Decision support system for tree care
Students Name: Dmytriv Alina Yuriivna
Qualification Level: magister
Speciality: Systems and Methods of Decision Making
Institute: Institute of Computer Science and Information Technologies
Mode of Study: full
Academic Year: 2022-2023 н.р.
Language of Defence: ukrainian
Abstract: One of the biggest environmental problems is the deterioration of forests. As of 2011, it was determined that the total area of Ukrainian forests is 10.4 million hectares, of which 9.6 million hectares are forested [1]. It turns out that the total forest coverage of the territory is 15.9%, although ecologists note that in fact, in recent years, the territory of Ukrainian forests has decreased to 11%. This forested area provides protection of water, soils and climate of Ukraine and the whole of Europe. Various diseases and pests, as well as natural disasters and human-caused pollution, affect trees very badly. However, poor quality of forest management and constant cutting down of forests cause a faster decrease in the area of forests. In forestry, satellite images are already used to analyze the forest area, but quite recently they have started using unmanned aerial vehicles with spectral cameras. Ukrainian scientists conducted several studies with the processing of satellite spectral images. This paper examines a study using the Global Forest Change data set and the Sentinel-2 satellite to decipher the species composition of forest stands in the "Boyar Forest Research Station" [2], as well as a study on the classification of mountain forests of the Carpathian National Nature Park to find out species composition of trees and their phytosanitary status [3]. Researches based on hyperspectral images from unmanned aerial vehicles are currently conducted mostly abroad. This paper reviews a study to assess the stage and severity of wilt infection in Chinese pine [4], and another study conducted in northern Portugal to identify defoliation disease in a natural black alder floodplain forest [5]. These researches used Random Forest algorithms and logistic regression, and as a result obtained data with high accuracy, compared to ground forest monitoring of the selected areas. Existing systems that solve the problem are mainly focused on forest inventory. They calculate tree metrics and store them, but do not analyze the data. That is why the user must independently determine the trees’ condition and form methods of treatment and prevention for trees, reviewing the obtained indicators. Among such analogues, the following systems are considered in this work: Tree Radar Unit, EOS Data Analytics Forest Monitoring, Katam Forest and CropsIT. Study object – the process of data analysis and implementation of necessary measures to improve the condition of trees. Scope of research – application of machine learning methods to detect damage to trees and form methods of their treatment and care. Goal of research: create a decision support system that will help identify possible tree diseases and suggest ways to treat and care them. When conducting a systematic analysis of the research object, a tree of goals is built to decompose the general goal and determine quality criterions. On the basis of these criterions and selected information systems alternatives, there is applied the analytical hierarchy method, it was determined that the type of system is a decision support system. Using the UML notation, the conceptual module of the system is built, namely, there are presented diagrams of use cases, classes, states, sequences, etc. In addition, the purpose of using the system for each user and its place of application are described in detail. The classification algorithm based on the decision tree method based on the Gini index was chosen as the system implementation method. It was decided to implement a decision support system for tree care using the Python programming language, namely using the Django web framework and a MySQL database. Layout is done using HTML and CSS. The result of the master’s thesis is a designed and implemented decision support system for tree care using the machine learning method of classification. The system helps to identify tree diseases based on certain characteristics by analyzing the files of the previous and current taxonomic characteristics. As a result, it provides the detection probability of three diseases, graphs for indicators and a list of recommendations for trees treatment and prevention. Keywords: decision support system, forest management, tree care, classification, decision tree. References. 1. General characteristic of Ukrainian forests. State Forest Resources Agency of Ukraine: website. URL: https://forest.gov.ua/en/areas-activity/forests-ukraine/general-characteristic-ukrainian-forests (Last accessed: 18.11.2022). 2. Georgian M.I., Myroniuk, V. V. Classification of tree species composition of forest stands using sentinel-2 satellite images. Forestry and horticulture: electronic scientific journal. URL: http://nbuv.gov.ua/UJRN/licgoc_2017_11_6 (Last accessed: 18.11.2022). 3. Lyalko V. I., Zholobak G. M., Hodorovsky A. J., Apostolov A. A., Sibirtseva O. M., Yelistratova L. O., Romanchuk I. F., Dorofey E. M. Space monitoring of the environmental – an effective mechanism of forest protection. Ukrainian Journal of Earth Remote Sensing. 2019. №20. P. 4-12. URL: https://ujrs.org.ua/ujrs/article/download/145/168 (Last accessed: 18.11.2022). 4. Run Yu, Lili Ren, Youqing Luo. Early detection of pine wilt disease in Pinus tabuliformis in North China using a field portable spectrometer and UAV-based hyperspectral imagery. Springer Open, Forest ecosystem. URL: https://forestecosyst.springeropen.com/articles/10.1186/s40663-021-00328-6 (Last accessed: 18.11.2022). 5. Guerra-Hernandez J., Diaz-Varela R.A., Avarez-Gonzalez J.G., Rodriguez-Gonzalez P.M. Assessing a novel modelling approach with high resolution UAV imagery for monitoring health status in priority riparian forests Springer Open, Forest ecosystem. URL: https://forestecosyst.springeropen.com/articles/10.1186/s40663-021-00342-8 (Last accessed: 18.11.2022).