Probability Of Extreme Events Research Articles

Extreme event probability estimation using PDE-constrained optimization and large deviation theory, with application to tsunamis

We propose and compare methods for the analysis of extreme events in complex systems governed by PDEs that involve random parameters, in situations where we are interested in quantifying the probability that a scalar function of the system's solution is above a threshold. If the threshold is large, this probability is small and its accurate estimation is challenging. To tackle this difficulty, we blend theoretical results from large deviation theory (LDT) with numerical tools from PDE-constrained optimization. Our methods first compute parameters that minimize the LDT-rate function over the set of parameters leading to extreme events, using adjoint methods to compute the gradient of this rate function. The minimizers give information about the mechanism of the extreme events as well as estimates of their probability. We then propose a series of methods to refine these estimates, either via importance sampling or geometric approximation of the extreme event sets. Results are formulated for general parameter distributions and detailed expressions are provided when Gaussian distributions. We give theoretical and numerical arguments showing that the performance of our methods is insensitive to the extremeness of the events we are interested in. We illustrate the application of our approach to quantify the probability of extreme tsunami events on shore. Tsunamis are typically caused by a sudden, unpredictable change of the ocean floor elevation during an earthquake. We model this change as a random process, which takes into account the underlying physics. We use the one-dimensional shallow water equation to model tsunamis numerically. In the context of this example, we present a comparison of our methods for extreme event probability estimation, and find which type of ocean floor elevation change leads to the largest tsunamis on shore.

Enhancement of extreme events through the Allee effect and its mitigation through noise in a three species system

We consider the dynamics of a three-species system incorporating the Allee Effect, focussing on its influence on the emergence of extreme events in the system. First we find that under Allee effect the regular periodic dynamics changes to chaotic. Further, we find that the system exhibits unbounded growth in the vegetation population after a critical value of the Allee parameter. The most significant finding is the observation of a critical Allee parameter beyond which the probability of obtaining extreme events becomes non-zero for all three population densities. Though the emergence of extreme events in the predator population is not affected much by the Allee effect, the prey population shows a sharp increase in the probability of obtaining extreme events after a threshold value of the Allee parameter, and the vegetation population also yields extreme events for sufficiently strong Allee effect. Lastly we consider the influence of additive noise on extreme events. First, we find that noise tames the unbounded vegetation growth induced by Allee effect. More interestingly, we demonstrate that stochasticity drastically diminishes the probability of extreme events in all three populations. In fact for sufficiently high noise, we do not observe any more extreme events in the system. This suggests that noise can mitigate extreme events, and has potentially important bearing on the observability of extreme events in naturally occurring systems.

Quantifying the contribution of an individual to making extreme weather events more likely

Probabilistic event attribution aims to quantify the role of anthropogenic climate change in altering the intensity or probability of extreme climate and weather events. It was originally conceived to calculate the costs associated with any increased likelihood of the meteorological event in question. However, only recently have such studies attempted to divide liability between polluting nations and ascribe a cost. Recent protests indicate a perception that older generations have the greater responsibility for climate change. In this paper, we examine how a portion of the cost of an event can be attributed to any individual person, according to their age and nationality. We demonstrate that this is quantitatively feasible using the example of the 2018 summer heatwave in eastern China and its impact on aquaculture. A relatively simple technique finds sample individuals responsible for between 0.53 and 18.10 yuan, increasing with their age and their country’s emissions over their adult lifetime since the first international consensus on carbon emissions was reached. This provides an illustration of the scale of such responsibilities, and how it is affected by national development and demographics. Such data can support decisions, at national and international levels, on how to fund recovery from climate impacts. It offers a simple quantitative approach for individuals to know their impact on the climate, or for governments to use in making policy decisions about how best to distribute costs of climate change.

Adaptation Strategies of Residential Buildings Based on a Health Risk Evaluation—A Case Study of Townhouses in Taiwan

Global warming increases the probability of extreme events and heat waves triggering severe impacts on human health, especially the elderly. Taiwan is an aged society, so residential buildings, which cannot withstand extreme temperature events, increase the risk of harm for the elderly. Furthermore, Taiwanese prefer to open the windows to reduce indoor high temperatures, which causes high levels of outdoor PM2.5 to flow indoors, leading to health risks. Therefore, this research proposes a strategy to create a house with a low temperature and a low PM2.5 health risk for the elderly based on building envelope renovation and windows user behavior patterns. The risk day is demonstrated as an index to evaluate the indoor environment quality, which is based on the number of days that exceed the health risk threshold. The results show that the performance improvement of the building envelope and control of the window opening timing can effectively reduce the risk days by 48.5%. This means that passive strategies cannot fully control health risks, and the use of equipment is necessary. Finally, if the current situation is maintained without any adjustment or strategy improvement, an additional 41.3% energy consumption must be paid every year to control health risks.

