Overview of Climate Data
Please add your comments on this section in this Google Doc.
Sources of Climate Data
Generally, climate data can be sourced from historical observation of climate using instrumentation or from climate model simulations. In the context of electricity system planning and design, future climatic conditions should be of primary interest, but observed historical climate data and products derived thereof may also be employed or used in the development of adequate future climate information.
Observed data is provided as point information at station locations. To provide continuous fields of climate information, station data can be interpolated spatially to produce gridded historical data. They provide a portrait of climate beyond the stations point locations but are impacted by potential interpolation errors, particularly in data sparce regions, or in regions of complex topography (e.g. mountainous areas). The approach that attempts to overcome these drawbacks is the reproduction of historical climate through the combination of a maximum of available climate information with a physical climate simulation approach known as climate reanalysis. Reanalysis products have substantially improved over recent years and are often used as surrogates for observational data in climate science and studies.
Observational data and reanalysis data provide a representation of the actual evolution and sequence of events as they occurred and provide robust estimates of climate when averaged over climatic periods (usually 30 years). Note, however, that due to natural climate variability, climate estimates would shift with slight temporal shifts of the period, or the data being used.
Simulated data from climate models used for electricity system planning and design should be sourced from internationally coordinated ensembles of climate model simulations or products derived thereof. These simulations typically cover multiple decades and are available for historical and future periods. They are fully consistent in time and space and are provided on grids. The grid resolution depends on the model or the data product.
Like observational data, bias-adjusted historical climate simulations may be used to derive climate estimates of historical climate. These estimates will not be identical to climate estimates from observed data yet will generally fall into the range of natural variability of the observed climate. Since climate model simulations extend into the future, estimates for future climate may also be derived. It is important to note that the sequence of events produced by a historical climate simulation is distinct from the sequence of historically observed events, although their respective climate estimate is the robust characteristic they have in common.
Summary of Types of Climate Datasets
The datasets discussed here can be grouped into the following categories:
Observational data are direct measurements from weather stations. These provide the most accurate and high-resolution temperature records, but their coverage is often limited to specific locations, leading to gaps in certain regions, particularly oceans and remote areas. Temporal gaps in the records may also be found in station data.
Gridded observational datasets address the spatial limitation of station data by interpolating their data across a defined grid, providing more comprehensive spatial coverage. These datasets, such as those produced by national meteorological agencies, offer a balance between accuracy and spatial representation, though uncertainties arise in data-sparse regions.
Reanalysis datasets combine historical observations with climate models to create consistent, long-term reconstructions of the atmosphere. They offer global coverage and high temporal resolution, making them valuable for analyzing past climate conditions and trends. However, reanalysis products rely on model-based data assimilation, which may introduce biases, especially in areas with limited observations.
Climate model data, often from global climate models (GCMs), simulate temperature under different greenhouse gas scenarios. These projections are essential for understanding future climate patterns and assessing potential impacts. While models provide large-scale trends, they lack the precision of observational data due to their coarse spatial resolution and inherent uncertainties in modeling processes.
Together, these datasets enable a comprehensive view of both historical and future climate conditions, supporting a wide range of applications in the electricity sector.
(from Demand)
Climate data
Climate data is essential for the electricity demand impact assessments under climate change. These data are primarily sourced from Global Climate Models (GCMs), which simulate the Earth’s climate system based on physical laws and boundary conditions. GCMs provide long-term projections of atmospheric variables such as temperature and precipitation, which are critical inputs to electricity demand models, especially in studies focusing on temperatrue-sensitive loads like heating and cooling.
