Volume XLII-4/W18
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 1131–1136, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-1131-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 1131–1136, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-1131-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

  19 Oct 2019

19 Oct 2019

ESTIMATING THE FOREST STAND VOLUME AND BASAL AREA USING PLEIADES SPECTRAL AND AUXILIARY DATA

M. Zahriban Heasari1, A. Fallah2, S. Shataee1, S. Kalbi2, and H. Persson3 M. Zahriban Heasari et al.
  • 1Dept. of Forestry, Gorgan University of Agricultural Sciences and Natural Resources, Iran
  • 2Dept. of Forestry, Sari University of Agricultural Sciences and Natural Resources, Iran
  • 3Dept. of Forest Resource Management, Forest Remote Sensing, Swedish University of Agricultural Sciences, Sweden

Keywords: Pleiades, Auxiliary data, Spectral data, nonparametric methods, Volume, Basal area

Abstract. The aim of this study was to evaluate the improvements of volume and basal area estimations, when spectral data from the Pleiades were complemented with auxiliary data. The study area was located in the Darabkola's forest of Sari, Iran. In-situ data were collected for 144 circular sample plots, with 17.84 m radius, which were distributed using a simple random sampling design. Tree information included diameter at breast height (DBH) of all trees within the sample plots, and the height of some trees. By using DBH and tree height, the volume and basal area per hectare was also computed for each plot. Geometric and radiometric corrections of spectral data were applied to the images. In addition, the auxiliary maps of slope, aspect, elevation, soil pH and texture (through ground sampling and interpolation), precipitation and temperature (through interpolation of climate stations) were prepared. Digital values corresponding to ground plots were extracted from spectral bands and auxiliary data and considered as independent variables while volume and basal area were selected as dependent variables. The forest modeling was carried out using a non-parametric method of random forest (RF), using 70% of the sample plots as training data. The results were validated using the remaining 30% sample plots. The results indicated that by using both spectral and auxiliary data, the RMSE was reduced by 5% compared to using only spectral data for volume modeling. The corresponding advantage of using both spectral and auxiliary data was 1% to 3% when basal area was modeled.