Dataset for "Deep learning models for estimating volume and Lorey's height across Nordic countries using optical and SAR satellite images" article

This repository contains the data products and code required to reproduce the results presented in the article "Deep Learning Models for Estimating Volume and Lorey's Height Across Nordic Countries Using Optical and SAR Satellite Images". The study investigates the use of U-Net deep learning models for estimating forest volume and Lorey's height. The models were first pretrained using wall-to-wall forest resource maps derived from Airborne Laser Scanning (ALS) data in Finland and Norway and subsequently fine-tuned using National Forest Inventory (NFI) plots from Norway. The performance of the U-Net models was compared against baseline methods, including k-nearest neighbours (kNN) and XGBoost. This repository includes: Analysis_of_models_plot.R – code used for model-level accuracy assessment and result visualisation; Analysis_of_stands.R – code used for stand-level accuracy assessment; combined_predictions_df_UNet_kNN_XGB_warea.csv – the observed and predicted dataset used for the accuracy assessments presented in the article. In addition, we provide the trained U-Net model weights used in the study for both forest volume and Lorey's height estimation in the Finnish and Norwegian study areas.