Continuous monitoring of land change activities and post-disturbance dynamics from Landsat time series: A test methodology for REDD+ reporting

Highlights

• Presented methods allow for monitoring activities and post-disturbance landscapes.

• Deforestation, driven by conversion to pastures, is increasing at very small rate.

• Less than a fifth of the area of deforestation was abandoned and left to regenerate.

• Using a buffer stratum around change areas increased precision in area estimates.

• Samples representing each period for which area estimates are desired were required.

Abstract

The REDD+ mechanism of UNFCCC was established to reduce greenhouse gases emissions by means of financial incentives. Of importance to the success of REDD+ and similar initiatives is the provision of credible evidence of reductions in the extent of land change activities that release carbon to the atmosphere (e.g. deforestation). The criteria for reporting land change areas and associated emissions within REDD+ stipulate the use of sampling-based approaches, which allow for unbiased estimation and uncertainty quantification. But for economic compensation for emission reductions to be feasible, agreements between participating countries and donors often require reporting every year or every second year. With the rates of land change typically being very small relative to the total study area, sampling-based approaches for estimation of annual or bi-annual areas have proven problematic, especially when comparing area estimates over time. In this paper, we present a methodology for monitoring and estimating areas of land change activity at high temporal resolution that is compliant with international guidelines. The methodology is based on a break detection algorithm applied to time series of Landsat data in the Colombian Amazon between 2001 and 2016. A biennial stratified sampling approach was implemented to (1) remove the bias introduced by the change detection and classification algorithm in mapped areas derived from pixel-counting; and (2) provide confidence intervals for area estimates obtained from the reference data collected for the sample. Our results show that estimating the area of land change, like deforestation, at annual or bi-annual resolution is inherently challenging and associated with high degrees of uncertainty. We found that better precision was achieved if independent sample datasets of reference observations were collected for each time interval for which area estimates are required. The alternative of selecting one sample of continuous reference observations analyzed for inference of area for each time interval did not yield area estimates significantly different from zero. Also, when large stable land covers (primary forest in this case, occupying almost 90% of the study area) are present in the study area in combination with small rates of land change activity, the impact of omission errors in the map used for stratifying the study area will be substantial and potentially detrimental to usefulness of land change studies. The introduction of a buffer stratum around areas of mapped land change reduced the uncertainty in area estimates by up to 98%. Results indicate that the Colombian Amazon has experienced a small but steady decrease in primary forest due to establishment of pastures, with forest-to-pasture conversion reaching 103 ± 30 kha (95% confidence interval) in the period between 2013 and 2015, corresponding to 0.22% of the study area. Around 29 ± 17 kha (95% CI) of pastureland that had been abandoned shortly after establishment reverted to secondary forest within the same period. Other gains of secondary forest from more permanent pastures averaged about 12 ± 11 kha (95% CI), while losses of secondary forest averaged 20 ± 12 kha (95% CI).

Introduction

Current tropical deforestation has been estimated to account for 7–14% of the annual CO₂ emissions released into the atmosphere by human activities whereas intact tropical primary forests sequester an equal amount (Achard et al., 2014; Goetz et al., 2015; Harris et al., 2012; Houghton et al., 2012). However, recent research suggests that a reduction in carbon density of tropical primary forest due to disturbance exceeds the emissions from deforestation, with the result that tropical forests are becoming a net source of carbon to the atmosphere (Baccini et al., 2017). The need for a reduction of emissions is thus more urgent than ever. Efforts to reduce global deforestation have led to the establishment of international frameworks like the United Nations Programme on Reducing Emissions from Deforestation and Forest Degradation (UN-REDD, 2016) that stipulate financial incentives to countries for reducing carbon emissions from tropical deforestation and forest degradation. For such frameworks to be successful, robust approaches that provide estimates of carbon emissions and removals with proper uncertainty metrics are required (IPCC, 2003). Methods to estimate carbon emissions and removals in the tropics typically rely on a gain/loss approach in which emission factors (i.e. carbon content per unit area per land cover type) and area of land change activities (i.e. areal extent of human activities that cause emission or removal of carbon such as deforestation, also called activity data) are multiplied (GFOI, 2016). Depending on the quantity of information required, and the degree of analytical complexity, the Intergovernmental Panel on Climate Change (IPCC) guidelines classifies the methodological approaches into three different Tiers: Tier 1, or the “default method”, relies on default emission factors data while Tier 2 requires country-specific emission factors; at Tier 3, higher-order methods typically include models and data that address national circumstances, and pixel- or stand-level tracking of land change activity over time (IPCC, 2003; GFOI, 2016). For representation of land areas and changes in area and condition, the IPCC identifies three approaches: Approach 1 does not include any direct data on land activities but simply country-scale area estimates of land categories at different times; Approach 2 requires a land change matrix, but without a spatial representation of the change; while Approach 3 requires a spatially and temporally explicit representation of land categories and conversions (GFOI, 2016). Following the Cancun Agreement of the United Nations Framework Convention on Climate Change (UNFCCC), countries that wish to report carbon emissions and removals under the requirements of IPCC guidelines need to create a system for Measurement, Reporting and Verification (MRV) for communication of the mitigation procedures and estimation approaches (UNFCCC, 2018). The national MRV system includes approaches for national forest monitoring in accordance with the IPCC Tier system (IPCC, 2006).

