머신러닝기법을 이용한 상수리나무의 수간고별 직경추정 연구
Estimation of Stem Taper for Quercus Acutissima Using Machine Learning Techniques
1
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
본 연구는 입목재적 계산에 기본이 되는 수간고별 직경 추정에 있어 머신러닝 기반 학습모델의 적용 가능성을 알아보기 위해 수행되었다. 이를위해 전국에 분포하는 상수리나무 표준목에서 얻은 자료를 이용하여 Kozak Model (Model 1)과 Random Forest Model (RF, Model 2 to 4)로부터수간고별 직경 값을 도출하였고, 이를 비교분석 하였다. RF는 입지환경 요소를 학습에 활용하여 3가지 모델 (Model 2 (수고, 흉고직경), Model 3 (수고, 흉고직경, 입지환경 요소), Model 4 (수고, 입지환경 요소))로 개발하였다. 모델의 적합도를 검증하기 위해 적합도 지수(Fitness Index), 편의(Bias), 평균절대편차(Mean Absolute Deviation), 평균 제곱근 오차(Root Mean Square Error)를 이용하여 모델의 적합성을 비교하였다. 분석결과, Kozak 모델(FI = 0.9695, Bias = 0.0292, RMSE = 1.5155, MAD = 1.1248) 에 비해 RF모델이 전반적으로 수간고에 따른 직경을 잘 예측하는모델로 나타났으며, RF 중에서도 입지환경 요소를 학습에 활용한 Model 3의 경우(FI = 0.9740, Bias = 0.0186, RMSE = 1.3995, MAD = 0.9746) 가장 우수한 성능을 보인 것으로 나타났다. 다만, 수간곡선을 통한 경향을 분석한 결과 일부 구간에서 수고의 변화에 따른 직경 추정에 이상치가발생하여 추후에는 파라미터조절이나 새로운 모델 적용 등을 통한 연구가 필요한 것으로 판단되었다. 본 연구를 통해 추정된 Random Forest 기반의 수간고별 직경추정 모델은 개체목 재적산출에 활용되어 산림경영을 위한 기초자료로 활용될 가능성이 있을 것으로 예상된다.
Abstract
This study was conducted to identify the applicability of a machine learning technique in the estimation of diameters by stem height that is a basic factor to measure stem volume. To this end, stem tapers were computed by using Kozak Model (Model 1) and Random Forest Model (RF, Model 2 to 4) based on data for samples of Quercus acutissima spread across South Korea to compare them. By including site characteristic factors in the learning, RF was developed into three different models: Model 2 with height and Diameter at Breast Height (DBH), Model 3 with height, DBH, and site characteristic factors, and Model 4 containing height and site characteristic factors. Compared were the values for Fitness Index(FI), Bias, Mean Absolute Deviation (MAD), and Root Mean Square Error (RMSE) to test the goodness of fit for each model. The findings showed the superiority of the RF models to the Kozak one(FI = 0.9695, Bias = 0.0292, RMSE = 1.5155, MAD = 1.1248) in estimating stem tapers. Among the RF types did Model 3 perform best (FI = 0.9740, Bias = 0.0186, RMSE = 1.3995, MAD = 0.9746). It should, however, be noted that the result from the analysis of trends of stem tapers revealed outliers in a few parts of the whole process. This requires further research on new approaches to control parameters.
