An ensemble climate projection for Africa

Outputs from climate models are the primary data source in climate change impact studies. However, their interpretation is not straightforward. In recent years, several methods have been developed in order to quantify the uncertainty in climate projections. One of the common assumptions in almost all these methods is that the climate models are independent. This study addresses the validity of this assumption for two ensembles of regional climate models (RCMs) from the Ensemble-Based Predictions of Climate Changes and their Impacts (ENSEMBLES) project based on the land cells covering Denmark. Daily precipitation indices from an ensemble of RCMs driven by the 40-yr ECMWF Re-Analysis (ERA-40) and an ensemble of the same RCMs driven by different general circulation models (GCMs) are analyzed. Two different methods are used to estimate the amount of independent information in the ensembles. These are based on different statistical properties of a measure of climate model error. Additionally, a hierarchical cluster analysis is carried out. Regardless of the method used, the effective number of RCMs is smaller than the total number of RCMs. The estimated effective number of RCMs varies depending on the method and precipitation index considered. The results also show that the main cause of interdependency in the ensemble is the use of the same RCM driven by different GCMs. This study shows that the precipitation outputs from the RCMs in the ENSEMBLES project cannot be considered independent. If the interdependency between RCMs is not taken into account, the uncertainty in the RCM simulations of current regional climate may be underestimated. This will in turn lead to an underestimation of the uncertainty in future precipitation projections.

Regional climate models (RCMs) are dynamical downscaling tools aimed to improve the modelling of local physical processes. Ensembles of RCMs are widely used to improve the coarse-grain estimates of global climate models (GCMs) since the use of several RCMs helps to palliate uncertainties arising from different dynamical cores and numerical schemes methods. In this paper, we analyse the differences and similarities in the climate change response for an ensemble of heterogeneous RCMs forced by one GCM (HadAM3H), and one emissions scenario (IPCC’s SRES-A2 scenario). As a difference with previous approaches using PRUDENCE database, the statistical description of climate characteristics is made through the spatial and temporal aggregation of the RCMs outputs into probability distribution functions (PDF) of monthly values. This procedure is a complementary approach to conventional seasonal analyses. Our results provide new, stronger evidence on expected marked regional differences in Europe in the A2 scenario in terms of precipitation and temperature changes. While we found an overall increase in the mean temperature and extreme values, we also found mixed regional differences for precipitation.

The study examines how regional climate models (RCMs) reproduce the diurnal temperature range (DTR) in their control simulations over Central Europe. We evaluate 30-year runs driven by perfect boundary conditions (the ERA40 reanalysis, 1961–1990) and a global climate model (ECHAM5) of an ensemble of RCMs with 25-km resolution from the ENSEMBLES project. The RCMs’ performance is compared against the dataset gridded from a high-density stations network. We find that all RCMs underestimate DTR in all seasons, notwithstanding whether driven by ERA40 or ECHAM5. Underestimation is largest in summer and smallest in winter in most RCMs. The relationship of the models’ errors to indices of atmospheric circulation and cloud cover is discussed to reveal possible causes of the biases. In all seasons and all simulations driven by ERA40 and ECHAM5, underestimation of DTR is larger under anticyclonic circulation and becomes smaller or negligible for cyclonic circulation. In summer and transition seasons, underestimation tends to be largest for the southeast to south flow associated with warm advection, while in winter it does not depend on flow direction. We show that the biases in DTR, which seem common to all examined RCMs, are also related to cloud cover simulation. However, there is no general tendency to overestimate total cloud amount under anticyclonic conditions in the RCMs, which suggests the large negative bias in DTR for anticyclonic circulation cannot be explained by a bias in cloudiness. Errors in simulating heat and moisture fluxes between land surface and atmosphere probably contribute to the biases in DTR as well.

Snow is important for many physical, social, and economic sectors in North America. In a warming climate, the characteristics of snow will likely change in fundamental ways, therefore compelling societal need for future projections of snow. However, many stakeholders require climate change information at finer resolutions that global climate models (GCMs) can provide. The North American Coordinated Regional Downscaling Experiment (NA-CORDEX) provides an ensemble of regional climate model (RCMs) simulations at two resolutions (~ 0.5° and ~ 0.25°) designed to help serve the climate impacts and adaptation communities. This is the first study to examine the differences in end of twenty-first-century projections of snow from the NA-CORDEX RCMs and their driving GCMs. We find that the broad patterns of change are similar across RCMs and GCMs: snow cover retreats, snow mass decreases everywhere except at high latitudes, and the duration of the snow covered season decreases. Regionally, the spatial details, magnitude, percent, and uncertainty of future changes vary between the GCM and RCM ensemble but are similar between the two resolutions of the RCM ensembles. An increase in winter snow amounts at high latitudes is a robust response across all ensembles. Percent snow losses are found to be more substantial in the GCMs than the RCMs over most of North America, especially in regions with high-elevation topography. Specifically, percent snow losses decrease with increasing elevation as the model resolution becomes finer.

