This is a python package that uses data from the Sentinel-2 (multispectral images) and the Sentinel-1 (synthetic apperture radar) satellite systems, along with data on vapour pressure deficit (VPD), in order to produce field-scale, continuous, weekly estimates of Leaf Area Index (LAI). Spatial data on climate variables (temperature and dew point), Sentinel-1 granules and Sentinel-2 images corresponding to the field and time-period of interest are provided to the algorithm. The algorithm splits every examined field into 25 bounding boxes (subfields). For each subfield and examined day the algorithm estimates the corresponding subfield-wide mean value for : (1) S1-VV/VH (10m, 2-3 day intervals), (2) S2-LAI (20-40m, irregular frequency due to cloudiness) and (3) VPD (5km, daily) . The algorithm uses 80% of the per-subfield dataset (all bounding boxes and examined dates) to train a Random Forest (RF) model that predicts LAI by using S1 VV, S1 VH, VPD and Day-of-Year as predictors. 20% of all the per-subfield data are used to validate the RF-predicted LAI (i.e. estimate R2 - coefficient of determination). Thereafter, the RF model is used to fill the gaps in the S2-based LAI weekly times-series. Each data point the final LAI time-series presents the field-mean LAI on the 1st day of every week of the examined period. The number of bounding boxes (subfields) and RF training/validation ratio (80%/20%) are hard-coded and they should be edited before package installation if necessary. The algorithm has been tested for managed grassland fields in the United Kingdom (UK). Its use for ecosystems other than grasslands and croplands is not recommended. Read the README.md file for information on system requirements and on how to install the package and implement the algorithm.
Myrgiotis, Vasilis. (2021). An algorithm that gap-fills Sentinel-2-based leaf area index (LAI) time-series using Sentinel-1-based backscatter and vapour pressure deficit (VPD) data, [software]. University of Edinburgh. School of GeoSciences. Global Change Ecology Lab. https://doi.org/10.7488/ds/3189.
|Date made available||26 Nov 2021|