참고문헌 (31)
- "Computer software. Vers. 3-3.7.4" (2019)[URL]
- Breiman L, "Random forests", Machine Learning, 45(1), pp.5~32 (2001)[DOI]
- Burkhart EH, "Modeling forest trees and stands", pp.458~ (2012), Springer Science and Business Media
- Diamantopoulou MJ, "Estimation of Weibull function parameters for modelling tree diameter distribution using least squares and artificial neural networks methods", Biosyst. Eng, 133, pp.33~45 (2015)
- Hastie T, "The elements of statistical learning: Data mining, inference, and prediction", pp.745~ (2009), Springer
- Kim DG, "Stem taper equation and bark thickness estimation equation for Pinus densiflora in Uljin, Gyeonbuk", Korean J. Forest Measurements, 5(1), pp.10~16 (2002)
- 고치웅, "변량지수식을 이용한 전국 상수리나무의 입목수간재적표 추정", 한국산림과학회지, 108(3), pp.357~363 (2019), 한국산림과학회[DOI]
- Korea Forest Service and Korea Forestry Promotion Institute, "The 7th national forest inventory and forest health monitering-Field manual-", Korea Forestry Promotion Institute, 86~ (2017)
- Kozak A, "A variable exponent taper equation", Can. J. For. Res., 18, pp.1363~1368 (1988)[DOI]
- Larson PR, "Stem form development of forest trees", Forest Science, 9(Suppl_2) (1963)[DOI]
- Larson PR, "Stem form of young Larix as influenced by wind and pruning", Forest Science, 11(4), pp.412~424 (1965)[DOI]
- Lee KS, "Pitch pine management plan(in korean)", National Institute of Forest Science, pp.11~16 (2007)
- Lee KS, "Forest measurement using air-borne Lidar (in korean)", Korea Institute of Science and Technology Information, 9~22 (2005)
- Lee WK, "Stem and stand taper model using spline function and linear equation", J. Korean Society of Forest Science, 83(1), pp.63~74 (1994)
- Louppe G, "Understanding Random Forests", Ph. D., pp.2014
- Max TA, "Segmented polynomial regression applied to taper equations", Forest Science, 22(3), pp.283~289 (1976)
- Muhairwe CK, "Effects of adding tree, stand, and site variables to Kozak’s variable-exponent taper equation", Canadian J. Forest Research, 24(2), pp.252~259 (1994)[DOI]
- Nunes MH, "Artificial intelligence procedures for tree taper estimation within a complex vegetation mosaic in Brazil", PloS One, 11(5), pp.1~16 (2016)[DOI]
- Özçelik R, "Estimating Crimean juniper tree height using nonlinear regression and artificial neural network models", For. Ecol, 306, pp.52~60 (2013)[DOI]
- Özçelik R., "Evaluation of potential modeling approaches for Scots pine stem diameter prediction in north-eastern Turkey", Computers and Electronics in Agriculture, 162, pp.773~782 (2019)[DOI]
- 박은정, "랜덤포레스트를 이용한 낙엽송과 편백의 적지적수도 제작: 경상남도를 대상으로", 농업생명과학연구, 53(6), pp.65~74 (2019), 농업생명과학연구원[DOI]
- Pedregosa F, "Scikit-learn : Machine Learning in Python", J. Machine Learning Research, 12, pp.2825~2830 (2011)
- Raschka S, "Python machine learning", pp.126~383 (2019), Gilbut
- Sachet JKA, "Tree value system:User guide (Vol. 234)", USDA, 45~ (1989)
- Sahin EK, "Assessing the predictive capability of ensemble tree methods for landslide susceptibility mapping using XGBoost, gradient boosting machine, and random forest", SN Applied Sciences, 2(7), pp.1~17 (2020)[DOI]
- Sakici OE, "Stem taper estimations with artificial neural networks for mixed Oriental beech and Kazdağı fir stands in Karabük region, Turkey", CERNE, 24, pp.439~451 (2018)
- Smith DM, "Some anatomical responses of loblolly pine to soil-water deficiencies", Tappi, 44(3), pp.179~185
- 손영모, "Stem taper Equations for Six Major Tree Species in Korea", 한국산림과학회지, 91(2), pp.213~218 (2002), 한국산림과학회
- Son YM, "Development of stem profile and taper equation for Quercus acuta in Wando", Korean J. Forest Measurements, (10), pp.1~6 (2007)
- Vahedi AA, "Artificial neural network application in comparison with modeling allometric equations for predicting above-ground biomass in the Hyrcanian mixedbeech forests of Iran", Biomass Bioenerg, 88, pp.66~76 (2016)
- Weiskittel AR, "Forest growth and yield modeling", pp.424~ (2011), Wiley-Blackwell