Many studies assessing future electricity demand have employed GCM outputs from the Coupled Model Intercomparison Project phases, particularly CMIP5 and CMIP6. For example, Fonseca et al. (2019) temperature projections from 20 CMIP5 GCMs to simulate intra-day electricity demand shifts. Similarly, Chidiac et al. (2022) relied on GCMs from the North American CORDEX, which has 23 CMIP6 to assess building heating and cooling demand in Ontario, while Hu et al. (2024) drew on CMIP5 from the EURO-CORDEX to construct the temperature response function. Amonkar et al. (2023) used ERA5 reanalysis data to analyze changes in peak and average thermal demand across the contiguous US. To explore a range of potential future emissions trajectories, studies commonly apply Representative Concentration Pathways (RCPs). These scenarios (i.e., RCP2.6, RCP4.5, RCP6.0, and RCP8.5) represent different greenhouse gas concentration trajectories based on varying degrees of pathway scenarios. The selection of RCPs provides a structured way to assess electricity demand sensitivity under different warming levels. For instance, Hu et al. (2024) employed three major RCPs (2.6, 4.5, and 8.5) to examine how demand profiles shift across Europe. Emodi et al. (2018) and Reyna and Chester (2017) similarly applied four RCPs including RCP6.0 to assess electricity demand responses under climate change, respectively. While Trotter et al. (2016) and Auffhammer et al. (2017) applied two RCPs (4.5 and 8.5), Lipson et al. (2019) only used the worst pathway, RCP8.5, to investigate climate change impact on energy demand through building-urban-atmosphere simulations. On the other hand, Chidiac et al. (2022) evaluated changes of HDDs and CDDs for buildings’ heating and cooling demand under the intermediate scenario, RCP4.5.
Despite the advances in climate data usage, the literature reveals several challenges and limitations. First, the choice of climate models and RCPs is not always clearly justified or consistent across studies, making comparative assessment difficult. Second, biases in raw climate model outputs can distort demand projections if not corrected or validated using observations. Third, the integration of climate data with energy system models often lacks explicit treatment of uncertainty. Fonseca et al. (2019) noted that climate variables alone explain a limited portion of electricity demand variability as socioeconomic, behavioral, and technological factors may play considerable roles. Additionally, while many studies have applied a single or limited number of climate models, ensemble approaches or various combinations of GCMs and RCMs, as used by Reyna and Chester (2017), Auffhammer et al. (2017), Lipson et al. (2019), Fonseca et al. (2019), Chidiac et al. (2022), Lee and Dessler (2022), and Bonkaney et al. (2023), are therefore essential to improve the robustness of results.
(from Demand)
Downscaling methods
Given that GCMs operate at relatively coarse spatial resolutions (typically 100-250 km), their outputs are often inadequate for regional scale modeling. Because electricity planning is typically sensitive to finer-scale temperature variations, it is essential that climate inputs local meteorological conditions with sufficient detail. To address this, two primary categories of downscaling techniques are widely applied in the literature: statistical downscaling and dynamical downscaling.
Statistical downscaling establishes empirical relationships between large-scale GCM variables and local climate observations, using historical data as a training base. These methods are computationally efficient and have been widely used where computational resources or high-resolution models are limited. For example, Berardi and Jafarpur (2020) employed CCWorldWeatherGen and WeatherShift™ tools to statistically downscale GCMs and generated site-specific future weather data in Canada. Similarly, Fonseca et al. (2019) and Auffhammer et al. (2017) applied the Multivariate Adaptive Constructed Analogs (MACA) method to statistically downscale CMIP5 GCM projections for United States. The resulting high-resolution weather data were used to simulate changes in electricity demand profiles under a warming climate. Trotter et al. (2016) also used trend-preserving statistical downscaling method to estimate climate-induced changes across Brazil.
In contrast, dynamical downscaling embeds higher-resolution Regional Climate Models (RCMs) within GCMs to explicitly simulate regional climate processes. This approach is particularly beneficial for capturing terrain-driven weather dynamics and localized extremes. Within the North American COPRDEX (NA-CORDEX), Chidiac et al. (2022) applied multiple RCMs including CRCM5, RCA4, RegCM4, WRF, CanRCM4, and HIRHAM5 driven by GCMs like HadGEM2-ES, CanESM2, MPI-ESM-LR, MPI-ESM-MR. EC-EARTH, and GFDL-ESM2M, to project changes in building-level heating and cooling demands in Ontario. Similarly, Hu et al. (2024) used the ERMO2009 RCM under the EURO-CORDEX project to model electricity demand changes across European countries, providing spatially disaggregated insights into future temperature-response dynamics. These dynamical approaches offer more physically consistent outputs than statistical methods while a higher computational cost is required.
Nevertheless, dynamical downscaling is not without limitations. Bias correction is typically necessary because RCM outputs may deviate from observed climatology. Moreover, the selection of GCM-RCM combinations can introduce additional variability into the projections, as a concern emphasized by Chilkoti et al. (2017). To address this, ensemble approaches using multiple model combinations have become more common to capture the spread of plausible futures and quantify projection uncertainty, although this adds additional computational costs.