Meteorological data in the Solargis database are derived from numerical weather models. A summary of meteorological models from which Solargis sources data is given below.
|Climate Forecast System Reanalysis (CFSR)||Climate Forecast System (MERRA-2)||Climate Forecast System (CFSv2)||Global Forecast System (GFSprod)|
|Time period||1994 to 2010||1994 to 2010||2011 to D-2||D-1 to D+9|
|Original spatial resolution||0.312° by ~0.312°||0. 5° by ~0.5°||0.205° by ~0.205°||0.205° by ~0.205°|
|Original time resolution||1 hour||1 hour||1 hour||3 hours|
Given that the original spatial resolution of the meteorological data from these models is in the range of 0.2 to 0.5 degrees, the data characterize a wider geographic region rather than a specific site. The original spatial resolution of the models is enhanced to 1 km for air temperature and atmospheric pressure by spatial disaggregation and use of the Digital Elevation Model SRTM-3. The spatial resolution of other parameters remains unchanged.
If meteorological measurements are available from the vicinity of a solar project site, the measurements can be used for accuracy enhancement of the historical data derived from meteorological models.
Time step and spatial resolution of Solargis meteorological parameters
|Meteorological parameter||Acronym||Unit||Time resolution||Spatial representation||Data source(s)|
|Air temperature at 2 metres (dry bulb)||TEMP||°C||1 hour||1 km||MERRA-2, CFSv2, and GFSprod|
|Relative humidity||RH||%||1 hour||1 km||MERRA-2, CFSv2, and GFSprod|
|Atmospheric pressure||AP||hPa||1 hour||1 km||MERRA-2, CFSv2, and GFSprod|
|Wind speed at 10 metres||WS||m/s2||1 hour||Original model||MERRA-2, CFSv2, and GFSprod|
|Wind direction at 10 metres||WD||°||1 hour||Original model||MERRA-2, CFSv2, and GFSprod|
|Precipitable Water||PWAT||kg/m2||1 hour||Original model||CFSR, CFSv2, and GFSprod|
|Snow water equivalent||SWE||kg/m2||daily||Original model||CFSR, CFSv2, and GFSprod|
|Precipitation||PREC||kg/m2||1 hour||Original model||CFSR, CFSv2, and GFSprod|
|Module temperature||TMOD||°C||10/15/30 mins||250 m||MERRA-2, CFSv2, and GFSprod + Global Tilted Irradiation from Solargis solar resource model|
Temperature data derived from the meteorological models are post-processed by Solargis to achieve improved accuracy and homogeneity over long time period. Overview of the post-processing methods applied are given below.
The original spatial resolution of temperature data derived from numerical weather models is too coarse to accurately represent temperature in regions with variations in elevation. This problem can be overcome if vertical rate of change in temperature is known. This rate is known as lapse rate. The lapse rate can change with time and from one location to another, it is influenced by weather patterns and local micro-climatic and topographic features. Normally, the temperature decreases as the site's altitude increases. However, near the surface, occurrence of temperature inversion (temperature increase with increasing altitude) is not uncommon. This is the typical case over cold surfaces, for instance.
In addition to temperature at 2 meters over the surface, weather models also provide temperature data at every model layer from the surface to the top of the atmosphere. Using the vertical profiles of temperature, a simplified parameterization of the vertical change of temperature is devised. The lapse rate calculated is then used for spatial downscaling of temperature data to 1 km resolution.
Homogeneization of MERRA-2 and CFSv2 datasets
Historical archive of air temperature data in the Solargis database is derived from 2 numerical weather models – MERRA-2 (for period up to 2010) and CFSv2 (for period 2011 onwards). At many locations worldwide, there is a non-negligible jump between the MERRA-2 and the CFSv2 temperature trends which appears unreal. In order to achieve a homogenized time series of temperature data (from 1994 to present time) a pixel-wise linear transformation of the MERRA-2 data to match the CFSv2 values has been done. This has been accomplished using the MERRA2 dataset as an underlying reference since it also covers the period for which CFSv2 outputs are available.
Unbiasing of MERRA-2/CFSv2 data
Despite the corrections discussed above, temperature data still contains, in general, some biases. These biases can be seen as either local-scale biases or large-scale biases. The former are biases produced by local features in the surroundings of the site of interest (e.g., topography, proximity to water reservoirs...) that are not resolved by the NWP models. On the contrary, large-scale biases are present and systematic over wider areas, such as over countries or continents. These are most typically produced by modelling issues and to some extent have been corrected by Solargis.
Ground measured temperature data from reliable sources have been used as a reference to correct the large-scale biases in the CFSR/CFSv2 database. The local-scale biases can also be corrected by Solargis - however this is a time intensive exercise and therefore only done on request as part of site specific resource assessment studies (offered as a consultancy service by Solargis).