인공신경망 기법을 이용한 청미천 유역 Flux tower 결측치 보정 = A point-scale gap filling of the flux-tower data using the artificial neural network

다국어 초록 (Multilingual Abstract)

In this study, we estimated missing evapotranspiration (ET) data at a eddy-covariance flux tower in the Cheongmicheon farmland site using the Artificial Neural Network (ANN). The ANN showed excellent performance in numerical analysis and is expanding in various fields. To evaluate the performance the ANN-based gap-filling, ET was calculated using the existing gap-filling methods of Mean Diagnostic Variation (MDV) and Food and Aggregation Organization Penman-Monteith (FAO-PM). Then ET was evaluated by time series method and statistical analysis (coefficient of determination, index of agreement (IOA), root mean squared error (RMSE) and mean absolute error (MAE). For the validation of each gap-filling model, we used 30 minutes of data in 2015. Of the 121 missing values, the ANN method showed the best performance by supplementing 70, 53 and 84 missing values, respectively, in the order of MDV, FAO-PM, and ANN methods. Analysis of the coefficient of determination (MDV, FAO-PM, and ANN methods followed by 0.673, 0.784, and 0.841, respectively.) and the IOA (The MDV, FAO-PM, and ANN methods followed by 0.899, 0.890, and 0.951 respectively.) indicated that, all three methods were highly correlated and considered to be fully utilized, and among them, ANN models showed the highest performance and suitability. Based on this study, it could be used more appropriately in the study of gap-filling method of flux tower data using machine learning method.

국문 초록 (Abstract)

본 연구에서는 청미천 유역에서의 플럭스타워에서 산출되는 증발산량의 결측값을 보완하기 위해 인공신경망(Artificial Neural Network, ANN)을 사용하였다. 비교 평가를 위해, Mean Diurnal Variation(MDV), Food and Agriculture Organization Penman-Monteith(FAO-PM) 방법들을 이용하여 증발산량을 산정하였고, ANN 방법을 이용한 결과와 비교하였다. 비교 평가 방법으로 시계열 방법 및 통계 분석(결정계수, IOA, RMSE, MAE)이 사용되었다. 각 gap-filling 모델의 검증을 위해 2015년의 30분 단위 데이터를 이용하였으며, 121개의 결측값 중 MDV, FAO-PM, ANN 방법 순으로 각각 70, 53, 54개의 결측값을 보완하여 모든 데이터가 관측되지 않은 36개의 데이터를 제외하면 각각 82.4%, 62.4%, 63.5%의 성능을 보였다. 결정계수(MDV, FAO-PM, ANN 방법 순으로 각각 0.673, 0.784, 0.841)와 IOA(MDV, FAO-PM, ANN 방법 순으로 각각 0.899, 0.890, 0.951)를 분석한 결과, 3가지 방법 모두 양질의 상관성을 보여 활용성이 충분하다고 판단되며, 이 중 ANN 모델이 가장 높은 적합도와 양질의 성능을 나타내었다. 본 연구를 기반으로 기계학습방법을 이용한 플럭스 타워 자료의 gap-filing 연구에 보다 적절하게 활용될 수 있을 것이다.

참고문헌 (Reference)

1 강민석, "한국 플럭스 관측망의 과거, 현재와 미래" 한국농림기상학회 20 (20): 1-4, 2018

2 김기영, "플럭스타워 기반 증발산의 결측자료 보정을 위한 gap-filling 산정 및 분석 - FAO-PM, MDV, Kalman filter을 이용하여 -" 한국방재학회 16 (16): 95-107, 2016

3 정대일, "증발량 관련 기후인자와 팬증발량의 변화 분석" 한국수자원학회 42 (42): 117-129, 2009

4 허선영, "머신러닝기반 범죄발생 위험지역 예측" 한국지리정보학회 21 (21): 64-80, 2018

5 렌민 유안, "경사진 KoFlux 산림유역에서 에디공분산 플럭스 산출에 미치는 좌표회전의 효과" 한국농림기상학회 9 (9): 100-108, 2007

6 Gao, H., "Using machine learning to predict suitable conditions for organic reactions" 4 (4): 1465-1476, 2018

7 Post, H., "Uncertainty analysis of eddy covariance CO2 flux measurements for different EC tower distances using an extended two-tower approach" 12 (12): 1205-, 2015

8 Papale, D., "Towards a standardized processing of net ecosystem exchange measured with eddy covariance technique : Algorithms and uncertainty estimation" 3 (3): 571-583, 2006

9 Kendale, S., "Supervised machine-learning predictive analytics for prediction of postinduction hypotension" 129 (129): 675-688, 2018

10 Choi, B., "Solving local minima problem with large number of hidden nodes on two-layered feed-forward artificial neural networks" 71 (71): 3640-3643, 2008

11 Ambas, V. T., "Sensitivity analysis of different evapotranspiration methods using a new sensitivity coefficient" 14 (14): 335-343, 2012

12 Hunter, D., "Selection of proper neural network sizes and architectures-A comparative study" 8 (8): 228-240, 2012

13 Allen, R. G., "Satellite-based energy balance for mapping evapotranspiration with internalized calibration(METRIC)-Model" 133 (133): 380-394, 2007

14 Park, J., "Satellite-based crop coefficient and evapotranspiration using surface soil moisture and vegetation indices in Northeast Asia" 156 : 305-314, 2017

15 Sheela, K.G, "Review on methods to fix number of hidden neurons in neural networks" 2013

16 Jensen, J. R., "Predictive modelling of coniferous forest age using statistical and artificial neural network approaches applied to remote sensor data" 20 (20): 2805-2822, 1999

17 Mitchell, T. M., "Machine learning" The McGraw-Hill Companies Inc 52-78, 1997

18 홍진규, "KoFlux 에디 공분산 자료 처리의 표준화" 한국농림기상학회 11 (11): 19-26, 2009

19 권효정, "KoFlux 관측지에서 에디 공분산 자료의 품질관리 및 보증" 한국농림기상학회 9 (9): 260-267, 2007

20 Atkinson, P. M., "Introduction neural networks in remote sensing" 18 (18): 699-709, 1997

21 Baldocchi, D., "FLUXNET : A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities" 82 (82): 2415-2434, 2001

22 Peterson, T. C., "Evaporation losing its strength" 377 (377): 687-688, 1995

23 Park, J., "Evaluation of statistical gap fillings for continuous energy flux(evapotranspiration)measurements for two different land cover types" 29 (29): 2021-2035, 2015

24 Marin, F. R., "Crop coefficient changes with reference evapotranspiration for highly canopy-atmosphere coupled crops" 163 : 139-145, 2016

25 Jean, N., "Combining satellite imagery and machine learning to predict poverty" 353 (353): 790-794, 2016

26 Roderick, M. L., "Changes in Australian pan evaporation from 1970 to 2002" 24 (24): 1077-1090, 2004

27 Tamura, S. I., "Capabilities of a four-layered feedforward neural network : Four layers versus three" 8 (8): 251-255, 1997

28 Panchal, G., "Behaviour analysis of multilayer perceptrons with multiple hidden neurons and hidden layers" 3 (3): 332-337, 2011

29 Chen, B., "Assessing tower flux footprint climatology and scaling between remotely sensed and eddy covariance measurements" 130 (130): 137-167, 2009

30 Langley, P., "Applications of machine learning and rule induction" 38 (38): 54-64, 1995

31 Sartori, M. A., "A simple method to derive bounds on the size and to train multilayer neural networks" 2 (2): 467-471, 1991