Estimation of forest area by large plot interpretation and division based on remote sensing.

Based on “forestland spatial distribution” database, multisource remote sensing (RS) data from 2010 to 2013, continuous forest inventory data in 2010 and other forestry management materials, the area of forest and other different land types in Longjiang Forest Industry Group (LFIG, in Heilongjiang Province) was estimated by interpretation and division on RS large plots. Results showed that: 1) 231 large plots with 20 km interval were set in LFIG, and the 4 km×4 km quadrat was suitable for monitoring with variation coefficient of forest land area of 14.41%. 2) Interpretation accuracy of land type area increased with resolution ratio of RS image. Interpretation accuracy of all land types in RS large plots reached 90.68%, highest as 96.82% of that in forestry land. 3) There was no significant difference among forestry land area, which was monitored by large plots at different scales. Forestry land of LFIG monitored by 4 km×4 km large plots occurred in a higher proportion of 89.47% in land, as almost 9 million hectare, which included 8.7 million hectare forest land. 4) Beside enclosure, afforestation was the main practice for the rise in forestland area. Conversely, deforestation, forest harvesting and forestland requisition occupation all caused decline of forestland area. Forestland area presented a net decrease of 59.4 thousand hectare from 2010 to 2013, i.e., an average reduction of 14.8 thousand hectare every year in LFIG. In conclusion, the RS large plot method proposed in this research for dynamic monitoring forest resources is reliable and worthy of promotion to relevant fields.