the NMME models are consistently more skilful than a statistical correlation, and these differences become larger with increasing lead time.

The Role of Atmospheric Drivers in a Sudden Transition of California Precipitation in the 2012/13 Winter

AbstractCalifornia experienced a sharp transition from wet to historically dry conditions during the 2012/13 winter. Such transitions in the precipitation regime have strong consequences for water management, but predicting them at the seasonal time scale remains very challenging as little is known about what drives them. The 2012/13 winter was characterized by a neutral El Niño Southern Oscillation, strong Madden‐Julian Oscillation (MJO) activity, and sudden stratospheric warming. This study examines the potential influence of these different drivers in atmospheric global climate model experiments with prescribed sea surface temperature (SST) and sea ice concentration and constrained tropical and/or Arctic variability. Our model results are compared to reforecasts from the North American Multi‐Model Ensemble (NMME) and Subseasonal Experiment (SubX)/Subseasonal to Seasonal (S2S) projects to determine which important processes may have been missed by the seasonal prediction systems. Our simulations suggest that the tropical warm‐west, cool‐east SST anomaly played an active role in the wet conditions of November–December 2012 through a typical tropical‐extratropical teleconnection. The tropical SST anomalies were not accurately predicted by the NMME models, and none of them predicted the wet anomalies in November–December. In contrast, the sudden transition toward dry conditions in January is found to be independent of tropical SST variability, and rather influenced by strong MJO activity over the western Pacific. The SubX/S2S models with a better MJO prediction tend to predict the dry conditions in January better. Our study highlights how intraseasonal processes superimpose on more persistent seasonal anomalies and induce regime shifts such as the wet‐to‐dry transition of the 2012/13 winter.

Assessing ENSO Summer Teleconnections, Impacts, and Predictability in North America

AbstractDuring the summer when an El Niño event is transitioning to a La Niña event, the extratropical teleconnections exert robust warming anomalies over the U.S. Midwest threatening agricultural production. This study assesses the performance of current climate models in capturing the prominent observed extratropical responses over North America during the transitioning La Niña summer, based on atmospheric general circulation model experiments and coupled models from the North American Multimodel Ensemble (NMME). The ensemble mean of the SST-forced experiments across the transitioning La Niña summers does not capture the robust warming in the Midwest. The SST-forced experiments do not produce consistent subtropical western Pacific (WP) negative precipitation anomalies and this leads to the poor simulations of extratropical teleconnections over North America. In the NMME models, with active air–sea interaction, the negative WP precipitation anomalies show better agreement across the models and with observations. However, the downstream wave train pattern and the resulting extratropical responses over North America exhibit large disagreement across the models and are consistently weaker than in observations. Furthermore, in these climate models, an anomalous anticyclone does not robustly translate into a warm anomaly over the Midwest, in disagreement with observations. This work suggests that, during the El Niño to La Niña transitioning summer, active air–sea interaction is important in simulating tropical precipitation over the WP. Nevertheless, skillful representations of the Rossby wave propagation and land–atmosphere processes in climate models are also essential for skillful simulations of extratropical responses over North America.

Did a skillful prediction of near-surface temperatures help or hinder forecasting of the 2012 US drought?

This study aims to understand the role of near-surface temperatures in predicting US climatic extremes using the North American Multi-Model Ensemble (NMME) system. Here, the forecasting skill was measured by anomaly correlation coefficient (ACC) between the observed and forecasted precipitation (PREC)/2-meter air temperature (T2m) anomalies over the contiguous United States (CONUS) during 1982–2012. The strength of the T2m–PREC coupling was measured by ACC between observed PREC and T2m or forecasted PREC and T2m over the CONUS. We also assessed the NMME forecasting skill for the summers of 2004 (spatial anomaly correlation between PREC and T2m: 0.05), 2011 (−0.65), and 2012 (−0.60) when the T2m–PREC coupling was weaker or stronger than the 1982–2012 climatology (ACC: −0.34). We found that most of the NMME models show the bias of stronger T2m–PREC coupling than the observed coupling over 1982–2012, indicating that they failed to reproduce the interannual variability of T2m–PREC coupling. Some NMME models with skillful prediction for T2m show the skillful prediction of the precipitation anomalies and US droughts in 2011 and 2012 via strong T2m–PREC coupling despite the fact that the forecasting skill is year-dependent and model-dependent. Most of the NMME models show the limited seasonal forecasting skill of the PREC surplus from active Atlantic tropical cyclones in the summer of 2004 and thus fail to reproduce weak T2m–PREC coupling. Lastly, we explored how the role of sea surface temperatures in predicting T2m and PREC. The findings of this study suggest a need for the selective use of the current NMME seasonal forecasts for US droughts and pluvials.

