|
In this study, two different machine learning approaches, namely Artificial Neural Network (ANN) [1-2], and supported Vector Regressions (SVR) [3-4] have been implemented and tested for estimating the forest biomass (t/ha) from the ESA airborne SAR missions. The study was carried out in the framework of the BRIX exercise, aimed at intercomparing biomass retrieval algorithms for P-band full-polarimetric SAR sensors in view of the upcoming ESA BIOMASS mission (a P-band synthetic aperture polarimetric radar).
Several strategies have been exploited, by developing “general” algorithms trained with data derived from the whole dataset and “specific” algorithms, trained with data derived from a single campaign, among Afrisar, Biosar and Tropisar. In all cases, the algorithms have been trained on a subset of the available data and validated on the remaining, obtaining correlation coefficients between R=0.82 and R= 0.94, with a RMSE between 15 t/ha and 70 t/ha, depending on the algorithm and on the dataset.
In the case of ANN, the validation of the “general” and “specific” algorithms resulted in a correlation coefficient between R=0.78 and R=0.94, depending on the dataset, with a RMSE between 15 and 60 t/ha and negligible BIAS.
The validation of the SVR algorithms resulted in a correlation coefficient between R=0.27 and R=0.90, depending on the dataset, with a corresponding RMSE between 25 and 77 t/ha and BIAS negligible in this case too.
After validation, both ANN and SVR algorithms have been applied to the whole SAR images available for generating the corresponding biomass maps.
References
[1]. Santi E., S. Paloscia, S. Pettinato and G. Fontanelli. “Application of artificial neural networks for the soil moisture retrieval from active and passive microwave spaceborne sensors,” Int. J. Appl. Earth Observ. Geoinf., vol 48, pp. 61–73, Jun. 2016.
[2]. Santi E., S. Paloscia, S. Pettinato, G. Fontanelli, M. Mura, C. Zolli, F. Maselli, M. Chiesi, L. Bottai, G. Chirici, 2017, The potential of multifrequency SAR images for estimating forest biomass in Mediterranean areas, Remote Sensing of Environment 200 (2017), pp. 63–73.
[3]. Pasolli, L., Notarnicola, C., Bruzzone, L., Bertoldi, G., Della Chiesa, S., Hell, V., Niedrist, G., Tappeiner, U., Zebisch, M., Del Frate, F., Vaglio Laurin, G. 2011. Estimation of Soil Moisture in an Alpine Catchment with RADARSAT2 Images. Hindawi Publishing Corporation, Applied and Environmental Soil Science, Article ID 175473, 12 pages, doi:10.1155/2011/175473
[4]. Pasolli, L., Notarnicola, C., Bertoldi, G., Bruzzone, L., Remelgado R., Greifeneder, F., Niedrist, G., Della Chiesa, Tappeiner, U., Zebisch, M. 2015. Estimation of Soil Moisture in Mountain Areas Using SVR Technique Applied to Multiscale Active Radar Images at C-Band. IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 8, NO. 1, JANUARY 2015, doi: 10.1109/JSTARS.2014.2378795.
|
