Vol. 36, issue 02, article # 7

Abstract:

Forest are among the main places on Earth where carbon is collected and accumulated. However, quantitative instrumental assessment of carbon fluxes is possible only for small-scale areas. When solving the scaling problem, we use machine learning methods, which can transform the values of the intensity of the Earth’s surface reflectance in different spectral intervals into ground-based in situ observations. The assessments of carbon fluxes by a regression neural network model of the multilayer perceptron type trained on FLUXNET network data for a station located in a boreal coniferous forest (56.4615°N, 32.9221°E) are presented. Using vegetation indicies NDVI and EVI measured by MODIS Aqua, air temperature at an altitude of 2 m, and total precipitation as input data, the model estimates of gross primary production (GPP), net ecosystem exchange (NEE), ecosystem respiration (TER), and some other parameters describing water and energy fluxes are calculated. Statistical estimation provides high values of the correlation coefficient and Nash–Sutcliffe coefficient on test dataset: R > 0.9 and NSE ≥ 0.87 for GPP and TER; R = 0.4 and NSE = 0.15 for NEE.
 

Keywords:

neural networks, machine learning, carbon fluxes, FLUXNET MODIS

References:

1. Baldocchi D., Falge E., Gu L., Olson R., Hollinger D., Running S., Anthoni P., Bernhofer Ch., Davis K., Evans R., Fuentes J., Goldstein A., Katul G., Law B., Lee X., Malhi Y., Meyers T., Munger W., Oechel W., Paw U.K.T., Pilegaard K., Schmid H.P., Valentini R., Verma S., Vesala T., Wilson K., Wofsy S. FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities // Bull. Am. Meteorol. Soc. 2001. V. 82, N 11. P. 2415–2434.
2. Gribanov K.G., Imasu R., Zakharov V.I. Neironnye seti dlya opredeleniya vysotnyh profilei CO₂ po dannym GOSAT/TANSO-FTS // Optika atmosf. i okeana. 2009. V. 22, N 9. P. 890–895; Gribanov K.G., Imasu R., Zakharov V.I. Neural networks for CO₂ profile retrieval from the data of GOSAT/TANSO-FTS // Atmos. Ocean. Opt. 2010. V. 23, N 1. P. 42–47.
3. Rolnick D., Donti P.L., Kaack L.H., Kochanski K., Lacoste A., Sankaran K., Slavin Ross A., Milojevic-Dupont N., Jaques N., Waldman-Brown A., Luccioni A., Maharaj T., Sherwin E.D., Karthik Mukkavilli S., Kording K.P., Gomes C., Ng A.Y., Hassabis D., Platt J.C., Creutzig F., Chayes J., Bengio Y. Tackling climate change with machine learning // ACM Computing Surveys (CSUR). 2022. V. 55, N 2. P. 1–96.
4. Heermann P.D., Khazenie N. Classification of multispectral remote sensing data using a back-propagation neural network // IEEE Trans. Geosci. Remote Sens. 1992. V. 30, N 1. P. 81–88.
5. Cao M., Marshall S., Gregson K. Global carbon exchange and methane emissions from natural wetlands: Application of a process-based model // J. Geophys. Res.: Atmos. 1996. V. 101, N D9. P. 14399–14414.
6. Dou X., Yang Y., Luo J. Estimating forest carbon fluxes using machine learning techniques based on eddy covariance measurements // Sustainability. 2018. V. 10, N 1. P. 203.
7. Yu T., Zhang Q., Sun R. Comparison of machine learning methods to up-scale gross primary production // Remote Sensing. 2021. Т. 13, № 13. С. 2448.
