The Use of Remotely Sensed Data and Polish NFI Plots for Prediction of Growing Stock Volume Using Different Predictive Methods

Abstract

Forest growing stock volume (GSV) is an important parameter in the context of forest resource management. National Forest Inventories (NFIs) are routinely used to estimate forest parameters, including GSV, for national or international reporting. Remotely sensed data are increasingly used as a source of auxiliary information for NFI data to improve the spatial precision of forest parameter estimates. In this study, we combine data from the NFI in Poland with satellite images of Landsat 7 and 3D point clouds collected with airborne laser scanning (ALS) technology to develop predictive models of GSV. We applied an area-based approach using 13,323 sample plots measured within the second cycle of the NFI in Poland (2010–2014) with poor positional accuracy from several to 15 m. Four different predictive approaches were evaluated: multiple linear regression, k-Nearest Neighbours, Random Forest and Deep Learning fully connected neural network. For each of these predictive methods, three sets of predictors were tested: ALS-derived, Landsat-derived and a combination of both. The developed models were validated at the stand level using field measurements from 360 reference forest stands. The best accuracy (RMSE% = 24.2%) and lowest systematic error (bias% = −2.2%) were obtained with a deep learning approach when both ALS- and Landsat-derived predictors were used. However, the differences between the evaluated predictive approaches were marginal when using the same set of predictor variables. Only a slight increase in model performance was observed when adding the Landsat-derived predictors to the ALS-derived ones. The obtained results showed that GSV can be predicted at the stand level with relatively low bias and reasonable accuracy for coniferous species, even using field sample plots with poor positional accuracy for model development. Our findings are especially important in the context of GSV prediction in areas where NFI data are available but the collection of accurate positions of field plots is not possible or justified because of economic reasons.

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

“…The learning rate was searched among values [0.00001, 0.0001, 0.0003, 0.0005, 0.001, 0.005]. For mini-batch size, two ranges of search values were used: [8,16,32,64,128] for the smallest sets of training data, and [32,64,128,256] for others. The number of training epochs was set to 250, which was considered adequate for convergence as the early stopping criterion interrupted training before this limit was reached.…”

“…A more sophisticated way would be to include the CHM at its original 1 m resolution or to use the ALS metrics from point clouds directly. This would most likely require a change to the DNN architecture, e.g., to select CNNs instead of fully connected dense layers [64,65]. Furthermore, as the number of trainable parameters would increase along with the number of predictive inputs, a larger amount of training data would probably be needed.…”

“…Additionally, in some other studies, deep learning models have not been outstandingly better than RF models. For instance, [64] reported that, in general, their deep learning method for growing stock volume prediction produced the best results, when compared to other methods (e.g., RF or k-NN), but that the differences were small. Using the Landsat and ALS predictors, they reported RMSE% = 24.16%, BIAS% = −2.2%, and R 2 = 0.38 with deep learning, and RMSE% = 25.18%, BIAS% = −5.32%, and R 2 = 0.39 with RF model.…”

Estimation of Boreal Forest Growing Stock Volume in Russia from Sentinel-2 MSI and Land Cover Classification

“…The simplest approaches are based on multispectral analysis of freely-available VNIR imagery having a spatial resolution of the order of 10 m or coarser [11][12][13][14][15][16][17]. Useful enrichment of the available feature space has been demonstrated using multitemporal datasets [18][19][20][21], incorporating texture measures [14,22] and field-derived or satellite-derived three-dimensional information [23][24][25][26][27][28][29][30]. Other approaches are based on the use of ultra-high-resolution VNIR imagery (usually not free of cost) [31,32], radar imagery [1,[33][34][35][36][37][38][39][40][41][42][43][44][45][46], or combinations of VNIR and radar imagery [47][48][49][50][51][52][53].…”

Operationalization of Remote Sensing Solutions for Sustainable Forest Management

“…The issue raised by Hawryło et al [5] could be used to explain the limited use of remotely sensed data in operational National Forest Inventories (NFIs) in many countries. That is, the NFIs often lack accurately georeferenced field plots.…”