Mapping forest dynamics in Bangladesh from satellite images: implications for the global REDD+ program

Abstract

Reducing emissions from deforestation and forest degradation (REDD+) is a global climate change mitigation initiative under negotiation by the United Nations Framework Convention on Climate Change (UNFCCC). This initiative provides financial incentives to developing countries for enhancing carbon stocks in their forests by abstaining from deforestation and forest degradation, which would lead to emission of CO2 into the atmosphere. To implement a REDD+ project, developing countries need to measure and monitor anthropogenic changes in their forests and as well as forest carbon to account for CO2 emissions and removals from such changes. It is also important to predict future scenarios of forest/landuse changes and to identify the areas at risk of changes so that appropriate conservation initiatives can be taken. Bangladesh is steadily progressing through its REDD+ roadmap. However, several key research issues still to address include: using remote sensing technology to detect deforestation, forest degradation and associated changes in forest carbon stock, and predict future forest/land-use pattern at local environmental setting. This study developed and evaluated approaches to detect the extent of historic (1995-2015) deforestation and forest degradation (Objective 1), to estimate the emissions of forest aboveground carbon due to deforestation and degradation activities (Objective 2), and to assess the future scenarios, driving forces and risk of deforestation and forest degradation (Objective 3) at Raghunandan Hill Reserve in northeastern Bangladesh using a combination of satellite and field-level data employing various geospatial, statistical and modeling tools. The thesis comprises seven chapters. The significance and background of the research along with aim, objectives, research questions and available literature on the topic were described in chapter 1, 2 and 3. Chapter 4 dealt with the first objective, where changes of forest areas to non-forest areas and one forest strata to another during 1995-2015 were detected with high accuracy (>90%) by applying Monte-Carlo spectral unmixing algorithm to Landsat images, followed by knowledge-based classification approach. The classification was verified using independent randomly drawn reference sample points using high-resolution Google Earth images. A post-classification comparison method was applied to quantify the spatial extent, location and rate of changes of forest classes by generating transition matrices. In chapter 5, emissions of carbon (t CO2e yr-1) from deforestation and forest degradation activities during 1995-2005 and 2005-2015 periods at Raghunandan forest were estimated using Landsat satellite and field-level biomass-carbon data applying regression analysis and geospatial techniques with acceptable accuracy. The accuracy of the estimate was verified using root mean square error iii (RMSE) of estimated carbon with field reference carbon. Results indicated that, during 1995-2005, the total estimated emission was 4589 tonnes (t) CO2e yr-1 from deforesta...

Targeting area and comparing the effect of different land use/land cover (LULC) scenarios on greenhouse gases (GHGs) emission reduction (Case study: Hyrcanian forests in Iran)

“…Because the emissions of GHGs are known to be strongly influenced by LULC change (Cooper et al, 2020;Hundera et al, 2020), scenario analysis with LULC models can play a major role in providing information to decision makers. By purpose of spatial targeting of REDD studies, LULC change models such as Land Change Modeler (LCM), Geomod, CA-MARCOV and CLUE-S have been used for the LULC changes prediction, specially forest loss trend (Feng et al, 2020;Tang et al, 2020;Parsamehr et al, 2019;Mena et al, 2017;Bununu et al, 2016;Kim, 2010;Hewson et al, 2019;Redowan, 2019). Some models such as LCM that permits the simulation of future scenario is integrated with a REDD steps to determine and model anthropogenic GHGs emission reductions (Bununu et al, 2016).…”