This study analyses the length and onset of the four seasons based on the annual climatic cycle of maximum and minimum temperatures. Previous studies focused over climatically homogeneous mid-high latitude areas, employing fixed temperature thresholds (related to climatic features such as freezing point) that can be inadequate when different climate conditions are present. We propose a method related to the daily minimum and maximum temperature 25th and 75th point-dependent climatic percentiles. It is applied to an ensemble of regional climate models (RCMs) of 25-km horizontal resolution over the peninsular Spain and Balearic Islands, where a large variety of climatic regimes, from alpine to semi-desertic conditions, are present. First, baseline climate (1961–2000) ERA40-forced RCM simulations are successfully compared with the Spain02 daily observational database, following astronomical season length (around 90 days). This result confirms the validity of the proposed method and capability of the RCMs to describe the seasonal features. Future climate global climate model-forced RCMs (2071–2100) compared with present climate (1961–1990) simulations indicate the disappearance of winter season, a summer enlargement (onset and end) and a slight spring and autumn increase.

ABSTRACTWe consider the impact of climate change on the wind energy resource of Ireland using an ensemble of regional climate model (RCM) simulations. The RCM used in this work is the Consortium for Small‐scale Modelling–climate limited‐area modelling (COSMO‐CLM) model.The COSMO‐CLM model was evaluated by performing simulations of the past Irish climate, driven by European Centre for Medium‐Range Weather Forecasts ERA‐40 data, and comparing the output with observations. For the investigation of the influence of the future climate under different climate scenarios, the Max Planck Institute's global climate model, ECHAM5, was used to drive the COSMO‐CLM model. Simulations are run for a control period 1961–2000 and future period 2021–2060. To add to the number of ensemble members, the control and future simulations were driven by different realizations of the ECHAM5 data. The future climate was simulated using the Intergovernmental Panel on Climate Change emission scenarios, A1B and B1.The research was undertaken to consolidate, and as a continuation of, similar research using the Rossby Centre's RCA3 RCM to investigate the effects of climate change on the future wind energy resource of Ireland. The COSMO‐CLM projections outlined in this study agree with the RCA3 projections, with both showing substantial increases in 60 m wind speed over Ireland during winter and decreases during summer. The projected changes of both studies were found to be statistically significant over most of Ireland. The agreement of the COSMO‐CLM and RCA3 simulation results increases our confidence in the robustness of the projections. Copyright © 2012 John Wiley & Sons, Ltd.

The Mediterranean basin is a region characterized by its vulnerability to changes in the water cycle. Hence, the impact of global warming on the water resources in the Mediterranean zone is one of the major concerns for the scientific community. The future climate projections used to elaborate the IPCC report of 2007 show great alterations in the evaporation and precipitation over the Mediterranean Sea at the end of 21st century. In this work we investigate the changes in the Mediterranean Sea water budget by using SRES‐A1B scenario experiments performed with high resolution (25 km) regional climate models (RCMs). The RCMs provide good estimates of the water budget components, in particular with a significant improvement of the runoff and Black Sea discharge terms compared to the coarser resolution general circulation models (GCMs) used in the last Intergovernmental Panel of Climate Change (IPCC) report. As for the case of GCMs, the RCMs show that the Mediterranean water budget is likely to be significantly altered at the end of 21st century. The response of the hydrological variables to global warming starts to be statistically significant from 2050, though some alterations are already observed before 2050. The RCMs predict an increase of evaporation, and a decrease of precipitation, and river and Black Sea discharge, yielding to a large increase of the Mediterranean fresh water deficit. The freshwater deficit for the period 2070–2099 related to 1950–1999 presents a mean increase of +40% for both RCMs and GCMs.

