Estimation of forestry-biomass using k-nearest neighbor(k-NN) method

Jung Soo Lee, Shigejiro Yoshida

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)


The purpose of this study was to estimate of forestry-biomass using by k-Nearest Neighbor (k-NN) algorithm with the Landsat TM and field survey data in research forest of Kangwon national university. Optimum reference plots (k) were selected to estimate the forest biomass based on the minimum horizontal reference area (HRA) and spatial filtering using DN (Digital number), NDVI (Normalized difference vegetation index) and TC (Tasseled cap). The accuracy of RMSE was better in the order: DN, NDVI, and TC, respectively. In the DN value application, the RMSE of coniferous and broadleaved trees had the minimum value when k=l 1 of HRA 4 km and k=6 of HRA 10 km with 7by7 filtering. The bias of each was overestimated by 1.0 t/ha and 1.2 t/ha respectively. On the other hand, the minimum RMSE of Pinus koraiensiss had at k=8 and HRA of 4 km without filtering and the bias were underestimated by 1.6 t/ha. As a result, the estimated total forestry biomass was 802,0001 and 252 t/ha for k-NN methods. The results were higher than the plot data estimation by 16 t/ha. In this study, it is able to precise forest biomass at regional forest.

Original languageEnglish
Pages (from-to)339-349
Number of pages11
JournalJournal of the Faculty of Agriculture, Kyushu University
Issue number2
Publication statusPublished - Sept 2013

All Science Journal Classification (ASJC) codes

  • Agronomy and Crop Science
  • Biotechnology


Dive into the research topics of 'Estimation of forestry-biomass using k-nearest neighbor(k-NN) method'. Together they form a unique fingerprint.

Cite this