ALBERT

Description: Tropical forests store 40–50 per cent of terrestrial vegetation carbon1. However, spatial variations in aboveground live tree biomass carbon (AGC) stocks remain poorly understood, in particular in tropical montane forests2. Owing to climatic and soil changes with increasing elevation3, AGC stocks are lower in tropical montane forests compared with lowland forests2. Here we assemble and analyse a dataset of structurally intact old-growth forests (AfriMont) spanning 44 montane sites in 12 African countries. We find that montane sites in the AfriMont plot network have a mean AGC stock of 149.4 megagrams of carbon per hectare (95% confidence interval 137.1–164.2), which is comparable to lowland forests in the African Tropical Rainforest Observation Network4 and about 70 per cent and 32 per cent higher than averages from plot networks in montane2,5,6 and lowland7 forests in the Neotropics, respectively. Notably, our results are two-thirds higher than the Intergovernmental Panel on Climate Change default values for these forests in Africa8. We find that the low stem density and high abundance of large trees of African lowland forests4 is mirrored in the montane forests sampled. This carbon store is endangered: we estimate that 0.8 million hectares of old-growth African montane forest have been lost since 2000. We provide country-specific montane forest AGC stock estimates modelled from our plot network to help to guide forest conservation and reforestation interventions. Our findings highlight the need for conserving these biodiverse9,10 and carbon-rich ecosystems.

Description: One of the most fundamental questions in ecology is how many species inhabit the Earth. However, due to massive logistical and financial challenges and taxonomic difficulties connected to the species concept definition, the global numbers of species, including those of important and well-studied life forms such as trees, still remain largely unknown. Here, based on global ground-sourced data, we estimate the total tree species richness at global, continental, and biome levels. Our results indicate that there are ∼73,000 tree species globally, among which ∼9,000 tree species are yet to be discovered. Roughly 40% of undiscovered tree species are in South America. Moreover, almost one-third of all tree species to be discovered may be rare, with very low populations and limited spatial distribution (likely in remote tropical lowlands and mountains). These findings highlight the vulnerability of global forest biodiversity to anthropogenic changes in land use and climate, which disproportionately threaten rare species and thus, global tree richness.

Description: Two novel satellite LiDAR missions —GEDI and ICESat-2— are currently operational and combined provide near-global measurements of forest height and structure. Such data underpin a new era of large-area approaches for measuring forest height in regrowing forests of different ages and assessing associated regrowth rates. Two LiDAR missions further allow for comparing independently derived forest heights and regrowth rates. This study utilized both GEDI and ICESat-2 measurements to assess regrowth rates in regrowing forests of different ages for the Brazilian state Rondônia. We considered 19 data subgroups stratified by beam strength, light condition, beam sensitivity, and waveform processing algorithm to assess the retrieval uncertainty and identify data subgroups associated with the most reliable regrowth estimates. The quality assessment of GEDI and ICESat-2 forest heights over four 50 km long airborne LiDAR strips determined a root mean square error of 4.14 m (CV = 17%) and 5.91 m (CV = 19%) and a mean error of 0.04 m and −2.81 m, respectively. A linear calibration model between satellite- and airborne-LiDAR heights was then derived for each data subgroup and used to calibrate satellite heights. Forest regrowth rates were subsequently estimated for each satellite mission using a space-for-time imputation with forest heights’ medians per stand age class. The total growth of GEDI and ICESat-2 median forest heights after 33 years was 20.17 m (SE = 1.3 m) and 20.13 m (SE = 2.8 m), respectively. However, when growth was approximated with different non-linear models, the total growth differed by up to 6%, and the average regrowth rate even by up to 23%. The study revealed that omitting either the calibration step or the removal of secondary-forest-border pixels would result in an underestimation of the regrowth rate by more than 20%. Furthermore, the ICESat-2 weak beams were found unreliable for regrowth retrieval. The study showed that the novel satellite LiDAR data and the proposed methods could assess median forest height growth over large areas. However, forest age errors should also be accounted for in the retrieval uncertainty before comparing the growth estimates across different regions. Code and data necessary to reproduce the results are freely available on GitHub and Zenodo.

