The paper presents the results of estimation of growing stock volume and live biomass in forest stands using combination of forest inventory measurements, multispectral satellite images RapidEye and digital elevation model (DEM). In a context of classification of remote sensing data we considered two nonparametric methods – k-Nearest Neighbors (k-NN) and Random Forest (RF). We concluded that RF outperforms kNN method nevertheless both of them provide quite accurate estimation of mean value of growing volume in a range of ±5 m3·ha-1, different components of aboveground biomass - ±1–2 t·ha-1
[1] Bilous А. М. Biological productivity and ecosystem functions of softwood deciduous forests in the Ukrainian Polissya : The Manuscript : 06.03.02 , 06.03.03 / Bilous Andrii – Kyiv, 2016. – 423 p.
[2] Breiman L. Random Forest / L. Breiman // Machine Learning. – 2001. – Vol. 45. – № 1. – P. 5–32.
[3] Crookston N. L. yaImpute: An R Package for ¬kNN Imputation / N. L. Crookston, A. O. Finley // Journal of Statistical Software. – 2008. – Vol. 23. – Issue 10. – 1–16.
[4] Imputing forest structure attributes from stand inventory and remote sensed data in Western Oregon, USA / A. T. Hudak, A. T. Haren, N. L. Crookston et al. // Forest Science. – 2014. – Vol. 60. – Issue. 2.– P. 253–269.
[5] McRoberts R. E. Estimation forest attribute parameters for small areas using nearest neighbors techniques / R. E. McRoberts // Forest Ecology and Management. – 2012. – Vol. 272. – P. 3–12.
[6] Tables and models of growth and productivity of forest of forming species of Northern Eurasia (standard and reference materials) – М.: 2006. – 803 p.
[7] Tomppo E. Satellite image-based National Forest Inventory of Finland / E. Tompo // International Archives of Photogrammetry and Remote Sensing. – 1991. – Vol. 28: 1–7. – P. 419–424.