Deep Learning for the Detection of Early Signs for Forest Damage Based on Satellite Imagery

Abstract

Abstract. We present an approach for detecting early signs for upcoming forest damages by training a Convolutional Neural Network (CNN) for the pixel-wise prediction of the remaining life-time (RLT) of trees in forests based on Sentinel-2 imagery. We focus on a scenario in which reference data are only available for a related task, namely for a bi-temporal pixel-wise classification of forest degradation. This reference is used to train a CNN for the pixel-wise prediction of forest degradation. In this context, we propose a new sub-sampling-based approach for compensating the effects of a heavy class imbalance in the training data. Using the resulting classification model, we predict semi-labels for images of a Sentinel-2 time series, from which training data for a CNN designed to regress the RLT can be derived after some label cleansing. However, due to data gaps in the time series, e.g. caused by clouds, only intervals can be derived for the target variable to be regressed, and for some training pixels one of the interval limits may even be unknown. Consequently, we propose a new loss function for training a CNN for regressing the RLT that only requires the known interval limits. The method is evaluated on a data set in Germany, covering a time-span of 5 years. We show that the proposed sub-sampling strategy for dealing with strong label imbalance when training the classifier significantly reduces the training time compared to other approaches. We further show that our model predicts the RLT with a maximum error of two months for 80% of the forest pixels that die within one year from the acquisition date of the Sentinel-2 image.

Earth-Observation-Based Monitoring of Forests in Germany—Recent Progress and Research Frontiers: A Review

“…The category of forest disturbance comprised 36 of 84 (43%) papers in this review [3,6,10,11,46,48,49,52,53,58,59,62,63,67,68,72,79,84,88,91,[94][95][96][97][98][99]103,110,111,[115][116][117][118]120] compared to 28 of 166 papers (17%) in our initial review [1]. Disturbances to forest systems include causes and effect; therefore, within this category of studies, broad themes emerged.…”

“…Regarding analysis, statistical methods were quite broad. Here, we noted an equal number of papers utilizing deep learning methods [53,88,91,111,115] and random forest [59,[97][98][99]116].…”

“…In addition to the further development of instruments, the use of artificial intelligence (e.g., convolutional neural networks) is becoming increasingly common, e.g., for the segmentation of deadwood objects from UAV data [115] or the detection of forest damage from satellite data [91,111]. For the training of a deep learning model, extensive, high-quality reference data are indispensable.…”