Description: Comparing the performance of different satellite sensors in global land cover change (LCC) monitoring is necessary to assess their potential and limitations for more accurate and operational LCC estimations. This paper aims to examine and compare the performance in LCC monitoring using three satellite sensors: PROBA-V, Landsat 8 OLI, and Sentinel-2 MSI. We utilized a unique set of global reference data containing four years of records (2015–2018) at 29,263 land cover change/no-change 100 × 100-m sites. The LCC monitoring was conducted using the BFAST(s)-Random Forest (BRF) change detection framework involving 15 global timeseries vegetation indices and three BFAST models. Due to the different spectral characteristics and data availability of the sensors, we designed 30 comparison scenarios to extensively evaluate their performance. The overall results were: 1) for global general LCC monitoring, Landsat 8 OLI slightly outperformed Sentinel-2, and PROBA-V performed the worst. The performance among the three sensors differed consistently despite different data availability and spectral observation regions. Sentinel-2 was more competitive with Landsat 8 when the red-edge 1 band was included; 2) Landsat 8 was more accurate in forest, herbaceous vegetation, and water monitoring. Sentinel-2 performed particularly well in wetland monitoring. In addition, we further observed: 3) missing data in time series decreased the accuracy in all sensors, but had little influence on the relative performance across sensors; 4) combining sensors would not necessarily improve the accuracy because the complementary effects enhanced the accuracy only when there was a large amount of data missing for all sensors; 5) the BRF framework maintained the performance gap among sensors, but obtained a higher and more balanced accuracy overall when compared with using BFAST methods alone, by involving ensemble learning with an embedded sample-balancing strategy; 6) among the random forest variables, the ‘magnitude’ proved to be the most important contributor, and the NDVI had the most consistently good performance across sensors when compared against other vegetation indices. All sensors using BRF still had some errors in change detection, with a tendency to underestimate the global LCC. A potential reason for this is the complexity of the diverse change/no-change characteristics at the global extent and the fact that smaller, more subtle LCCs might not be well detected. These limitations could be addressed by taking advantage of ensemble learning approaches with a combination of multiple independent region/thematic-adapted LCC monitoring models and using the original Sentinel-2 (10 m) and Landsat 8 (30 m) in the future.

Description: Several forest change detection algorithms are available for tracking and quantifying deforestation based on dense Landsat and Sentinel time series satellite data. Only few also capture regrowth after clearing in an accurate and continuous way across a diversity of forest types (including dry and seasonal forests) and are thus suitable to address the need for better information on secondary forest succession and for assessing forest restoration activities. We present a new change detection algorithm that makes use of the flexibility of kernel density estimations to create a forest reference phenology, taking into account all historical phenological variations of the forest rather than smoothing these out by curve fitting. The AVOCADO (Anomaly Vegetation Change Detection) algorithm allows detection of anomalies with a spatially explicit likelihood measure. We demonstrate the flexibility of the algorithm for three contrasting sites using all available Landsat time series data; ranging from tropical rainforest to dry miombo forest ecosystems, with different time series data densities, and characterized by different forest change types (e.g. selective logging, shifting cultivation). We found that the approach produced in general high overall accuracies (〉 90%) across these varying conditions, but had lower accuracies in the dry forest site with a slight overestimation of disturbances and regrowth. The latter was due to the similarity of crops in the time series NDMI signal, causing false regrowth detections. In the moist forest site the low producer accuracies in the intact forest and regrowth class was due to its very small area class (most forest disappeared by the nineties). We showed that the algorithm is capable of capturing small-scale (gradual) changes (e.g. selective logging, forest edge logging) and the multiple changes associated to shifting cultivation. The performance of the algorithm has been shown at regional scale, but if larger scale studies are required a representative selection of reference forest types need to be selected carefully. The outputs of the change maps allow the estimation of the spatio-temporal trends in the proportions of intact forest, secondary forest and non-forest - information that is useful for assessing the areas and potential of secondary forests to accumulate carbon and forest restoration targets. The algorithm can be used for disturbance and regrowth monitoring in different ecozones, is user friendly, and open source.