While tropical deforestation and associated carbon emissions have been extensively studied during the last three decades (Achard et al., 2002; Baccini et al., 2012; Brown, 1997; DeFries et al., 2002; FAO, 1993; Hansen et al., 2013), the last couple of years have witnessed remarkable developments in environmental remote sensing. The opening of the Landsat archive in 2008 (Woodcock et al., 2008) has allowed for production of global maps of forest cover change (Hansen et al., 2013; Kim, 2010) and time series analysis of satellite data to study changes on the land surface (see for example Kennedy et al., 2010; Verbesselt et al., 2010; Zhu and Woodcock, 2014a, Zhu and Woodcock, 2014b). New missions with global acquisition strategies and free data policies are already in orbit (Sentinel-2A, -2B and Landsat-8) and more are forthcoming (Landsat-9, -10 and Sentinel-2C, -2D). In addition, statistical protocols for unbiased estimation of area have become an integral part of forest and land cover monitoring (McRoberts, 2011; Olofsson et al., 2013; Stehman, 2013). Together, these advancements enable a more comprehensive analysis of land change that meets the highest requirements of IPCC for land representation. Still, there are relatively few studies in the scientific literature focused on the use of these methods for advancing operational forest monitoring in MRV systems. Notable exceptions are the Guyana MRV system that conforms to the IPCC Approach 3 for multiple land cover classes (GFOI, 2016); the national forest monitoring system of Peru that employs Landsat-based time series analysis and unbiased estimation of forest cover change (Potapov et al., 2014); the PRODES system of Brazil (Instituto Nacional de Pesquisas Espaciais (INPE), 2016) based on manual interpretation of Landsat imagery; and the Mexican MAD-MEX system (Gebhardt et al., 2014) that uses time-series analysis, segmentation and approaches for statistical inference. Colombia has experienced an increase in forest monitoring capacity with a Government agency (Instituto de Hidrología, Meteorología y Estudios Ambientales, IDEAM) dedicated to the establishment of a forest monitoring system (IDEAM, 2016). The Colombian system is built upon good practices in remote sensing and sampling-based estimation, including stratified estimation and implementation of new algorithms that make use of the Landsat archive. The aforementioned forest monitoring systems are impressive and have provided valuable information on the state of tropical forests. Still, what is missing is a system that tracks the conversions between the six IPCC land categories, including the dynamics of post-disturbance landscapes, at high temporal and spatial resolution, coupled with unbiased estimation protocols for provision of biennial estimates of activity data.

In this paper we test a methodology for continuous monitoring and estimation of areas of land cover and land change that is compliant with IPCC Approach 3 for representation of land. The methodology builds on recent advancements in the field of environmental remote sensing, using algorithms for time series analysis (Zhu and Woodcock, 2014a) and estimation protocols (Olofsson et al., 2014; Stehman, 2013). The performance of the methodology is tested for the Colombian Amazon between 2001 and 2016.