Artificial Intelligence Applied to Parameter Enhancement for Carbon Flux Monitoring by Satellite in the Amazon Region

• Jean A. C. Dias UFPA
• Leonardo de O. Tamasauskas UFPA
• Pedro H. do V. Guimarães UFPA
• Alan B. S. Corrêa UFPA
• João D. C. D. Neto UFPA
• Albert E. C. dos Santos UFPA
• Danilo Souza CARBONEXT
• Ermínio R. Paixão UFPA
• José G. dos S. Fernandes UFPA
• Gabriel B. Costa UFPA
• Marcos C. da R. Seruffo UFPA

DOI: https://doi.org/10.5753/wcama.2024.2073

Abstract

The monitoring of atmospheric carbon flux has great importance in the comprehension of the ecosystems behaviors, being described in the measures of Gross Primary Production and Net Primary Production (GPP and NPP). Therefore, this article aims to apply the use of a machine learning algorithm to improve the parameters of the satellite product MOD17, in order to approximate their estimates of GPP and NPP in the Amazon to the Flux Towers’ data at Santarém, in Brazil, and Iquitos, in Peru. Comparisons using the new obtained parameters demonstrated a reduction of the Root-Mean-Square Error (RMSE) in GPP of up to 9.72% and of the Mean Absolute Error (MAE) in NPP of up to 37.8%, indicating more precise and stable estimations.

References

Anić, M., Ostrogović Sever, M. Z., Alberti, G., Balenović, I., Paladinić, E., Peressotti, A., Tijan, G., Večenaj, Z., Vuletić, D., and Marjanović, H. (2018). Eddy covariance vs. biometric based estimates of net primary productivity of pedunculate oak (quercus robur l.) forest in croatia during ten years. Forests, 9(12).

Chen, X., Wei, D., and Wang, X. (2023). A modified sea ice area empirical prediction model based on multi-source satellite remote sensing data for liaodong bay of the bohai sea, china. In Journal of Physics: Conference Series, volume 2486, page 012020. IOP Publishing.

Chu, H., Christianson, D. S., Cheah, Y.-W., Pastorello, G., O’Brien, F., Geden, J., Ngo, S.-T., Hollowgrass, R., Leibowitz, K., Beekwilder, N. F., et al. (2023). Ameriflux base data pipeline to support network growth and data sharing. Scientific Data, 10(1):614.

Costa, G. B., Santos e Silva, C. M., Mendes, K. R., dos Santos, J. G., Neves, T. T., Silva, A. S., Rodrigues, T. R., Silva, J. B., Dalmagro, H. J., Mutti, P. R., et al. (2022). Wue and co2 estimations by eddy covariance and remote sensing in different tropical biomes. Remote Sensing, 14(14):3241.

Danelichen, V. H., Biudes, M. S., Velasque, M., Machado, N. G., Gomes, R. S., Vourlitis, G. L., and Nogueira, J. S. (2015). Estimating of gross primary production in an amazon-cerrado transitional forest using modis and landsat imagery. Anais da Academia Brasileira de Ciências, 87:1545–1564.

Endsley, K. A., Zhao, M., Kimball, J. S., and Devadiga, S. (2023). Continuity of global modis terrestrial primary productivity estimates in the viirs era using model-data fusion. Journal of Geophysical Research: Biogeosciences, 128(9):e2023JG007457.

Feitosa, T. B., Fernandes, M. M., Santos, C. A. G., da Silva, R. M., Garcia, J. R., de Araujo Filho, R. N., de Moura Fernandes, M. R., and da Cunha, E. R. (2023). Assessing economic and ecological impacts of carbon stock and land use changes in brazil’s amazon forest: A 2050 projection. Sustainable Production and Consumption, 41:64–74.

GMAO (2015). MERRA-2: 2d, 1-Hourly, Time-Averaged, Single-Level, Assimilation, Single-Level Diagnostics V5.12.4. Greenbelt, MD, USA, Goddard Earth Sciences Data and Information Services Center (GES DISC). Global Modeling and Assimilation Office (GMAO). DOI: 10.5067/VJAFPLI1CSIV.

Hong Tao, Manqi Li, M. W. and Lü, G. (2019). Genetic algorithm-based method for forest type classification using multi-temporal ndvi from landsat tm imagery. Annals of GIS, 25(1):33–43.

Huang, X., Xiao, J., Wang, X., and Ma, M. (2021). Improving the global modis gpp model by optimizing parameters with fluxnet data. Agricultural and Forest Meteorology, 300:108314.

Liu, C., Pan, G., Song, D., and Wei, H. (2023). Air quality index forecasting via genetic algorithm-based improved extreme learning machine. IEEE Access, 11:67086–67097.

Roman, T., Griffis, T., Kolka, R., Wayson, C., Lilleskov, E., del Castillo Torres, D., Malaverri, L. F., and Marin, J. E. R. (2021). Ameriflux fluxnet1f pe-qfr quistococha forest reserve, ver. 3-5, ameriflux amp, (dataset). DOI: 10.17190/AMF/1832157.

Saleska, S. (2019). Ameriflux base br-sa1 santarem-km67-primary forest, ver. 5-5, ameriflux amp, (dataset). DOI: 10.17190/AMF/1245994.

Wang, H., Li, X., Ma, M., and Geng, L. (2019). Improving estimation of gross primary production in dryland ecosystems by a model-data fusion approach. Remote Sensing, 11(3).