Explore the vast range of titles available.

Long‐standing simulation errors limit the utility of climate models. Overlooked are tropical‐wide errors, with sea surface temperature (SST) biasing high or low across all the tropical ocean basins. Our analysis based on Coupled Model Intercomparison Project (CMIP) multi‐model ensembles shows that such SST biases can be classified into two types: one with a broad meridional structure and of the same sign across all basins that is highly correlated with the tropical mean; and one with large inter‐model variability in the cold tongues of the equatorial Pacific and Atlantic. The first type can be traced back to biases in atmospheric simulations of cloud cover, with cloudy models biasing low in tropical‐wide SST. The second type originates from the diversity among models in representing the thermocline depth; models with a deep thermocline feature a warm cold tongue on the equator. Implications for inter‐model variability in precipitation climatology and SST threshold for convection are discussed. Key Points Our analysis suggests two types of tropical‐wide SST biases in climate models The first type originates from biases in atmospheric simulations of cloud cover The second type is linked to oceanic representation of the thermocline depth

Forecasts of maximum and minimum air temperatures are essential to mitigate the damage of extreme weather events such as heat waves and tropical nights. The Numerical Weather Prediction (NWP) model has been widely used for forecasting air temperature, but generally it has a systematic bias due to its coarse grid resolution and lack of parametrizations. This study used random forest (RF), support vector regression (SVR), artificial neural network (ANN) and a multi‐model ensemble (MME) to correct the Local Data Assimilation and Prediction System (LDAPS; a local NWP model over Korea) model outputs of next‐day maximum and minimum air temperatures ( Tmaxt+1 and Tmint+1) in Seoul, South Korea. A total of 14 LDAPS model forecast data, the daily maximum and minimum air temperatures of in‐situ observations, and five auxiliary data were used as input variables. The results showed that the LDAPS model had an R2 of 0.69, a bias of −0.85 °C and an RMSE of 2.08 °C for Tmaxt+1 forecast, whereas the proposed models resulted in the improvement with R2 from 0.75 to 0.78, bias from −0.16 to −0.07 °C and RMSE from 1.55 to 1.66 °C by hindcast validation. For forecasting Tmint+1, the LDAPS model had an R2 of 0.77, a bias of 0.51 °C and an RMSE of 1.43 °C by hindcast, while the bias correction models showed R2 values ranging from 0.86 to 0.87, biases from −0.03 to 0.03 °C, and RMSEs from 0.98 to 1.02 °C. The MME model had better generalization performance than the three single machine learning models by hindcast validation and leave‐one‐station‐out cross‐validation. Key Points Machine learning based bias correction of air temperature forecasts of a numerical model All machine learning models improved prediction skills of air temperature An ensemble of three machine learning models resulted in more robust bias correction than individual machine learning models

Data assimilation (DA) aims to optimally combine model forecasts and observations that are both partial and noisy. Multi‐model DA generalizes the variational or Bayesian formulation of the Kalman filter, and we prove that it is also the minimum variance linear unbiased estimator. Here, we formulate and implement a multi‐model ensemble Kalman filter (MM‐EnKF) based on this framework. The MM‐EnKF can combine multiple model ensembles for both DA and forecasting in a flow‐dependent manner; it uses adaptive model error estimation to provide matrix‐valued weights for the separate models and the observations. We apply this methodology to various situations using the Lorenz96 model for illustration purposes. Our numerical experiments include multiple models with parametric error, different resolved scales, and different fidelities. The MM‐EnKF results in significant error reductions compared to the best model, as well as to an unweighted multi‐model ensemble, with respect to both probabilistic and deterministic error metrics. Plain Language Summary Forecasts that combine multiple imperfect models of a system are used in many fields, including the physical, natural and socio‐economic sciences. In particular, data assimilation (DA), the process by which observations are integrated with model forecasts, is critical in the prediction of chaotic systems. Multi‐model DA (MM‐DA) unifies multi‐model forecast combination and DA into a single process. Here, we significantly improve on previous formulations of MM‐DA by accounting for model error, and formulate a multi‐model ensemble Kalman filter appropriate for high‐dimensional systems. Key Points Multiple models and observations can be optimally combined for data assimilation (DA) and forecasting using multi‐model DA We formulate a multi‐model ensemble Kalman filter (MM‐EnKF), which incorporates model error and is appropriate for high‐dimensional models Using numerical experiments, we show that the MM‐EnKF can significantly outperform the best model and an unweighted multi‐model ensemble

