Deep Neural Networks with Transfer Learning for Forest Variable Estimation Using Sentinel-2 Imagery in Boreal Forest

Abstract

Estimation of forest structural variables is essential to provide relevant insights for public and private stakeholders in forestry and environmental sectors. Airborne light detection and ranging (LiDAR) enables accurate forest inventory, but it is expensive for large area analyses. Continuously increasing volume of open Earth Observation (EO) imagery from high-resolution (<30 m) satellites together with modern machine learning algorithms provide new prospects for spaceborne large area forest inventory. In this study, we investigated the capability of Sentinel-2 (S2) image and metadata, topography data, and canopy height model (CHM), as well as their combinations, to predict growing stock volume with deep neural networks (DNN) in four forestry districts in Central Finland. We focused on investigating the relevance of different input features, the effect of DNN depth, the amount of training data, and the size of image data sampling window to model prediction performance. We also studied model transfer between different silvicultural districts in Finland, with the objective to minimize the amount of new field data needed. We used forest inventory data provided by the Finnish Forest Centre for model training and performance evaluation. Leaving out CHM features, the model using RGB and NIR bands, the imaging and sun angles, and topography features as additional predictive variables obtained the best plot level accuracy (RMSE% = 42.6%, |BIAS%| = 0.8%). We found 3×3 pixels to be the optimal size for the sampling window, and two to three hidden layer DNNs to produce the best results with relatively small improvement to single hidden layer networks. Including CHM features with S2 data and additional features led to reduced relative RMSE (RMSE% = 28.6–30.7%) but increased the absolute value of relative bias (|BIAS%| = 0.9–4.0%). Transfer learning was found to be beneficial mainly with training data sets containing less than 250 field plots. The performance differences of DNN and random forest models were marginal. Our results contribute to improved structural variable estimation performance in boreal forests with the proposed image sampling and input feature concept.

Boreal Forest Height Mapping using Sentinel-1 Time Series and improved LSTM model

“…Observed experimental results are encouraging further investigations and are generally in line with other reported studies in boreal forest biome [3], [7], [48], [49]. Obtained accuracies are notably higher than in several other studies using Sentinel-1 data or Sentinel-2 datasets or their combinations, and compare well versus earlier multisensor EO data studies [3], [20], [48]–[52]. Several datasets with high potential for forest variable retrieval, such as e.g.…”

“…In boreal region, studies on forest variable prediction using Sentinel-1 or Landsat data report prediction accuracies within the range of 35-60% rRMSE [3], [48], while proposed model utilizing Sentinel-1 time series data reached rRMSE as small as 18 %. Predictions obtained using traditional ML models were within the same accuracy range as in recently published studies using Sentinel-2 and Landsat [53], while predictions using different versions of LSTM models appeared more accurate.…”

Improved Semisupervised UNet Deep Learning Model for Forest Height Mapping With Satellite SAR and Optical Data

“…Obtained results compare favourably to previous studies on forest height prediction. In boreal region, reported forest height accuracies with Sentinel-2 and Landsat data were 35-60% RMSE [19], [49], while proposed DL model reached as small as 24 % RMSE on plot-level and 15.4% on standlevel. Obtained predictions with ML models and Sentinel-2 data are within the same accuracy range as in recent published studies using Sentinel-2 and Landsat [50], while our predictions using deep learning models are much more accurate.…”

“…Transfer learning entails adapting a pretrained model and fine-tuning it using available sample of training data. One if such approaches has been recently demonstrated with optical Sentinel-2 data over boreal forest in Finland [19]. Semi-supervised learning, in order to improve modeling, concentrates on improving the use of unlabeled training data.…”

Integrating the new age of bioeconomy and Industry 4.0 into biorefinery process design

“…For example, recent studies reported that satellite data and machine-learning techniques were used for forest classification, obtaining higher accuracy (90% to 91%) . Another study used ANN to predict growing stock volume in forested areas (Astola et al 2021). In addition, a recent study developed a method to prepare forest fire susceptibility mapping, considering numerous factors, e.g., altitude, rainfall, wind effect, temperature, slope aspect, distances to roads and settlements, land use, and soil type, among others.…”