Impact of large wildfires on PM<sub>10</sub> levels and human mortality in Portugal

Abstract. Uncontrolled wildfires have a substantial impact on the environment, the economy and local populations. According to the European Forest Fire Information System (EFFIS), between 2000 and 2013 wildfires burned up to 740 000 ha of land annually in the south of Europe, Portugal being the country with the highest percentage of burned area per square kilometre. However, there is still a lack of knowledge regarding the impacts of the wildfire-related pollutants on the mortality of the country's population. All wildfires occurring during the fire season (June–July–August–September) from 2001 and 2016 were identified, and those with a burned area above 1000 ha (large fires) were considered for the study. During the studied period (2001–2016), more than 2 million ha of forest (929 766 ha from June to September alone) were burned in mainland Portugal. Although large fires only represent less than 1 % of the number of total fires, in terms of burned area their contribution is 46 % (53 % from June to September). To assess the spatial impact of the wildfires, burned areas in each region of Portugal were correlated with PM10 concentrations measured at nearby background air quality monitoring stations. Associations between PM10 and all-cause (excluding injuries, poisoning and external causes) and cause-specific mortality (circulatory and respiratory) were studied for the affected populations using Poisson regression models. A significant positive correlation between burned area and PM10 was found in some regions of Portugal, as well as a significant association between PM10 concentrations and mortality, these being apparently related to large wildfires in some of the regions. The north, centre and inland of Portugal are the most affected areas. The high temperatures and long episodes of drought expected in the future will increase the probabilities of extreme events and therefore the occurrence of wildfires.

A Spatial Structure Variable Approach to Characterize Storm Events for Coastal Flood Hazard Assessment

Over the last decades, the evaluation of hazards and risks associated with coastal flooding has become increasingly more important in order to protect population and assets. The general purpose of this research was to assess reliable coastal flooding hazard maps due to overflow and wave overtopping. This paper addresses the problem of defining credible joint statistics of significant wave heights Hs and water levels ζ, focusing on the selection of the sample pair that characterizes each sea storm, to evaluate the occurrence probability of extreme events. The pair is selected maximizing a spatial structure variable, i.e., a linear combination of Hs and ζ, specific to each point of the area at risk. The structure variable is defined by the sensitivity of the flooding process to Hs and ζ, as found by analyzing a set of inundation maps produced through a Simplified Shallow-Water numerical model (SSW). The proposed methodology is applied to a coastal stretch in the Venetian littoral (Italy), by means of a 30 year-long time series recorded at the “Acqua Alta” oceanographic research tower, located in the Northern Adriatic Sea in front of the Venetian lagoon. The critical combination of Hs and ζ forming the structure variable is presented in a map, and it can be related to the topography and the presence of mitigation measures. The return period associated with the two recent large storms that occurred in this area in 2018 and 2019 is also investigated. The proposed procedure gives credible occurrence probabilities for these events, whereas other approaches would consider them extremely unlikely.

Extreme floods of Venice: characteristics, dynamics, past and future evolution (review article)