Evapotranspiration (ET) accounts for the majority of water available from precipitation in the terrestrial water cycle, and improvements to the accuracy, resolution, and coverage of ET data can enhance hydrologic models and assessments. The OpenET collaboration of six remotely sensed ET modeling teams has demonstrated that an ensemble approach to ET estimation generally provides improved accuracy relative to individual ensemble members. The performance of individual models has been shown to vary by land cover type and climate zone, but a thorough study of the variables that influence model performance differences has not yet been conducted. In this paper, we model the performance of OpenET models relative to flux tower data as a function of variables such as land cover type and precipitation. These performance models are used to map estimated OpenET model performance across the conterminous United States. We develop relative weights based on these modeled performance metrics and show that a performance‐weighted ensemble improves accuracy relative to the current OpenET ensemble method to varying degrees. The monthly mean absolute error of the weighted ensemble is reduced relative to the current method by 8% in agricultural settings, by 23% in shrublands and mixed forests, and by 5% in grasslands and evergreen forests. We produce weight maps that can be used to generate performance‐weighted ensemble values for OpenET data. The results can be used to inform model selection and provide insight about the controls on model performance that could lead to model refinement.

Since its establishment in 2005, the APEC Climate Center (APCC) has pioneered advancements in seasonal climate prediction through its Multi‐Model Ensemble (MME) system, integrating the world's most diverse range of dynamical climate models. Over the past two decades, APCC has incorporated long‐range forecasts from more than 60 model versions, contributed by 21 institutions across 11 countries. This study presents the first diachronic assessment of the APCC MME system evolution, focusing on operational model transitions and associated substantial 34% improvement in global forecast skill. Despite these significant advances, challenges persist, particularly in predicting precipitation over land areas in the northern extratropics and in overcoming the spring predictability barriers. Our findings underscore the crucial role of model innovation, increased diversity, and international collaboration in advancing seasonal prediction. As the first study of its kind, this work provides key insights into the evolution of climate modeling, providing a foundation for future forecasting improvements.

We investigate why the North American Multi‐Model Ensemble (NMME) upper‐level height forecast for December–February (DJF) 2023/24 differs from the expected El Niño response. These atypical height anomalies emerged despite the fact a strong El Niño was forecast. The analysis focuses on diagnosing the NMME forecasts of DJF 2023/24 for SSTs and 200‐hPa heights initialized at the beginning of November 2023 relative to other ensemble mean NMME DJF forecasts dating back to 1982. The results demonstrate that forecasts of the 200‐hPa height anomalies had a large contribution from warming trends in global SSTs. It is the combination of trends and the expected El Niño teleconnection that results in the forecast height anomalies. Increasingly, for forecasts of geopotential height anomalies during the recent El Niño winters, the amplitude of trends is nearly equal to the signal from El Niño and has implications for the climatological base period selection for seasonal forecasts. Plain Language Summary Seasonal forecasts are cast as anomalies as users want to know what can be expected beyond the typical seasonal swings of the climate. This necessitates a choice for the climatological base period relative to which forecast anomalies are computed. It, however, poses a challenge under rapid climate change. In this scenario, climate trends become part of the real‐time forecast anomalies, and if the climatological base period is sufficiently different, may even start to dominate. This was the case for the NMME DJF 2023/24 forecast of 200‐hPa heights which was forecast to be a strong El Niño, and yet, forecast for 200‐hPa heights differed from typical El Niño signal. The analysis implies that seasonal forecasts for some variables, consideration of trends is important and reliance on expected signal from El Niño—Southern Oscillation alone may not be sufficient. Key Points The North American Multi‐Model Ensemble (NMME) seasonal forecasts for December–February (DJF) 2023/24 upper‐level height differ from the expected El Niño signal It is the combination of trends in heights and the expected El Niño signal that results in the forecast NMME ensemble mean heights anomalies The forecast of trends is increasingly important to account for NMME forecast anomaly and their amplitude in recent years can be of same magnitude as the signal from El Niño

