High-resolution dynamical downscaling experiment outputs data over Reunion and Mauritius islands in the South-West Indian Ocean

The present article describes a dataset encompassing model outputs generated by the Weather Research and Forecasting (WRF) regional climate model. A high-resolution (1km) downscaling simulation was performed over two tropical islands, Reunion and Mauritius, situated in the South-West Indian Ocean (SWIO), with initial and boundary conditions provided by the ERA5 reanalysis with a global resolution of 0.25° × 0.25°. The simulation used three nested domains sequentially configured with spatial resolutions of 9, 3, and 1km, respectively, with a downscaling ratio of 3. The physical configurations of this simulation were determined through previous modeling studies and sensitivity tests. The published simulation data currently covers a period of 10 years, starting from 1991 (with the possibility to be extended to 30 years). Over 60 output variables were selected for publication with open access, including those related to the intermittent energy resources (e.g., surface solar radiation and its direct/diffuse components, wind speed/direction at multiple vertical levels, and precipitation, of interest for the run-off-river hydropower), as well as the widely used climatic/meteorological variables (e.g., temperature, pressure, humidity, etc.) at a temporal resolution varying from a day up to 30 minutes. All the data are available through an open-access data server, where an intelligent algorithm is applied to simplify the download process for data users. For the first time, a long-term, high-resolution climate/meteorological dataset covering Reunion and Mauritius has been simulated and published as open-access data, yielding substantial benefits to studies on climate modeling, weather forecasting, and even those related to climate change in the SWIO region. In particular, this dataset will enable a better understanding of the temporal and spatial characteristics of intermittent climate-related energy resources, consequently facilitating their implementation towards a green and low-carbon future.


a b s t r a c t
The present article describes a dataset encompassing model outputs generated by the Weather Research and Forecasting (WRF) regional climate model.A high-resolution (1km) downscaling simulation was performed over two tropical islands, Reunion and Mauritius, situated in the South-West Indian Ocean (SWIO), with initial and boundary conditions provided by the ERA5 reanalysis with a global resolution of 0.25 °× 0.25 °.The simulation used three nested domains sequentially configured with spatial resolutions of 9, 3, and 1km, respectively, with a downscaling ratio of 3. The physical configurations of this simulation were determined through previous modeling studies and sensitivity tests.The published simulation data currently covers a Dataset link: High-resolution dynamical downscaling experiment outputs data over Reunion and Mauritius islands in the South-West Indian Ocean (Original data)

Keywords:
Renewable energy resources Meteorological information Downscaling Regional climate modeling WRF Simulation output period of 10 years, starting from 1991 (with the possibility to be extended to 30 years).Over 60 output variables were selected for publication with open access, including those related to the intermittent energy resources (e.g., surface solar radiation and its direct/diffuse components, wind speed/direction at multiple vertical levels, and precipitation, of interest for the run-off-river hydropower), as well as the widely used climatic/meteorological variables (e.g., temperature, pressure, humidity, etc.) at a temporal resolution varying from a day up to 30 minutes.All the data are available through an open-access data server, where an intelligent algorithm is applied to simplify the download process for data users.For the first time, a longterm, high-resolution climate/meteorological dataset covering Reunion and Mauritius has been simulated and published as open-access data, yielding substantial benefits to studies on climate modeling, weather forecasting, and even those related to climate change in the SWIO region.In particular, this dataset will enable a better understanding of the temporal and spatial characteristics of intermittent climate-related energy resources, consequently facilitating their implementation towards a green and low-carbon future.
© 2023 The Author(s The value of this dataset can be summarized as (but not limited to) the following: • Weather forecasting: A large regional climate dataset helps in providing detailed information about climate patterns and trends at a regional and even local scale with extended temporal coverage, which is essential for accurate weather forecasting.• Climate change detection and adaptation: high-resolution climate datasets with long temporal coverage help to identify past climate change and its impacts at local scales, which is crucial for developing adaptation and mitigation strategies, especially for geographically isolated islands.• Physical process study: high-resolution data allows the analysis of fine-scale processes in climate dynamics, which can reveal new insights into how climate processes work and help to better understand the impacts of climate change at local scales.• Intermittent climate-related energy resources analysis: long-term and high-resolution data helps to get accurate knowledge of the quantity and spatiotemporal variability of climaterelated energy resources, such as solar, wind, and hydro (run-off-river) energy resources, at given locations.That will be valuable for the future production of decarbonized energy with a significant penetration of intermittent energy sources.• Data validation and evaluation: this dataset with various variables allows for comparison with other datasets, such as ground-based measurements, satellite observations, reanalysis, or outputs from other models, for validation or evaluation purposes.• Topography impact study: the simulation of two closed, isolated islands with similar climate conditions but contrasting topographies ( Fig. 2 ) allows to study the effects of topography on many aspects of climate, such as vegetation or local wind systems enforced by the topography.

Objective
The SWIO is a critical maritime zone subject to multiple climatic hazards, such as tropical cyclones, floods, and droughts, which profoundly impact the region's populations and ecosystems [1] .The WRF simulation presented in this article, with a spatial resolution of 1km, can advance climate-related research in the SWIO region by providing the first-ever high-resolution open-access climate dataset.This data can contribute to a better understanding of regional climate variability, extreme weather events [2] , and the regional impacts of global climate change [3] and facilitate assessments of renewable energy resources [4] for the energy transition in the region [ 5 , 6 ].