Abstract. Floods in the Venice city centre result from the superposition of several factors: astronomical tides; seiches; and atmospherically forced fluctuations, which include storm surges, meteotsunamis, and surges caused by atmospheric planetary waves. All these factors can contribute to positive water height anomalies individually and can increase the probability of extreme events when they act constructively. The largest extreme water heights are mostly caused by the storm surges produced by the sirocco winds, leading to a characteristic seasonal cycle, with the largest and most frequent events occurring from November to March. Storm surges can be produced by cyclones whose centres are located either north or south of the Alps. Historically, the most intense events have been produced by cyclogenesis in the western Mediterranean, to the west of the main cyclogenetic area of the Mediterranean region in the Gulf of Genoa. Only a small fraction of the inter-annual variability in extreme water heights is described by fluctuations in the dominant patterns of atmospheric circulation variability over the Euro-Atlantic sector. Therefore, decadal fluctuations in water height extremes remain largely unexplained. In particular, the effect of the 11-year solar cycle does not appear to be steadily present if more than 100 years of observations are considered. The historic increase in the frequency of floods since the mid-19th century is explained by relative mean sea level rise. Analogously, future regional relative mean sea level rise will be the most important driver of increasing duration and intensity of Venice floods through this century, overcompensating for the small projected decrease in marine storminess. The future increase in extreme water heights covers a wide range, largely reflecting the highly uncertain mass contributions to future mean sea level rise from the melting of Antarctica and Greenland ice sheets, especially towards the end of the century. For a high-emission scenario (RCP8.5), the magnitude of 1-in-100-year water height values at the northern Adriatic coast is projected to increase by 26–35 cm by 2050 and by 53–171 cm by 2100 with respect to the present value and is subject to continued increase thereafter. For a moderate-emission scenario (RCP4.5), these values are 12–17 cm by 2050 and 24–56 cm by 2100. Local subsidence (which is not included in these estimates) will further contribute to the future increase in extreme water heights. This analysis shows the need for adaptive long-term planning of coastal defences using flexible solutions that are appropriate across the large range of plausible future water height extremes.

Event attribution of Parnaíba River floods in Northeastern Brazil

AbstractThe climate modeling techniques of event attribution enable systematic assessments of the extent that anthropogenic climate change may be altering the probability or magnitude of extreme events. In the consecutive years of 2018, 2019, and 2020, rainfalls caused repeated flooding impacts in the lower Parnaíba River in Northeastern Brazil. We studied the effect that alterations in precipitation resulting from human influences on the climate had on the likelihood of flooding using two ensembles of the HadGEM3‐GA6 atmospheric model: one driven by both natural and anthropogenic forcings; and the other driven only by natural atmospheric forcings, with anthropogenic changes removed from sea surface temperatures and sea ice patterns. We performed hydrological modeling to base our assessments on the peak annual streamflow. The change in the likelihood of flooding was expressed in terms of the ratio between probabilities of threshold exceedance estimated for each model ensemble. With uncertainty estimates at the 90% confidence level, the median (5% 95%) probability ratio at the threshold for flooding impacts in the historical period (1982–2013) was 1.12 (0.97 1.26), pointing to a marginal contribution of anthropogenic emissions by about 12%. For the 2018, 2019, and 2020 events, the median (5% 95%) probability ratios at the threshold for flooding impacts were higher at 1.25 (1.07 1.46), 1.27 (1.12 1.445), and 1.37 (1.19 1.59), respectively; indicating that precipitation change driven by anthropogenic emissions has contributed to the increase of likelihood of these events by about 30%. However, there are other intricate hydrometeorological and anthropogenic processes undergoing long‐term changes that affect the flood hazard in the lower Parnaíba River. Trend and flood frequency analyses performed on observations showed a nonsignificant long‐term reduction of annual peak flow, likely due to decreasing precipitation from natural climate variability and increasing evapotranspiration and flow regulation.

Stability of spatial dependence structure of extreme precipitation and the concurrent risk over a nested basin

Extreme low and high precipitation can cause natural disasters (e.g., drought and flood events) which have devastatingeffects on natural environment and the human society. In this study, we present a framework for assessing the stability of spatial dependence of extreme precipitation and their concurrence probability in a nested catchment. The Pearl River basin was used as a case study to test the applicability of the framework. This framework is threefold. This first part involves modeling spatial dependence of both summer maxima and winter minima precipitation between pairs of scattered gauges over the basin by employing the pair-copula constructions as a baseline model. The second part of this framework aims to identify the variability of spatial dependence as a result of the effects of external drivers. Specifically, variations in the spatial dependence of extreme precipitation in the Pearl River basin over different periods associated with the influence of large-scale climate signals were examined. The final part quantifies the probability of occurring extreme precipitation-related events simultaneously over broader areas related to a specific event at a given gauge and the possible spatial extents being affected. The results indicate that the presented framework is able to capture the large-scale spatial dependence structures of both summer maxima and winter minima precipitation. It also allows for explicitly estimating the pairwise concurrence probability of extreme high and low precipitation events and event-impacted areas across the basin.

