A bark beetle infestation predictive model based on satellite data in the frame of decision support system TANABBO

Abstract

Biogeosciences and Forestry Biogeosciences and Forestry A bark beetle infestation predictive model based on satellite data in the frame of decision support system TANABBO Renata Ďuračiová (1) , Milan Muňko (1) , Ivan Barka (2) , Milan Koreň (3) , Karolina Resnerová (4) , Jaroslav Holuša (4) , Miroslav Blaženec (5) , Mária Potterf (6) , Rastislav Jakuš (4-5) The European spruce bark beetle Ips typographus L. causes significant economic losses in managed coniferous forests in Central and Northern Europe. New infestations either occur in previously undisturbed forest stands (i.e., spot initiation) or depend on proximity to previous years' infestations (i.e., spot spreading). Early identification of newly infested trees over the forested landscape limits the effective control measures. Accurate forecasting of the spread of bark beetle infestation is crucial to plan efficient sanitation felling of infested trees and prevent further propagation of beetle-induced tree mortality. We created a predictive model of subsequent year spot initiation and spot spreading within the TANABBO decision support system. The algorithm combines open-access Landsat-based vegetation change time-series data, a digital terrain model, and forest stand characteristics. We validated predicted susceptibility to bark beetle attack (separately for spot initiation and spot spreading) against beetle infestations in managed forests in the Bohemian Forest in the Czech Republic (Central Europe) in yearly time steps from 2007 to 2010. The predictive models of susceptibility to bark beetle attack had a high degree of reliability (area under the ROC curve-AUC: 0.75-0.82). We conclude that spot initiation and spot spreading prediction modules included within the TANABBO model have the potential to help forest managers to plan sanitation felling in managed forests under pressure of bark beetle outbreak.

: An eXplainable Framework to Map Bark Beetle Infestation in Sentinel-2 Images

“…For example, the French Ministry of Agriculture and Food commissioned the inventory map of the bark beetle infestation hotspots observed in October 2018, to assess the damage in spruce forests of the northeast of France following the 2018 bark beetle outbreak. 1 This kind of inventory results, although not distinguishing the stage of the infestation, may also support locating undiscovered previous breeding sites and prompting the application of field monitoring in the vicinity of the detected hotspots, to identify recent infestations in the neighbour areas [6], [7].…”

“…The FCN-8 has also recently been adopted various remote sensing tasks (e.g., [40] and [41]). In this study, we used standard implementations of both U-Net 5 and FCN-8 6 . In both cases, we adopted the Tversky loss to address the issue of data imbalance [42].…”

Predicting the Susceptibility to Pityokteines curvidens Using GIS with AHP and MaxEnt Models in Fir Forests

“…An AUC score of 1 indicates a perfect prediction of the model. Five categories are used to categorize the AUC score: poor (0.5-0.6), medium (0.6-0.7), good (0.7-0.8), very good (0.8-0.9), and perfect (0.9-1.0) [33,42,45,[61][62][63][64]. For the development of the model, 75% of the infested stands were utilized as training data, while the remaining 25% were used as validation data.…”

Assessment of Machine Learning Algorithms for Modeling the Spatial Distribution of Bark Beetle Infestation

“…Only classes of strong and very strong damage of spruce forest stands [27] were considered for this study. A similar approach was used by [28,29].…”