Ensembles of high‐resolution regional climate model ( RCM ) simulations are crucial for assessing regional climate change and the associated uncertainties. This article presents an RCM ensemble generation technique which explores uncertainties arising from the positioning of synoptic systems in the large‐scale atmospheric forcing by shifting the atmospheric fields from global climate model ( GCM ) runs horizontally. Here, we discuss how the so‐called Atmospheric Forcing Shifting ( AFS ) affects temperature and precipitation over Europe for the period 1980–1984. We use ERA ‐40 reanalysis data in which the atmospheric fields are shifted to each direction by 25 and 50 km, respectively, to run RCM simulations with COSMO‐CLM at 50‐km resolution. The analysis of the AFS ensemble includes comparisons with E‐ OBS observations and COSMO‐CLM runs driven by different GCMs . AFS has an evident effect on the spatiotemporal distributions of temperature and particularly precipitation, which is most pronounced during hydrological summer (May to October) when spatial weather patterns are more variable. Furthermore, AFS produces realistic changes in the likelihood and intensity of extreme precipitation. The changes induced by AFS depend strongly on orography, i.e. precipitation increases are likely to occur where moist air masses are shifted towards the windward mountain side and vice versa. Thus, increasing the RCM ensemble spread by means of AFS is a simple and useful method for sampling observed climate statistics and assessing the variability and changes in mean and extreme climate.

The study examines links of summer heat waves and winter cold spells in Central Europe to atmospheric circulation and specifically its persistence in an ensemble of regional climate models (RCMs). We analyse 13 RCMs driven by the ERA-40 reanalysis and compare them against observations over reference period 1971–2000. Using objective classification of circulation types and an efficiency coefficient with a block resampling test, we identify circulation types significantly conducive to heat waves and cold spells. We show that the RCMs have a stronger tendency to group together days with very high or low temperature and tend to simulate too many heat waves and cold spells, especially those lasting 5 days and more. Circulation types conducive to heat waves in summer are characterized by anticyclonic, southerly and easterly flow, with increasing importance of warm advection during heat waves. Winter cold spells are typically associated with easterly and anticyclonic flow, and the onset of cold spells tends to be linked to northerly and cyclonic flow with cold advection. The RCMs are generally able to reproduce the links between circulation and heat waves or cold spells, including the radiation-to-advection effect for heat waves and the opposite advection-to-radiation effect for cold spells. They capture relatively well also changes of mean temperature anomalies during sequences of given circulation types, namely the tendency towards temperature increase (decrease) during those types conducive to heat waves (cold spells). Since mean lengths of all circulation supertypes are overestimated in the RCMs, we conclude that the overly persistent circulation in climate models contributes to the overestimated frequency of long heat waves and cold spells. As these biases are rather general among the examined RCMs and similar drawbacks are likely to be manifested in climate model simulations for the twenty-first century, the results also suggest that climate change scenarios for spells of days with high or low temperatures should be interpreted with a view to the models’ limitations in simulating the recent climate.

Abstract. Including the impacts of climate change in decision making and planning processes is a challenge facing many regional governments including the New South Wales (NSW) and Australian Capital Territory (ACT) governments in Australia. NARCliM (NSW/ACT Regional Climate Modelling project) is a regional climate modelling project that aims to provide a comprehensive and consistent set of climate projections that can be used by all relevant government departments when considering climate change. To maximise end user engagement and ensure outputs are relevant to the planning process, a series of stakeholder workshops were run to define key aspects of the model experiment including spatial resolution, time slices, and output variables. As with all such experiments, practical considerations limit the number of ensemble members that can be simulated such that choices must be made concerning which global climate models (GCMs) to downscale from, and which regional climate models (RCMs) to downscale with. Here a methodology for making these choices is proposed that aims to sample the uncertainty in both GCM and RCM ensembles, as well as spanning the range of future climate projections present in the GCM ensemble. The RCM selection process uses performance evaluation metrics to eliminate poor performing models from consideration, followed by explicit consideration of model independence in order to retain as much information as possible in a small model subset. In addition to these two steps the GCM selection process also considers the future change in temperature and precipitation projected by each GCM. The final GCM selection is based on a subjective consideration of the GCM independence and future change. The created ensemble provides a more robust view of future regional climate changes. Future research is required to determine objective criteria that could replace the subjective aspects of the selection process.

Projected future regional climate changes in Luxembourg are assessed based on a six-member ensemble of regional climate models (RCM) from the ENSEMBLES project. The key aspects are projected changes in air temperature and their impacts on vegetation. Up to now, there have been only few assessments of future climate conditions for Luxembourg. As agriculture is the dominant land use in Luxembourg, possible effects on crops and vegetation in general are highly relevant. Different RCMs at 25 km spatial and a daily temporal resolution, ranging from 1961 to 2100 based on the SRES A1B emission scenario are used. To reduce systematic biases in the RCM-derived time series, a bias correction is applied. Multi-model annual mean temperatures are projected to increase by 3.1 °C between the reference time span (1961 to 1990) and the far future (2069 to 2098). Clear change signals are found in seasonal bivariate frequency distributions of air temperature and precipitation. Derived impacts are an elongation of the thermal vegetation period by 6.2 days per decade due to an earlier onset in spring; growing degree day sums show a substantial increase leading to potentially better growth conditions; the earlier onset of the vegetation period causes an increase in late frost risk, especially in the near future (2021 to 2050) projections compared to the reference period.

