180 likes | 322 Views
Development of a reference data base and validation concept for RS-based forest cover change products in Fiji.
E N D
Development of a reference data base and validation concept for RS-based forest cover change products in Fiji Johannes Eberenz1, Johannes Reiche1, Samuela Lagataki², Akosita Lewai², Wolf Forstreuter³1) Wageningen University, The Netherlands; 2) Fiji Forestry Department (FFD); 3) SOPAC, South Pacific Counsel, Fiji
Contents • Introduction • Forest cover change data • Methodology 3.1 Database design 3.2 Validation of RS-based products • Preliminary results • Discussion
1 Introduction: Context • Study Program: • Final year of MSc. Geo-Information Science, Wageningen University, The Netherlands • 2012 – 2014 • Focus on RS • Internship exchange: • Internship at Fiji Forestry Department • October 2013 – March 2014 • Supervised by Johannes Reiche • Funded by GIZ
1 Introduction: Motivation and Goal Final Product: Databasethat provides the framework to harmonize different reference datasets to assess forest change Validation concept to use this DB for assessing RS-derived changes (bi-temporal, yearly changes) • Reference data is crucial • For Fiji, different reference datasets are availbale • Problems: • Different formats • No common interface • Unknown quality
1 Introduction: Concept of a Reference Database Harvest & Planting maps Reference Database Quality checks Digitized VHR Validation of RS-products IntegrationHarmonizing Existing databases RS-based change products
2 Data: Different Change data sources Logging, harvest and reforestation datasets • Selective logging of indigenous forests (FFD) • National forest inventory (FFD) • Harvest and replantation (Fiji Pine Ltd., Fiji Hardwood Ltd.) • Change areas derived from multi-temporal very high resolution (VHR) datasets (e.g. WorldView) • All datasets require quality control for spatial and information consistency
2 Data: Example VHR images KOMPSAR example (Detail) 2008-09-03 2011-12-15 Panchromatic (1m) Multispectral (4m)
3.1 Methods: Database design Views Change ChangeArea Study Area Land Use ChangeType
3.1 Methods: Data Integration Digitized VHR Harvest & Planting maps • Check input data quality of test areas (comparison to RS-products, jointly with FFD) • Develop a workflow for different data sources, vector and raster data • Demo implementation with selected datasets Quality checks IntegrationHarmonizing Reference Database
3.2 Methods: Validation concept • Traditional methods: • Confusion matrix • Accuracy measures: Overall accuracy, Kappa, producers & users accuracy per class • Advanced methods: • Stratified estimation • Latent class analysis Reference Database Validation of RS-products RS-based change products
3.1 Methods: Demo Implementation • Simple Schema (Single Table) • Use available spatial data infrastructure: • SOPAC GeoNode • Local solution at forestry department • Stepwise integration of available data • Different data sources • Vector and raster data • Demonstrate use: • Demo validations • Workshop SOPAC
4 Prelimary Results: Demo Validation Reference data:Logging year form Fiji-Pine RS-based forest cover change product:Change year from Landsat time series
5 Discussion • Final Product: • Database that provides the framework to harmonize different reference datasets to assess forest change • Validation concept to use this DB for assessing RS-derived changes (bi-temporal, yearly changes) • Challenges: • Different data quality • Validation of selective logging • ...?
Vinaka! Contact: Johannes.eberenz@wur.nl