OBIA4RTM Demonstration Data

Published: 17 June 2019| Version 2 | DOI: 10.17632/vs55cwssyh.2


This dataset provides sample data demonstrating the capacities of the OBIA4RTM tool. OBIA4RTM combines radiative transfer modelling (RTM) of vegetation with object-based image analysis (OBIA). Its main purpose is to provide vegetation parameters such as Leaf Area Index (LAI) or leaf Chlorophyll a+b content (CAB) on a per-object rather than per pixel base. In this dataset, the OBIA4RTM tool was applied to two Sentinel-2 scenes covering an agricultural area in Southern Germany. Field parcels were used as image objects that were delineated from high-resolution ortho-photography and classified into vegetated and non-vegetated parcels using a Support Vector Machine trained on manually selected samples. For each of the two scenes - dating back on the 6th and 18th of July 2017 - the canopy RTM ProSAIL was run in forward mode and the synthetic spectra stored in a Lookup-Table (LUT). For parameter retrieval, the 5 closest matches between spectra in the LUT and a given observed satellite spectrum averaged per parcel were used. Matches were found in terms of the lowest Root Mean Squared Error (RMSE). The utilized vegetation parameterisation is provided additionally. The results include the Leaf Area Index (LAI), the Chlorophyll a+b content (CAB) of leaves and the fraction of brown leaves (Cbrown). In addition, the retrieval error in terms of RMSE is provided together with the average of the 5 best matching synthetic spectra in the LUT to a given object-based spectrum. This allows for evaluating the quality of the inversion results and enables user to further improve the results by applying a more appropiate vegetation parameterisation. The structure of the dataset (see below) is straightforward: - The "Field Parcels" folder contains an ESRI shapefile with the field parcels as well as the classification results for the two image acquisition dates - The "ProSAIL Parametersisation" directory provides the vegetation parameters used to run the ProSAIL model. - The actual results are stored as ESRI-shapefiles in "Retrieved Vegetation Parameters" folder containing the LAI, CAB, Fraction of brown leaves and the RMSE as well as inverted Sentinel-2 spectra - "Sentinel-2 data" contains the utilized Sentinel-2 data as GeoTiff clipped to the study area in Level-2A This information should allow for reproducing the results using the freely available base version of OBIA4RTM (for research and education) or within other software packages. All geodata is projected in UTM-Zone 32N, WGS-84.


Steps to reproduce

The Sentinel-2 data was downloaded from Copernicus Scientific Hub in Level-1C and corrected for atmospheric effects and resampled to a spatial resolution of 20m using Sen2Core. Afterwards, the data was clipped to the extent of the study area. Only 9 out of the Sentinel-2 bands were kept (see used_S2_bands.txt). For retrieving the reflectance per object (=field parcel) GDAL together with Python 3.7 was used to calculate zonal statistics per image object. The mean reflectance per band was stored on a per field parcel base and then used for conducting the inversion.


Remote Sensing, Radiative Transfer, Field Crops, Monitoring in Agriculture