Remote sensing-based forest health monitoring systems – case studies from Czechia and Slovakia

Aim of this paper is to present the remote sensing-based systems of forest health assessment in the Czech Republic and Slovakia, and to analyse both their strengths and weaknesses. Nationwide assessment of forest health in the Czech Republic is based on the interpretation of Sentinel–2 satellite data using novel approaches for cloud-free image synthesis based on all available satellite observations. A predictive statistical model to yield time series of leaf area index (LAI) from satellite observations is developed above extensive in-situ data, including LAI and forest defoliation assessment. Forest health is evaluated for each pixel from yearly changes of forest LAI, while the country-wise assessment of the health status is performed at the cadastral level. Methodology developed for Slovakia is based on a two-phase regression sampling. The first phase of the procedure provides an initial fast estimate of forest damage using only satellite observations (visible and infrared channels from Landsat or Sentinel–2 systems). The second phase refines the result of the first phase using data from a ground damage assessment (site-level defoliation from ICP Forests database). Resulting forest health assessment over the whole forest area is presented in 10 defoliation classes. The Czech Republic shows 1.6% of heavily damaged forests, 12.5% of damaged forests, 79.2% of forests with stable conditions, 6.3% of regenerated forests and 0.4% of strongly regenerated forests. In Slovakia, the total share of damaged stands (i. e. with defoliation higher than 40%) increased from 6 – 8% in 2003 – 2011 to 13 – 15% in 2012 – 2017. Both methodologies conduct nationwide assessment of forest health status in a fast and automatized way with high accuracy and minimal costs. The weaknesses are, for example, a high computational demands for production cloud free mosaics, inability to identify initial phases of forest health decline, exclusion of stands older than 80 years (in the Czech Republic) and inability to differentiate between harvested and severely damaged stands (in Slovakia). Finally, the paper outlines future development of both methodologies.