The Alaskan Summer 2019 Extreme Heat Event: The Role of Anthropogenic Forcing, and Projections of the Increasing Risk of Occurrence

AbstractExtreme heat occurred over Alaska in June–July 2019, posing risks to infrastructure, ecosystem, and human health. It is vital to improve our understanding of the causes of such events and the extent to which anthropogenic forcing may alter their likelihood and magnitude. Here, we use multiple large ensembles of climate models, comprising thousands of simulated years, to investigate these issues. Our results suggest that the presence of anthropogenic radiative forcing increased the likelihood of the 2019 extreme heat event by as much as 6%. Further we show the rate of occurrence of such an extreme heat event is likely to substantially increase in the future with increasing levels of atmospheric greenhouse gases. While uncertainty in projected climate risk from model choice leads to a broad range of future extreme heat event probabilities, some models project that with rapidly increasing levels of greenhouse gases the likelihood of such events would exceed 75% by 2090.

Understanding human influence on climate change in China.

China's climate has been warming since the 1950s, with surface air temperature increasing at a rate higher than the global average. Changes in climate have exerted substantial impacts on water resources, agriculture, ecosystems and human health. Attributing past changes to causes provides a scientific foundation for national and international climate policies. Here, we review recent progress in attributing the observed climate changes over past decades in China. Anthropogenic forcings, dominated by greenhouse gas emissions, are the main drivers for observed increases in mean and extreme temperatures. Evidence of the effect of anthropogenic forcings on precipitation is emerging. Human influence has increased the probability of extreme heat events, and has likely changed the occurrence probabilities for some heavy precipitation events. The way a specific attribution question is posed and the conditions under which the question is addressed present persistent challenges for appropriately communicating attribution results to non-specialists.

Intraseasonal Variability of Sea Level in the Western North Pacific

AbstractSea levels in the Western North Pacific (WNP) are presented with anomalous intraseasonal variations. This study examines the response of sea level in the WNP to the atmospheric Intraseasonal Oscillation modes, namely the Madden–Julian Oscillation (MJO) and the Boreal Summer Intraseasonal Oscillation (BSISO), using the 25 years (1993–2017) of satellite altimetry and barotropic model output. In winter, the MJO has significant effects on the component of sea level due to the instant wind and atmospheric pressure effects (high‐frequency), showing an eastward propagation pattern in most regions, with the strongest effects in the western marginal seas. The MJO‐associated pattern of dynamical (low‐frequency) component of sea level propagates southward, with the significant effects mostly in the tropics. In summer, the BSISO‐associated pattern of the high‐frequency component of sea level moves from southwest to northeast, with the largest anomalies in the middle of WNP (20°N‐30°N), while the strongest BSISO effects on the low‐frequency component are detectable mostly in the coasts of China and east of the Philippines. The MJO and BSISO can also modulate the probability of extreme sea level events. In winter, during phases 2–5, MJO increases the chance of extreme high events in the high‐frequency component of sea level by 100%–200% in the western coasts and the tropics. In summer, in BSISO phases 6–7, the chance of extreme high events in the high‐frequency component of sea level is enhanced by >300% in the South China Sea and east of the Philippines.

Subseasonal prediction of springtime Pacific–North American transport using upper-level wind forecasts

Abstract. Forecasts of Pacific jet variability are used to predict stratosphere-to-troposphere transport (STT) and tropical-to-extratropical moisture export (TME) during boreal spring over the Pacific–North American region. A retrospective analysis first documents the regionality of STT and TME for different Pacific jet patterns. Using these results as a guide, Pacific jet hindcasts, based on zonal-wind forecasts from the European Centre for Medium-Range Weather Forecasting Integrated Forecasting System, are utilized to test whether STT and TME over specific geographic regions may be predictable for subseasonal forecast leads (3–6 weeks ahead of time). Large anomalies in STT to the mid-troposphere over the North Pacific, TME to the west coast of the United States, and TME over Japan are found to have the best potential for subseasonal predictability using upper-level wind forecasts. STT to the planetary boundary layer over the intermountain west of the United States is also potentially predictable for subseasonal leads but likely only in the context of shifts in the probability of extreme events. While STT and TME forecasts match verifications quite well in terms of spatial structure and anomaly sign, the number of anomalous transport days is underestimated compared to observations. The underestimation of the number of anomalous transport days exhibits a strong seasonal cycle, which becomes steadily worse as spring progresses into summer.