As climate change continues to impact the planet, the importance of forests is becoming increasingly emphasized. The International Co-operative Program on the Assessment and Monitoring of Air Pollution Effects on Forests (ICP Forests) has been monitoring and assessing forests in 40 countries since 1985. In Republic of Korea, the first Forest Health Management (FHM) survey was a nationwide sample point assessment conducted between 2011 and 2015. However, there are limitations in representing the health of forests that occupy 63.7% of Korea’s land area due to the nature of sample point surveys, which survey a relatively small area. Accordingly, a species richness map was created to promote species diversity in forest health evaluations in Republic of Korea. The map was created using data from the first FHM survey, which examined 28 factors with 12 survey indicators in four categories: tree health, vegetation health, soil health, and atmospheric health. We conducted an ensemble modeling of species distribution for woody plant species that are major habitats in Republic of Korea. To select the species, we used the first FHM survey data and chose those with more than 100 sample points, resulting in a total of 11 species. We then created the species richness map of Republic of Korea by overlaying their distributions. To verify the accuracy of the derived map, an independent verification was conducted using statistical verification and external data from the National Natural Environment Survey. To support forest management that accounts for climate change adaptation, the derived species richness map was validated based on the vegetation climate distribution map of the Korean Peninsula, which was published by the Korea National Arboretum. The map confirmed that species richness is highest around the boundary of the deciduous forest in the central temperate zone and lowest around the evergreen and deciduous mixed forest in the southern temperate zone. By establishing this map, it was possible to confirm the spatial distribution of species by addressing the limitations of direct surveys, which are unable to represent all forests. However, it is important to note that not all factors of the first FHM survey were considered during the spatialization process, and the target area only includes Republic of Korea. Thus, further research is necessary to expand the target area and include additional items.

A new 46‐year hindcast dataset for seasonal‐to‐annual ensemble predictions has been created using a multi‐model ensemble of 5 state‐of‐the‐art coupled atmosphere‐ocean circulation models. The multi‐model outperforms any of the single‐models in forecasting tropical Pacific SSTs because of reduced RMS errors and enhanced ensemble dispersion at all lead‐times. Systematic errors are considerably reduced over the previous generation (DEMETER). Probabilistic skill scores show higher skill for the new multi‐model ensemble than for DEMETER in the 4–6 month forecast range. However, substantially improved models would be required to achieve strongly statistical significant skill increases. The combination of ENSEMBLES and DEMETER into a grand multi‐model ensemble does not improve the forecast skill further. Annual‐range hindcasts show anomaly correlation skill of ∼0.5 up to 14 months ahead. A wide range of output from the multi‐model simulations is becoming publicly available and the international community is invited to explore the full scientific potential of these data.

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 modeling 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. We found 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: (a) no individual model performed best at every depth and horizon (days in the future), and (b) 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 utilizing MMEs, rather than individual models, in operational forecasts. Key Points Aggregated lake temperature forecast skill was higher for multi‐model ensemble (MME) forecasts than individual model forecasts Including baseline empirical models (day‐of‐year, persistence) with process models improved MME forecast performance MME forecasts improved forecast skill by “hedging,” as no individual model performed best at all horizons or depths

We consider paradigms for interpretation and analysis of the CMIP3 ensemble of climate model simulations. The dominant paradigm in climate science, of an ensemble sampled from a distribution centred on the truth, is contrasted with the paradigm of a statistically indistinguishable ensemble, which has been more commonly adopted in other fields. This latter interpretation (which gives rise to a natural probabilistic interpretation of ensemble output) leads to new insights about the evaluation of ensemble performance. Using the well‐known rank histogram method of analysis, we find that the CMIP3 ensemble generally provides a rather good sample under the statistically indistinguishable paradigm, although it appears marginally over‐dispersive and exhibits some modest biases. These results contrast strongly with the incompatibility of the ensemble with the truth‐centred paradigm. Thus, our analysis provides for the first time a sound theoretical foundation, with empirical support, for the probabilistic use of multi‐model ensembles in climate research.