Evaluation of the North American multi-model ensemble for monthly precipitation forecast

The North American multi-model ensemble (NMME) is a multi-model seasonal forecasting system consisting of a collection of models generated from several climate modelling centers. This research examined the monthly precipitation in North Maluku generated by five NMME models. The purpose of this research is to assess the performance of monthly precipitation prediction by using RMSE and Rank Histogram analysis. The NMME models are verified against observed precipitation. The analysis shows that they are biased and underdispersive. Among the five NMME models, the Center for Ocean-Land-Atmosphere Studies (COLA) exhibits the best predictive skill. The performances of the Canadian Meteorological Centre (CMC) are relatively worse than that of the other models. The COLA model shows relatively high skill when used to forecast May-November monthly precipitation. Meanwhile, the National Oceanic and Atmospheric Administration (NOAA)’s Geophysical Fluid Dynamics Laboratory (GFDL) model shows high skill in December-April periods. The ensemble forecast is calibrated with the BMA approach in order to obtain reliable forecasts.

Improving the precipitation forecasts of the North-American multi model ensemble (NMME) over Sistan basin

Poor forecasting of climate variables at river basin scale, leads to situations which entail considerable risks and losses in various sectors such as the agriculture, the environment and water resources. Lack of accurate sub-seasonal and seasonal precipitation predictions are major management and operation predicaments in the transboundary Sistan Basin, located in Afghanistan and Iran, and selected as the case study in this paper. The outputs of theNorth-American Multi Model Ensemble (NMME), as a seasonalforecasting system, were utilized in the target region. In order to evaluate the performance of the raw outputs of four NMME models (namely: NCEP-CFSv2, CMC-CanCM3, CMC2-CanCM4, NCAR-CCSM4), the individual model ensemble mean was compared to the observations. Results highlighted the need for post-processing of the NMME outputs. For this purpose, three different commonly used methods, including GrandNMME, bias correction quantile mapping (QM) and Copula approaches, alongside a novel hybrid technique, were applied to improve future predictions of precipitation patterns for the hindcast period of 1982–2010 and forecast period of 2012–2016. All four methods were performed for each month in each 0.5-degree cell over the following one to six months. The findings demonstrated the performance of different NMME models were not the same over time and space. Among the models, the NCEP-CFSv2 and CMC2-CanCM4 seemed to best capture the spatial variability of precipitation in the study area. In addition, they performed better in estimating the amounts of precipitation. Among the methods, the QM, the Copula and the hybrid were all successful in improving the geographic patterns of the data. Furthermore, they were all able to enhance the accuracy of the data. At different timescales, the hybrid method showed substantial improvements over the seasonal predictions. Overall, the findings illustrated that collective use of methods may narrow the potential uncertainty in projections of regional changes.

Exploring the predictability of within-season rainfall statistics of the Bangladesh monsoon using North American Multimodel Ensemble outputs

Improvement of rainfall forecasts on seasonal to within-season timescales is crucial for many vulnerable regions and nations, including Bangladesh. For South Asia, seasonal predictability of rainfall can be quite challenging, and Bangladesh has limited predictive skill with respect to total seasonal rainfall due to its weak relationship with ENSO variability. The relationship between total seasonal monsoon rainfall from June through September (JJAS), as simulated by the North American Multimodel Ensemble (NMME), and within-season observed rainfall statistics for Bangladesh were explored, employing 25-year cross-validations at lead times up to 3 months. The model hindcasts of total JJAS rainfall demonstrate only low-to-modest skill at predicting total observed seasonal rainfall, but a more robust predictive relationship is found for the number of observed dry and wet spells within the season, more so than with the number of extreme dry or wet days. A small ensemble of NMME models could be used to provide more valuable information to aid decision-makers than previously thought, with important implications for agricultural decision-making and climate services.

Post-processing of the North American multi-model ensemble for monthly forecast of precipitation based on neural network models

The aim of this paper is to investigate the ability of artificial neural network (ANN) models for post-processing the monthly precipitation forecasts under North American multi-model ensemble (NMME) project and proposing a new multi-model ensemble neural network (MME-NN) model. Monthly precipitation hindcasts of eight models from NMME project are considered in this study. Multi-layer perceptron neural networks are employed for post-processing the output of the models in comparison with PERSIANN-CDR climatology data. Also, utilizing a multi-criteria decision-making approach, NMME models are ranked for each month. The study is implemented over Iran and detailed discussions are provided. The results indicate that the skill of NMME models is different for each month and for each region of the country. Also, it is shown that the neural network outperforms all NMME models for all months. By using the ranking of the models, for each month, the NMME models are ordered based on their skill and a monthly rank is devoted for each model.