The projected effects of urbanization and climate change on summer thermal environment in Guangdong-Hong Kong-Macao Greater Bay Area of China

The individual and combined impacts of near-future urbanization and climate change on thermal environment in July over the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) are projected. Results indicate that both the urbanization and climate change lead to near-future temperature rise, with comparable daily mean urban-warming of ~0.51 °C and ~ 0.70 °C respectively. In which, the two factors also enhance the probability of extreme heat events, especially the role from climate change. However, the relative contributions of two factors on human thermal comfort are very different. Urbanization leads to drier daytime but warmer nighttime, resulting in more comfortable daytime but uncomfortable nighttime. Under the changing climate, significant and uniform increase of Humidex (HI) can be found in whole day, suggesting future climate change intensifies human discomfort thoroughly. Climate change takes up ~93.33% of the total increased HI, attributing to both the caused temperature rise and humidity increase. Contrarily, urbanization only contributes ~7.41%, because the urban-reduced humidity offsets the effect of urbanized warming. It is further revealed that exposure duration of dangerous discomfort (HI ≥ 45) for the GBA residents will increase at least 8.87% in 2030 under the effects of climate change and urbanization, suggesting necessity of mitigating climate change in future megalopolis.

Analysis of Dangerous Sea States in the Northwestern Mediterranean Area

Extreme sea waves, although rare, can be notably dangerous when associated with energetic sea states and can generate risks for the navigation. In the last few years, they have been the object of extensive research from the scientific community that helped with understanding the main physical aspects; however, the estimate of extreme waves probability in operational forecasts is still debated. In this study, we analyzed a number of sea-states that occurred in a precise area of the Mediterranean sea, near the location of a reported accident, with the objective of relating the probability of extreme events with different sea state conditions. For this purpose, we performed phase-resolving simulations of wave spectra obtained from a WaveWatch III hindcast, using a Higher Order Spectral Method. We produced statistics of the sea-surface elevation field, calculating crest distributions and the probability of extreme events from the analysis of a long time-series of the surface elevation. We found a good matching between the distributions of the numerically simulated field and theory, namely Tayfun second- and third- order ones, in contrast with a significant underestimate given by the Rayleigh distribution. We then related spectral quantities like angular spreading and wave steepness to the probability of occurrence of extreme events finding an enhanced probability for high mean steepness seas and narrow spectra, in accordance with literature results, finding also that the case study of the reported accident was not amongst the most dangerous. Finally, we related the skewness and kurtosis of the surface elevation to the wave steepness to explain the discrepancy between theoretical and numerical distributions.

Urban adaptive transportation system in extreme precipitation: System dynamic analysis based on Shanghai

Extreme precipitation is a major challenge affecting the normal functions of urban transportation, particularly in the context of climate change. National and municipal administrations have responded to the increased probability of extreme precipitation events by expanding drainage facilities, or by warning the public in time. In this study, we applied a dynamic system framework to examine how the interdependent factors in extreme precipitation from engineering and society interact with each other and form an adaptive system. Based on the concept of a complex adaptive system, we identified the adaptive units and paths of the system. We set the model parameters using data from Shanghai, and simulated the dynamic process of the system to 2032. We found that the effects of extreme precipitation on an urban transportation infrastructure can be controlled based on the long-term operation of the adaptive system, and that the adaptive units in the physical, social, and economic subsystems are connected with each other.

The influence of finite depth on the evolution of extreme wave statistics in numerical wave tanks

Model tests with unidirectional random wave fields are common for coastal engineering purposes. In this paper, we investigate the finite water depth effect on both the probability of extreme events and the averaged shape of them in such experimental conditions. We simulate some typical model test conditions with a numerical tool, which solves the fully non-linear wave equations. We find the analytical solution based on Nonlinear Schrödinger equation (NLS) to be accurate for low steepness and relatively deep water cases. However, the analytical solution underestimates the kurtosis at the steady state for high steepness cases in shallow water, and also gives zero value of kurtosis at critical water depth, whereas a small but non-zero value of kurtosis is observed in the numerical tank. We also investigate the averaged shape of the extreme events, which are modified by nonlinear physics over the distance. The horizontal asymmetry is significant initially but is greatly reduced to a steady state, and the contraction of the wave groups grows monotonically until the steady state is reached along the wave tank. Finite water depth has limited effects on the averaged shape of large events, especially for the extremely steep waves.

