Nonparametric machine learning for mapping forest cover and exploring influential factors

Abstract

The contribution of forest ecosystem services to human well-being varies over space following the dynamics in forest cover. Use of machine learning models is increasing in projecting forest cover changes and investigating the drivers, yet references are still lacking for selecting machine learning models for spatial projection of forest cover patterns. ObjectivesWe assessed the ability of nonparametric machine learning techniques to project the spatial distribution of forest cover and identify its drivers using a case study of Tasmania, Australia. MethodsWe developed, evaluated, and compared the performance of four nonparametric machine learning models: support vector regression (SVR), artificial neural networks (ANN), random forest (RF), and gradient boosted regression trees (GBRT). ResultsThe results demonstrated that RF far outperformed the other three models in both fitting and projection accuracy, and required less computional costs. GBRT outperformed SVR and ANN in projection accuracy. However, RF exhibited serious overfitting due to the full growth of its decision trees. The influence rankings of explanatory variables on spatial patterns of forest cover were different under the four models. Land tenure type and rainfall were identified among the top four most influential variables by all four models. The ranking produced by the RF model was significantly different with topographic factors associated with land clearing and production costs (elevation and distance to timber facilities) being the two most influential variables. ConclusionsWe encourage practitioners to consider nonparametric machine learning methods, especially RF, when facing problems of complex environmental data modelling.

Assessing Biotic and Abiotic Effects on Biodiversity Index Using Machine Learning

“…Also, in a study in Western Himalaya [76] that used machine learning methods including classification regression tree (CART), random forest (RF), and support vector machine (SVM) algorithms showed results similar to the present study; the authors concluded that RF model has a higher accuracy in forest fire burn area. A study in Tasmania, Australia [77] used support vector regression (SVR), artificial neural networks (ANN), random forest (RF), and gradient boosted regression trees (GBRT) for mapping forest cover and exploring influential factors, and their findings were in line with the results of our study. In terms of projection accuracy, and required less computational costs, RF far outperformed the other three models [77].…”

“…A study in Tasmania, Australia [77] used support vector regression (SVR), artificial neural networks (ANN), random forest (RF), and gradient boosted regression trees (GBRT) for mapping forest cover and exploring influential factors, and their findings were in line with the results of our study. In terms of projection accuracy, and required less computational costs, RF far outperformed the other three models [77].…”

A hybrid deep-learning approach for complex biochemical named entity recognition

“…Moreover, they are context-sensitive, and require expert knowledge in specific fields and substantial efforts in maintaining rules and dictionaries. To overcome these drawbacks, the Hidden Markov Hodel (HMM), the maximum entropy model (MaxEnt), the Conditional Random Field (CRF), and classical machine learning methods such as the support vector machines [12] have gradually replaced the aforementioned traditional methods.…”

Utilização de dados hiperespectrais para diferenciação de variedades de cana-de-açúcar (>i<Saccharum officinarum>/i<) L.

“…Ressalta-se que a simples extração da informação dos sensores hiperespectrais ou a utilização de índices espectrais de vegetação, não permitem de forma simples a distinção de variedades de cana, não gerando resultados satisfatórios(FRASSON et al, 2007;. Nestes casos é necessário o uso de técnicas de aprendizagem de máquinas, as quais têm grande vantagem ao analisar dados com alta dimensionalidade, não lineares, provendo assim uma melhor acurácia(PARK et al, 2014;LIU et al, 2020). Tais técnicas vem sendo empregadas em diversas que as demais técnicas(LIU et al, 2020).…”

“…Nestes casos é necessário o uso de técnicas de aprendizagem de máquinas, as quais têm grande vantagem ao analisar dados com alta dimensionalidade, não lineares, provendo assim uma melhor acurácia(PARK et al, 2014;LIU et al, 2020). Tais técnicas vem sendo empregadas em diversas que as demais técnicas(LIU et al, 2020). Outra questão importante é com relação ao número de amostras que, quando reduzidas em comparação ao número de variáveis, podem acarretar performance ruim do modelo, o que eventualmente é resolvido com técnicas de redução de dimensionalidade, como a análise de componentes principais (PCA).…”