High-resolution microclimatic grids for the Bohemian Forest Ecosystem

Here, we provide spatially continuous, high-resolution (5 m) microclimate grids covering all 923 km² of the Bohemian Forest Ecosystem (BFE), i.e. the complete area of the Šumava (Czech Republic) and Bavarian Forest (Germany) National Parks.

To derive these grids, we established a dense network of 288 microclimatic sites that continuously measured air, near-ground, and soil temperature every 15 minutes from 12 October 2019 to 11 October 2020. We used Tomst Thermologgers and TMS-4 loggers equipped with the same MAXIM/DALLAS Semiconductor DS7505U+ thermometers, which have an accuracy of ± 0.5 °C and a resolution of 0.0625 °C. 

We combined the measured microclimate temperature with LiDAR derived land surface topography and forest structure through boosted spatial generalized additive models (GAMs).

We validated the resulting microclimatic grids with an independent network of forest weather stations and compared these microclimatic grids with SoilTemp (soil temperature), ForestTemp (near-ground forest understorey temperature), and downscaled ERA5-Land (air temperature). The developed BFE microclimatic grids were closer to independently measured temperatures than any other alternative and captured high microclimatic variability controlled jointly by land surface topography and forest structure. 

Our microclimatic grids represent accurate, high-resolution spatial variation of mean annual soil temperature, mean, maximum, and minimum air temperature at two heights, and growing degree days at 200 cm. 

The dataset contains 9 microclimatic grids (GeoTIFF format, coordinate system EPSG 31468, resolution 5 m).

Extent: 4587063, 5399139: 4646023, 5451289

The name of the file is “name.tif”, where "name" represents the abbreviation of the microclimatic variable (see names below).

The values are in °C (°C d for GDD), and the data can be readily imported into standard geographical information system software (e.g., QGIS) or accessed in statistical software (e.g., R). The datasets do not include colour schemes.

Measured variable (depth/height)  
 - Microclimatic variable    Abbreviation  (Units)

Soil temperature (-8 cm)             
 - Mean temperature = mean temperature          T.soil_8_cm.mean    (°C)
            
Near-ground air temperature (15 cm)    
    - Mean temperature = mean temperature       T.air_15_cm.mean     (°C)
    - Maximum temperature = 95^th percentile of daily maximum temperatures     T.air_15_cm.max.95p    (°C)
    - Minimum temperature = 5^th percentile of daily minimum temperatures        T.air_15_cm.min.5p    (°C)
    - Growing degree days = sum of degree days above base temperature (base 5°C)      T.air_15_cm.GDD5    (°C d)
            
Air temperature (200 cm)    

    - Mean temperature = mean temperature    T.air_200_cm.mean    (°C)
    - Maximum temperature = 95^th percentile of daily maximum temperatures     T.air_200_cm.max.95p    (°C)
    - Minimum temperature = 5^th percentile of daily minimum temperatures     T.air_200_cm.min.5p    (°C)
    - Growing degree days = sum of degree days above base temperature (base 5°C)      T.air_200_cm.GDD5    (°C d)

BFE_script_and_data.zip contains

     - Microclimatic variables, topography and forest structure variables for all stations used for modelling (BFE_data2.RData). It includes the dataframe “zeta” with 62 variables for all 288 plots, SpatialPointsDataFrame “zeta.sp” with the spatial data of all plots and their unique ID “ID_lokalit” and a simple list with the names of predictor variables, “stack_names”. 

     - Script BFE_microclimate_maps_model_script_final.R used for statistical modelling and prediction.
     - microclimate2predict.csv - list of microclimate variables for prediction used in the script.

The data used for prediction cannot be made publicly available.  
The LIDAR-derived topography and forest structure rasters can be obtained upon request from the authors and with a licence from LIDAR data providers: Administration of Bavarian Forest National Park and Šumava National Park Administration.

Detailed description will be available in an associated journal article.

Version 2 includes improved rasters, the GeoTIFFs include an alpha channel for transparency of pixels with NA and the full script used for processing.

Version 3 includes updated models that remove artifacts around forest edges in some variables and further improve the reliability of the raster..
 - GDD at 15 cm was added
 - Uncertainty raster added 

The dataset is presneted in a Data descriptor paper: Brůna J., Macek M., Man M., Hederová L., Klinerová T., Moudrý V., Heurich M., Červenka J., Wild J. & Kopecký M. (2026) High-resolution microclimatic grids for the Bohemian Forest Ecosystem based on in situ measurements. Scientific Data 13: 246. https://doi.org/10.1038/s41597-026-06566-z