<p>We use an unprecedented ensemble of regional climate model (RCM) projections over seven regional CORDEX domains to provide, for the first time, an RCM-based global view of monsoon changes at various levels of increased greenhouse gas (GHG) forcing. All regional simulations are conducted using RegCM4 at a 25km horizontal grid spacing using lateral and lower boundary forcing from three General Circulation Models (GCMs), which are part of the fifth phase of the Coupled Model Inter-comparison Project (CMIP5). Each simulation covers the period from 1970 through 2100 under two Representative Concentration Pathways (RCP2.6 and RCP8.5). Regional climate simulations exhibit high fidelity in capturing key characteristics of precipitation and atmospheric dynamics across monsoon regions in the historical period. In the future period, regional monsoons exhibit a spatially robust delay in the monsoon onset, an increase in seasonality, and a reduction in the rainy season length at higher levels of radiative forcing. All regions with substantial delays in the monsoon onset exhibit a decrease in pre-monsoon precipitation, indicating a strong connection between pre-monsoon drying and a shift in the monsoon onset. The weakening of latent heat driven atmospheric warming during the pre-monsoon period delays the overturning of atmospheric subsidence in the monsoon regions, which defers their transitioning into deep convective states. Monsoon changes under the RCP2.6 scenario are mostly within the baseline variability. </p>

ABSTRACTHuman‐induced climate change is reshaping wind patterns across Canada, posing significant challenges for sectors such as wind energy and infrastructure planning. This study assesses the capability of regional climate models (RCMs) in simulating near‐surface wind speed (WS) across Canada by analysing outputs from various RCM ensembles, which downscale CMIP5 global climate model (GCM) output, including the NA‐CORDEX multi‐model ensemble (at 0.22° resolution) and the CanRCM4 single‐model large ensemble (at 0.44° resolution). These RCM outputs are compared against observational data, two reanalysis data sets (ERA5 and AgERA5), and GCM ensembles from CMIP5 and CMIP6. The evaluation examines the models' ability to replicate historical WS distributions, biases in mean and extreme WS, trends and temporal variability. The findings reveal that, despite the higher spatial resolution of RCMs, their added value over the GCM ensembles is limited, raising concerns about the reliability of RCM‐derived WS projections for climate services without further bias adjustment or statistical downscaling. The inability of both RCMs and GCMs to accurately simulate WS trends diminishes confidence in future WS projections, potentially leading to inadequate risk assessments and insufficient preparation for the impacts of climate change on vital sectors like energy and infrastructure.

Deep Learning (DL) has recently emerged as an innovative tool to downscale climate variables from large-scale atmospheric fields under the perfect prognosis (PP) approach. Different Convolutional Neural Networks (CNN) have been applied under present-day conditions with promising results, but little is known about their suitability for extrapolating future climate change conditions. Here, we analyze this problem from a multi-model perspective, developing and evaluating an ensemble of CNN-based downscaled projections (DeepESD) for temperature and precipitation over the European EUR-44i (0.5º) domain, based on eight GCMs from the Coupled Model Intercomparison Project Phase 5 (CMIP5). To our knowledge, this is the first time that CNNs have been used to produce multi-model ensembles of downscaled projections, allowing to quantify inter-model uncertainty in climate change signals. The results are compared with those corresponding to an EUR-44 ensemble of regional climate models (RCMs) showing that DeepESD reduces distributional biases in the historical period. Moreover, the resulting climate change signals are broadly comparable to those obtained with the RCMs, with similar spatial structures. As for the uncertainty of the climate change signal (measured on the basis of inter-model spread), DeepESD yields a smaller uncertainty for precipitation, but a similar uncertainty for temperature. To facilitate further studies of this downscaling approach we follow FAIR principles and make publicly available the code (a Jupyter notebook) and the DeepESD dataset. In particular, DeepESD is published at the Earth System Grid Federation (ESGF), as the first continental-wide PP dataset contributing to CORDEX (EUR-44).

Climate change impacts on phenology and yield of hazelnut in Australia