Description: For monitoring and reporting forest carbon stocks and fluxes, many countries in the tropics and subtropics rely on default values of forest aboveground biomass (AGB) from the Intergovernmental Panel on Climate Change (IPCC) guidelines for National Greenhouse Gas (GHG) Inventories. Default IPCC forest AGB values originated from 2006, and are relatively crude estimates of average values per continent and ecological zone. The 2006 default values were based on limited plot data available at the time, methods for their derivation were not fully clear, and no distinction between successional stages was made. As part of the 2019 Refinement to the 2006 IPCC Guidelines for GHG Inventories, we updated the default AGB values for tropical and subtropical forests based on AGB data from 〉25 000 plots in natural forests and a global AGB map where no plot data were available. We calculated refined AGB default values per continent, ecological zone, and successional stage, and provided a measure of uncertainty. AGB in tropical and subtropical forests varies by an order of magnitude across continents, ecological zones, and successional stage. Our refined default values generally reflect the climatic gradients in the tropics, with more AGB in wetter areas. AGB is generally higher in old-growth than in secondary forests, and higher in older secondary (regrowth 〉20 years old and degraded/logged forests) than in young secondary forests (⩽20 years old). While refined default values for tropical old-growth forest are largely similar to the previous 2006 default values, the new default values are 4.0–7.7-fold lower for young secondary forests. Thus, the refined values will strongly alter estimated carbon stocks and fluxes, and emphasize the critical importance of old-growth forest conservation. We provide a reproducible approach to facilitate future refinements and encourage targeted efforts to establish permanent plots in areas with data gaps.

Description: Community-based monitoring (CBM) is one of the- most sustainable ways of establishing a national forest monitoring system for successful Reduce Emissions from Deforestation and Forest Degradation (REDD+) implementation. In this research, we present the details of the National Forest Conservation Program (PNCB—Programa Nacional de Conservación de Bosques para la Mitigación del Cambio Climático), Peru, from a satellite-based alert perspective. We examined the community’s involvement in forest monitoring and investigated the usability of 1853 CBM data in conjunction with 445 satellite-based alerts. The results confirm that Peru’s PCNB contributed significantly to the REDD+ scheme, and that the CBM data provided rich information on the process and drivers of forest change. We also identified some of the challenges faced in the existing system, such as delays in satellite-based alert transfer to communities, sustaining community participation, data quality and integration, data flow, and standardization. Furthermore, we found that mobile devices responding to alerts provided better and faster data on land-use, and a better response rate, and facilitated a more targeted approach to monitoring. We recommend expanding training efforts and equipping more communities with mobile devices, to facilitate a more standardized approach to forest monitoring. The automation and unification of the alert data flow and incentivization of the participating communities could further improve forest monitoring and bridge the gap between near-real-time (NRT) satellite-based and CBM systems.

