Production workshop on development of national land cover data for Afghanistan

The service aims to develop an operational system for annual land cover mapping and change analysis using a standard methodology and consistent datasets. It is a collaborative effort between SERVIR-HKH at ICIMOD and SERVIR-Mekong at the Asian Disaster Preparedness Center, with additional support from the United States Forest Services (USFS) and SilvaCarbon. The RLCMS uses the architecture of cloud- based, state-of-the-art remote sensing science and technology, such as Google Earth Engine, and a standard set of input data sources to generate high-quality land cover maps on a regular basis at the regional and national levels. The National Land Cover Monitoring Systems (NLCMS) for Afghanistan will generate consistent land cover data at the national level which can also be harmonized at the regional level in the RLCMS.In Afghanistan, government agencies use land cover data for agriculture, the national forest inventory, and other forest management purposes. National-level land cover assessments were carried out in 1972, 1993, and 2010 using satellite data from different sensors; each assessment used different land cover classes/legends. No land cover assessments at the national level have been conducted.

In this aspect, the SERVIR-HKH Initiative and partners are now convening a production workshop focused on building team capacity and refining algorithms for deriving key land cover “primitive” elements, devising methods for assembling the data primitives into target land cover classes, and formulating explicit accuracy assessment procedures. The five-day event will train partner teams in the use and implementation of primitive development, data assembly algorithms, accuracy assessment procedures and implications, and documentation of how these elements fit within the overall system.

Objectives

• Review training and validation samples collected for Afghanistan and refine procedures for creating primitive classes
• Use Collect Earth Online (CEO) to target the concentration of additional reference data to be used to improve the performance of the primitive classifier
• Design, test, and refine procedures for assembling data primitives and other information sources into explicit land cover classes
• Specify appropriate accuracy assessment procedures and products that will provide critical context to data users; finalize validation protocols and develop a project plan to finalize validation data collection in CEO
• Refine the logic and work plan for the NLCMS to meet identified end-user needs

Expected Outcomes

• A set of explicit procedures and tools for combining the data primitives and (potentially) other information into final land cover classes as set out in the agreed national typology
• Specification of reference data needs and procedures to support (potentially iterative) accuracy assessment of land cover products produced by the NLCMS
• A draft version of land cover for Afghanistan

Prerequisites

• Google Earth Engine account
• CEO data collected from Afghanistan
• Computer with Chrome browser installed
• Reference data and land cover maps