Future extreme heat wave events using Bayesian heat wave intensity-persistence day-frequency model and their uncertainty

Heat wave events are occurring more often and over a wider area around the world, and in the coming era of global warming, their intensity will intensify and their persistence days will be longer. In this study, the relationship between the occurrence probability, intensity, and persistency days of future extreme heat wave events is explained using the concept of IDF (rainfall intensity-duration-frequency) curves for extreme rainfall events. The uncertainty of the non-stationary heat wave intensity-persistence day-frequency (HPF) model is analyzed using Bayesian inference. Non-stationary HPF curves were applied to a total of 16 future daily maximum surface air temperature ensembles projected at six major sites in Korea. From the ensemble of future climate information, it was found that the intensity of the extreme heat wave event for persistence day of 2-day estimated in 2050 would be more likely to rise in the range of 1.23–1.69 °C (under RCP 4.5) and 1.15–1.96 °C (under RCP 8.5) than that estimated in 2010. It was also found that the uncertainty resulting from parameter estimation of the HPF model was greater than the uncertainty resulting from the inter-model variability of various climate model combinations. When reflecting the uncertainty resulting from the estimation of model parameters, the 95% confidence interval of the delta change for the heat wave intensity projected in 2050 from 2010 was estimated to be 0.53–4.94 °C (under RCP 4.5) and 0.89–5.57 °C (under RCP 8.5).

Are Kenya Meteorological Department heavy rainfall advisories useful for forecast-based early action and early preparedness for flooding?

Abstract. Preparedness saves lives. Forecasts can help improve preparedness by triggering early actions as part of pre-defined protocols under the Forecast-based Financing (FbF) approach; however it is essential to understand the skill of a forecast before using it as a trigger. In order to support the development of early-action protocols over Kenya, we evaluate the 33 heavy rainfall advisories (HRAs) issued by the Kenya Meteorological Department (KMD) during 2015–2019. The majority of HRAs warn counties which subsequently receive heavy rainfall within the forecast window. We also find a significant improvement in the advisory ability to anticipate flood events over time, with particularly high levels of skill in recent years. For instance actions with a 2-week lifetime based on advisories issued in 2015 and 2016 would have failed to anticipate nearly all recorded flood events in that period, whilst actions in 2019 would have anticipated over 70 % of the instances of flooding at the county level. When compared against the most significant flood events over the period which led to significant loss of life, all three such periods during 2018 and 2019 were preceded by HRAs, and in these cases the advisories accurately warned the specific counties for which significant impacts were recorded. By contrast none of the four significant flooding events in 2015–2017 were preceded by advisories. This step change in skill may be due to developing forecaster experience with synoptic patterns associated with extremes as well as access to new dynamical prediction tools that specifically address extreme event probability; for example, KMD access to the UK Met Office Global Hazard Map was introduced at the end of 2017. Overall we find that KMD HRAs effectively warn of heavy rainfall and flooding and can be a vital source of information for early preparedness. However a lack of spatial detail on flood impacts and broad probability ranges limit their utility for systematic FbF approaches. We conclude with suggestions for making the HRAs more useful for FbF and outline the developing approach to flood forecasting in Kenya.

Strong regional influence of climatic forcing datasets on global crop model ensembles

We present results from the Agricultural Model Intercomparison and Improvement Project (AgMIP) Global Gridded Crop Model Intercomparison (GGCMI) Phase I, which aligned 14 global gridded crop models (GGCMs) and 11 climatic forcing datasets (CFDs) in order to understand how the selection of climate data affects simulated historical crop productivity of maize, wheat, rice and soybean. Results show that CFDs demonstrate mean biases and differences in the probability of extreme events, with larger uncertainty around extreme precipitation and in regions where observational data for climate and crop systems are scarce. Countries where simulations correlate highly with reported FAO national production anomalies tend to have high correlations across most CFDs, whose influence we isolate using multi-GGCM ensembles for each CFD. Correlations compare favorably with the climate signal detected in other studies, although production in many countries is not primarily climate-limited (particularly for rice). Bias-adjusted CFDs most often were among the highest model-observation correlations, although all CFDs produced the highest correlation in at least one top-producing country. Analysis of larger multi-CFD-multi-GGCM ensembles (up to 91 members) shows benefits over the use of smaller subset of models in some regions and farming systems, although bigger is not always better. Our analysis suggests that global assessments should prioritize ensembles based on multiple crop models over multiple CFDs as long as a top-performing CFD is utilized for the focus region.