Description: Over the past decade, several global maps of above-ground biomass (AGB) have been produced, but they exhibit significant differences that reduce their value for climate and carbon cycle modelling, and also for national estimates of forest carbon stocks and their changes. The number of such maps is anticipated to increase because of new satellite missions dedicated to measuring AGB. Objective and consistent methods to estimate the accuracy and uncertainty of AGB maps are therefore urgently needed. This paper develops and demonstrates a framework aimed at achieving this. The framework provides a means to compare AGB maps with AGB estimates from a global collection of National Forest Inventories and research plots that accounts for the uncertainty of plot AGB errors. This uncertainty depends strongly on plot size, and is dominated by the combined errors from tree measurements and allometric models (inter-quartile range of their standard deviation (SD) = 30–151 Mg ha−1). Estimates of sampling errors are also important, especially in the most common case where plots are smaller than map pixels (SD = 16–44 Mg ha−1). Plot uncertainty estimates are used to calculate the minimum-variance linear unbiased estimates of the mean forest AGB when averaged to 0.1∘. These are used to assess four AGB maps: Baccini (2000), GEOCARBON (2008), GlobBiomass (2010) and CCI Biomass (2017). Map bias, estimated using the differences between the plot and 0.1∘ map averages, is modelled using random forest regression driven by variables shown to affect the map estimates. The bias model is particularly sensitive to the map estimate of AGB and tree cover, and exhibits strong regional biases. Variograms indicate that AGB map errors have map-specific spatial correlation up to a range of 50–104 km, which increases the variance of spatially aggregated AGB map estimates compared to when pixel errors are independent. After bias adjustment, total pantropical AGB and its associated SD are derived for the four map epochs. This total becomes closer to the value estimated by the Forest Resources Assessment after every epoch and shows a similar decrease. The framework is applicable to both local and global-scale analysis, and is available at https://github.com/arnanaraza/PlotToMap. Our study therefore constitutes a major step towards improved AGB map validation and improvement.

Description: Tropical deforestation continues at alarming rates with profound impacts on ecosystems, climate, and livelihoods, prompting renewed commitments to halt its continuation. Although it is well established that agriculture is a dominant driver of deforestation, rates and mechanisms remain disputed and often lack a clear evidence base. We synthesize the best available pantropical evidence to provide clarity on how agriculture drives deforestation. Although most (90 to 99%) deforestation across the tropics 2011 to 2015 was driven by agriculture, only 45 to 65% of deforested land became productive agriculture within a few years. Therefore, ending deforestation likely requires combining measures to create deforestation-free supply chains with landscape governance interventions. We highlight key remaining evidence gaps including deforestation trends, commodity-specific land-use dynamics, and data from tropical dry forests and forests across Africa.

Description: An increase in the frequency and severity of disturbances (such as forest fires) is putting pressure on the resilience of the Amazon tropical forest; potentially leading to reduced ability to recover and to maintain a functioning forest ecosystem. Dense and long-term satellite time series approaches provide a largely untapped data source for characterizing disturbance- recovery forest dynamics across large areas and varying types of forests and conditions. Although large-scale forest recovery capacity metrics have been derived from optical satellite image time series and validated over various ecosystems, their sensitivity to disturbance (e.g. disturbance magnitude, disturbance timing, and recovery time) and environmental data characteristics (e.g. noise magnitude, seasonality, and missing values) are largely unknown. This study proposes an open source simulation framework based on the characteristics of sampled original satellite image time series to (i) compare the reliability of recovery metrics, (ii) evaluate their sensitivity with respect to environmental and disturbance characteristics, and (iii) evaluate the effect of pre-processing techniques on the reliability of the recovery metrics for abrupt disturbances, such as fires, in the Amazon basin forests. The effect of three pre-processing techniques were evaluated: changing the temporal resolution, noise removal techniques (such as time series smoothing and segmenting), and using a varying time span after the disturbance to calculate recovery metrics. Here, reliability is quantified by comparing derived and theoretical values of the recovery metrics (RMSE and R2). From the three recovery metrics evaluated, the Year on Year Average (YrYr) and the Ratio of Eighty Percent (R80p) are more reliable than the Relative Recovery Index (RRI). Time series segmentation tends to improve the reliability of recovery metrics. Recovery metrics derived from temporal dense Landsat time series tend to show a higher reliability than those derived from time series aggregated to quarterly or annual values. Although the framework is demonstrated on Landsat time series of the Amazon tropical forest, it can be used to perform such test on other datasets and ecosystems.