2026-10-06T11:33:05Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34037553 2026-10-01T01:31:31Z category_24019 category_28819 item_type_3 month_year_10_2026

National forest type classification of Nepal: 15 forest types at 30 m from Landsat 8 (2013/14) Shiva Khanal (13245867) Forest ecosystems Spatial data and applications Nepal Forest Type Map Himalaya Vegetation Spatial Data <p dir="ltr">This dataset provides a wall-to-wall forest type map of Nepal at 30 m spatial resolution, classifying the forested area of the country into 15 forest types. The map was produced to support forest monitoring by forest type and REDD+ Measurement, Reporting and Verification (MRV) at national and sub-national scales. It represents forest conditions for 2013/14.</p><p dir="ltr">METHODS</p><p dir="ltr">Landsat 8 OLI surface data from 2013/14 (bands 2 to 7) were segmented into image objects using object-based image analysis (OBIA) in eCognition. The objects were classified with a Classification and Regression Tree (CART) algorithm. Predictor variables included spectral information from Landsat 8 bands 2 to 7, Haralick texture features, and terrain variables. Classification was restricted to forest using a forest mask from Department of Forest Research and Survey (DFRS). Training and reference data were field data from the Forest Resource Assessment (FRA) Nepal project, DFRS. </p><p dir="ltr">ACCURACY</p><p dir="ltr">Accuracy was assessed with 597 reference observations</p><p dir="ltr">Overall accuracy: 69.85% (95% CI: 66.08% to 73.61%)</p><p dir="ltr">Kappa: 0.63 (95% CI: 0.58 to 0.67)</p><p dir="ltr">User's / producer's accuracy by class:</p><p dir="ltr">TMH 0.77 / 0.80; UMH 0.68 / 0.89; LMH 0.70 / 0.79; S 0.69 / 0.60; Pr 0.62 / 0.63; Pw 0.62 / 0.47; Q 0.76 / 0.32; A 0.80 / 0.50; Ce 0.50 / 0.20; Bu 0.25 / 0.50; Td n/a / 0.00; KS/SK 0.00 / 0.00.</p><p dir="ltr">Jw, Ct and Sp were not represented in the validation sample. Accuracies for minor classes rest on very few samples and should be interpreted with caution.</p><p dir="ltr">FILE DETAILS</p><p dir="ltr">Format: GeoTIFF, single band, 8-bit unsigned integer, categorical</p><p dir="ltr">CRS: WGS 84 / UTM zone 44N (EPSG:32644), units metres</p><p dir="ltr">Pixel size: 30 m x 30 m (pixel area 0.09 ha)</p><p dir="ltr">Dimensions: 26,891 columns x 13,384 rows</p><p dir="ltr">Origin (upper left): 408375, 3345435</p><p dir="ltr">Extent: 408375 to 1215105 E; 2943915 to 3345435 N (approx. 80.05 to 88.41 E, 26.44 to 30.24 N)</p><p dir="ltr">A raster attribute table (.vat.dbf / .aux.xml) gives the class code, abbreviation and full name.</p><p dir="ltr">CLASS CODES AND MAPPED AREA</p><p dir="ltr">Code | Abbr. | Forest type | Area (ha) | Share of forest (%)</p><p dir="ltr">1 | TMH | Terai Mixed Hardwood | 1,445,743 | 24.6</p><p dir="ltr">2 | S | Sal (Shorea robusta) | 896,772 | 15.3</p><p dir="ltr">4 | LMH | Lower Mixed Hardwood | 1,002,071 | 17.1</p><p dir="ltr">5 | UMH | Upper Mixed Hardwood | 1,070,602 | 18.2</p><p dir="ltr">6 | Pr | Pinus roxburghii | 496,648 | 8.5</p><p dir="ltr">7 | KS/SK | Acacia catechu & Dalbergia sissoo | 80,412 | 1.4</p><p dir="ltr">8 | Pw | Pinus wallichiana | 198,013 | 3.4</p><p dir="ltr">9 | Q | Quercus spp. | 449,006 | 7.6</p><p dir="ltr">10 | Jw | Juglans wallichiana | 6,091 | 0.1</p><p dir="ltr">11 | Td | Tsuga dumosa | 18,460 | 0.3</p><p dir="ltr">12 | A | Abies spectabilis & Abies pindrow | 105,667 | 1.8</p><p dir="ltr">13 | Ct | Cupressus torulosa | 19,948 | 0.3</p><p dir="ltr">15 | Ce | Cedrus deodara | 27,566 | 0.5</p><p dir="ltr">16 | Bu | Betula utilis | 36,453 | 0.6</p><p dir="ltr">17 | Sp | Picea smithiana | 20,767 | 0.4</p><p dir="ltr">Total mapped forest: 5,874,219 ha (65,269,096 pixels).</p><p dir="ltr">Codes 3 and 14 are not used.</p><p><br></p><p dir="ltr">USAGE NOTES</p><p dir="ltr">The map shows forest type within the forest mask only; non-forest areas are NoData. Areas are pixel counts and have not been adjusted for classification error, so they should not be read as unbiased area estimates. Rare and spectrally similar classes (e.g. KS/SK, Td, Ce, Bu) have low or unknown accuracy. Users should aggregate classes or apply error-adjusted area estimation where appropriate.</p><p><br></p><p dir="ltr">ACKNOWLEDGEMENTS</p><p dir="ltr">DFRS/Forest Resource Assessment Nepal Project for field data, computing facilities and eCognition software; NASA/USGS for Landsat 8 data.</p><p dir="ltr">RELATED MATERIAL</p><p dir="ltr">Khanal, S. Integration of object-based image analysis with machine learning algorithm for forest type classification in Nepal. Poster, Asia-Pacific Network for Global Change Research (APN) [event, year]. https://www.apn-gcr.org/wp-content/uploads/2020/09/20IGM-NS06-Khanal.pdf</p> 2026-10-01T01:31:31Z Dataset Dataset 10.6084/m9.figshare.34037553.v1 https://figshare.com/articles/dataset/National_forest_type_classification_of_Nepal_15_forest_types_at_30_m_from_Landsat_8_2013_14_/34037553 CC BY 4.0