8. Polyakov A.V. Ispol'zovanie metoda iskusstvennyh neironnyh setei pri vosstanovlenii vertikal'nyh profilei atmosfernyh parametrov // Optika atmosf. i okeana. 2014. V. 27, N 1. P. 34–39; Polyakov A.V. The method of artificial neural networks in retrieving vertical profiles of atmospheric parameters // Atmos. Ocean. Opt. 2014. V. 27, N 3. P. 247–252.
9. Zeng J., Matsunaga T., Tan Z.-H., Saigusa N., Shirai T., Tang Y., Peng S., Fukuda Y. Global terrestrial carbon fluxes of 1999–2019 estimated by upscaling eddy covariance data with a random forest // Scientific Data. 2020. V. 7, N 1. P. 1–11.
10. Tramontana G., Jung M., Schwalm C.R., Ichii K., Camps-Valls G., Ráduly B., Reichstein M., Arain M.A., Cescatti A., Kiely G., Merbold L., Serrano-Ortiz P., Sickert S., Wolf S., Papale D. Predicting carbon dioxide and energy fluxes across global FLUXNET sites with regression algorithms // Biogeosciences. 2016. N 13. P. 4291–4313. DOI: 10.5194/bg-13-4291-2016.
11. Kurbatova J., Li C., Varlagin A., Xiao X., Vygodskaya N. Modeling carbon dynamics in two adjacent spruce forests with different soil conditions in Russia // Biogeosciences. 2008. V. 5, N 4. P. 969–980.
12. Zhengxing W., Chuang L., Alfredo H. From AVHRR-NDVI to MODIS-EVI: Advances in vegetation index research // Acta Ecologica Sinica. 2003. V. 23, N 5. P. 979–987.
13. ORNL DAAC 2018. MODIS and VIIRS Land Products Global Subsetting and Visualization Tool. ORNL DAAC, Oak Ridge, Tennessee, USA. Accessed February, 2022. Subset obtained for MYD13Q1 product at 56,46,32,92, time period: [04.07.2002] to [31.12.2014], and subset size: 0.250 ´ 0.250 km. DOI: 10.3334/ORNLDAAC/1379.
14. NASA Earth Data Portal; Mission and Measurments; MODIS Terra&Aqua. URL: https://ladsweb.modaps. eosdis.nasa.gov/missions-and-measurements/modis/.
15. ECMWF Copernicus Climate Change Data Service; ERA5 hourly land products. URL: https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land?tab=form.
16. Baldocchi D., Hicks B., Meyers T. Measuring biosphere-atmosphere exchanges of biologically related gases with micrometeorological methods // Ecology. 1988. V. 69. P. 1331–1340. DOI: 10.2307/1941631.
17. Schmid H.P. Source areas for scalars and scalar fluxes // Bound.-Layer Meteorol. 1994. V. 67. P. 293–318. DOI: 10.1007/BF00713146.
18. Reichle D.E. The Global Carbon Cycle and Climate Change: Scaling Ecological Energetics from Organism to the Biosphere. Elsevier, 2020. P. 150–152. DOI: 10.1016/C2019-0-01382-9.
19. Nash J.E., Sutcliffe J.V. River flow forecasting through conceptual models part I – A discussion of principles // J. Hydrol. 1970. V. 10, N 3. P. 282–290.
20. Bottou L. Large-scale machine learning with stochastic gradient descent // Proc. COMPSTAT'2010. 2010. P. 177–186.
21. Paszke A., Gross S., Massa F., Lerer A., Bradbury J., Chanan G., Killeen T., Lin Z., Gimelshein N., Antiga L., Desmaison A., Kopf A., Yang E., DeVito Z., Raison M., Tejani A., Chilamkurthy S., Steiner B., Fang L., Bai J., Chintala S. PyTorch: An imperative style, high-performance deep learning library // Adv. Neural Inform. Proces. Syst. 2019. V. 32. P. 8024–8035.
22. Huang N., Wang L., Zhang Y., Gao S., Niu Z. Estimating the Net Ecosystem Exchange at global FLUXNET sites using a random forest model // IEEE J. Sel. Top. Appl. Earth Observ. Remote Sens. 2021. V. 14. P. 9826–9836. DOI: 10.1109/JSTARS.2021.3114190.