Data description
The dataset consists of selected output variables (see Table 1 ) from the Weather Research and Forecasting (WRF) regional climate model, restored in the format of Network Common Data Form (netCDF; [7] ), which is machine-independent, direct-access and self-describing [8] .Model outputs and corresponding metadata can be easily accessed via programming languages such as Python, R, etc., or graphic user interfaces such as Ncview ( http://meteora.ucsd.edu/∼ pierce/ncview _ home _ page.html ) or Panoply ( https://www.giss.nasa.gov/tools/panoply/).Spatially, the published output variables are either on a single surface level, being in 3 dimensions (time, longitude, latitude), or on the surface and 36 hybrid eta levels in the atmosphere, being in 4 dimensions (time, longitude, latitude, level).These eta levels are calculated by WRF (option AUTO_LEVELS_OPT = 2), with a considerable number of levels spanning the first kilometer of the atmosphere, a region of primary interest for wind energy installation.Temporally, these variables are stored at different frequencies, including 30-minute, hourly, 3-hourly, 6-hourly, and daily.All available variables are listed in Table 1 .The names of the WRF output files are made up of elements described above, including variable name, spatial resolution, driven model, and frequency, which are separated by underscores ('_') in the file name and appear in the following order: where VariableName corresponds to the variable names as in Table 1 ; Domain is "REU-MAU", and the resolution is "1km", denoting a domain covering Reunion and Mauritius at 1km spatial resolution; DrivenModel is the "ECMWF-ERA5-reanalysis"; RCMModel-VersionID is "WRF-v421"; Frequency can be one of the following: "30min", "hour", "3hr", "6hr", or "day"; StartTime and EndTime are in the format of "YYYYMMDDhhmm" in UTC; and the 'nc' suffix stands for the NetCDF format.As an example, an output file can be named as follow: _ 199201010 0 0 0 − 199212310 0 0 0 .nc ", From the example above, one could know this output file consists of data stored in NetCDF (nc) format of the variable of "2-meter air temperature" (T2) over a domain covering Reunion and Mauritius (REU-MAU) at 1km spatial resolution (1km), in a frequency of hour (hour) from 19920101 to 29921231, taken from a simulation by "WRF-v421" driven by "ECMWF-ERA5reanalysis".

Initial and boundary conditions and nested domains
The non-hydrostatic WRF Model has been used previously in Reunion to represent the general climate conditions [9] and even for extremes [10] .In the present experiment, the dynamical downscaling simulation was conducted with WRF version 4.2.1 [11] , driven by 6-hourly ERA5 reanalysis data [12] on pressure levels.Land-use categories were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS).Three downscaling domains were progressively nested in a one-way mode with increasing horizontal resolutions, as depicted in Fig. 1 .The firstguess domain, denoted as d01, which has a mesh size of 300 × 225 pixels in the longitudinal and latitudinal directions respectively, was interpolated by WRF using the ERA5 forcing data.The second domain (d02) has 240 × 181 pixels, and the third one (d03) has 360 × 270 pixels, focusing specifically on the study area, i.e., Reunion and Mauritius.

Physical configuration
The physical configuration of the WRF simulation has been chosen based on previous simulation studies [ 9 , 14-17 ] and sensitivity tests, in which surface air temperature, surface solar radiation (the energy resource of solar photovoltaic), and 10-meter wind speed and direction were compared with ground-based measurements from the French national meteorological service (i.e., Météo-France) in Reunion and from the Indian Ocean Solar Network (IOS-net, https://galilee.univ-reunion.fr/thredds/catalog.html ) in Mauritius.A seasonal comparison based on hourly outputs was performed for 2017, a year without substantial climate variability such as El Niño-Southern Oscillation or Indian Ocean Dipole to avoid the possible large-scale impact in Reunion [18] .The tested physical options are listed in Table 2 (i.e., the column headers), including the Planetary Boundary Layer, Cumulus, Microphysics, and Land Surface Model.The optimal physical configurations of the WRF simulation were then determined based on the results of these sensitivity tests (i.e., the best-performing physical schemes as depicted by a star symbol in each column of that table), along with other physical schemes applied in the simulation but not tested in the sensitivity study, such as the RRTMG scheme for Longwave and Shortwave ra-

Fig. 1 .
Fig. 1.Domain setting of the WRF simulation.Spatial resolution and domain sizes (number of pixels in longitudinal and latitudinal directions) are shown inside each domain.

Fig. 2 .
Fig. 2. Topography of Reunion (left) and Mauritius (right) in meters based on the ASTER Global Digital Elevation Model from NASA Jet Propulsion Laboratory [13] .The highest summit in Reunion, depicted by a red triangle, is the dormant Piton des Neiges volcano, peaking at over 30 0 0 m asl (above sea level) in the center (red triangle on the left plot).The other one is the active Piton de La Fournaise volcano at 2560 m asl in the east (green triangle on the left plot).Between these two lies a 1500 m-high plateau.On the other side, Mauritius has a relatively flat topography, with a summit measuring about 800 meters in elevation in the southwestern region (indicated on the plot by the red triangle).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Methodology, Software, Visualization, Formal analysis, Data curation, Writing -original draft, Writing -review & editing, Supervision, Resources, Project administration; Béatrice Morel: Conceptualization, Supervision, Resources, Funding acquisition, Writing -review & editing; Swati Singh: Methodology, Software, Formal analysis, Validation, Writingreview & editing; Alexandre Graillet: Data curation, Writing -original draft, Writing -review & editing; Julien Pergaud: Methodology, Software, Resources; Remy Ineza Mugenga: Methodology, Software, Data curation; Lwidjy Baraka: Data curation; Marie-Dominique Leroux: Data curation; Patrick Jeanty: Resources, Funding acquisition; Mathieu Delsaut: Data curation; Tyagaraja S.M. Cunden: Methodology, Software; Girish Kumar Beeharry: Methodology, Software; Roddy Lollchund: Methodology, Software.

Table 1
WRF output variables at 1km spatial resolution.