Markedly divergent estimates of Amazon forest carbon density from ground plots and satellites

Deforestation in Latin America in the 2000s predominantly occurred outside of typical mature forests

The role of tropical forests in the global carbon budget remains controversial, as carbon emissions from deforestation are highly uncertain. This high uncertainty arises from the use of either fixed forest carbon stock density or maps generated from satellite-based optical reflectance with limited sensitivity to biomass to generate accurate estimates of emissions from deforestation. New space missions aiming to accurately map the carbon stock density rely on direct measurements of the spatial structures of forests using lidar and radar. We found that lost forests are special cases, and their spatial structures can be directly measured by combining archived data acquired before and after deforestation by space missions principally aimed at measuring topography. Thus, using biomass mapping, we obtained new estimates of carbon loss from deforestation ahead of forthcoming space missions. Here, using a high-resolution map of forest loss and the synergy of radar and lidar to estimate the aboveground biomass density of forests, we found that deforestation in the 2000s in Latin America, one of the severely deforested regions, mainly occurred in forests with a significantly lower carbon stock density than typical mature forests. Deforestation areas with carbon stock densities lower than 20.0, 50.0, and 100.0 Mg C/ha accounted for 42.1%, 62.0%, and 83.3% of the entire deforested area, respectively. The average carbon stock density of lost forests was only 49.13 Mg C/ha, which challenges the current knowledge on the carbon stock density of lost forests (with a default value 100 Mg C/ha according to the Intergovernmental Panel on Climate Change Tier 1 estimates, or approximately 112 Mg C/ha used in other studies). This is demonstrated over both the entire region and the footprints of the spaceborne lidar. Consequently, our estimate of carbon loss from deforestation in Latin America in the 2000s was 253.0 ± 21.5 Tg C/year, which was considerably less than existing remote-sensing-based estimates, namely 400-600 Tg C/year. This indicates that forests in Latin America were most likely not a net carbon source in the 2000s compared to established carbon sinks. In previous studies, considerable effort has been devoted to rectify the underestimation of carbon sinks; thus, the overestimation of carbon emissions should be given sufficient consideration in global carbon budgets. Our results also provide solid evidence for the necessity of renewing knowledge on the role of tropical forests in the global carbon budget in the future using observations from new space missions.

LiDAR-based reference aboveground biomass maps for tropical forests of South Asia and Central Africa

Accurate mapping and monitoring of tropical forests aboveground biomass (AGB) is crucial to design effective carbon emission reduction strategies and improving our understanding of Earth's carbon cycle. However, existing large-scale maps of tropical forest AGB generated through combinations of Earth Observation (EO) and forest inventory data show markedly divergent estimates, even after accounting for reported uncertainties. To address this, a network of high-quality reference data is needed to calibrate and validate mapping algorithms. This study aims to generate reference AGB datasets using field inventory plots and airborne LiDAR data for eight sites in Central Africa and five sites in South Asia, two regions largely underrepresented in global reference AGB datasets. The study provides access to these reference AGB maps, including uncertainty maps, at 100 m and 40 m spatial resolutions covering a total LiDAR footprint of 1,11,650 ha [ranging from 150 to 40,000 ha at site level]. These maps serve as calibration/validation datasets to improve the accuracy and reliability of AGB mapping for current and upcoming EO missions (viz., GEDI, BIOMASS, and NISAR).

A new circa 2007 biomass map for China differs significantly from existing maps

The forest area of China is the fifth largest of any country, and unlike in many other countries, in recent decades its area has been increasing. However, there are substantial differences in estimates of the amount of carbon this forest contains, ranging from 3.92 to 17.02 Pg C for circa 2007. This makes it unclear how the changes in China's forest area contribute to the global carbon cycle. We generate a circa 2007 aboveground biomass (AGB) map at a resolution of 50 m using optical, radar and LiDAR satellite data. Our estimates of total carbon stored in the forest in China was 9.52 Pg C, with an average forest AGB of 104 Mg ha-1. Compared with three existing AGB maps, our AGB map showed better correlation with a distributed set of forest inventory plots. In addition, our high resolution AGB map provided more details on spatial distribution of forest AGB, and is likely to help understand the carbon storage changes in China's forest.

Sensing Forests Directly: The Power of Permanent Plots

The need to measure, monitor, and understand our living planet is greater than ever. Yet, while many technologies are applied to tackle this need, one developed in the 19th century is transforming tropical ecology. Permanent plots, in which forests are directly sensed tree-by-tree and species-by-species, already provide a global public good. They could make greater contributions still by unlocking our potential to understand future ecological change, as the more that computational and remote technologies are deployed the greater the need to ground them with direct observations and the physical, nature-based skills of those who make them. To achieve this requires building profound connections with forests and disadvantaged communities and sustaining these over time. Many of the greatest needs and opportunities in tropical forest science are therefore not to be found in space or in silico, but in vivo, with the people, places and plots who experience nature directly. These are fundamental to understanding the health, predicting the future, and exploring the potential of Earth's richest ecosystems. Now is the time to invest in the tropical field research communities who make so much possible.

Evaluation of multifrequency SAR data for estimating tropical above-ground biomass by employing radiative transfer modeling

The retrieval of the biophysical parameters and subsequent estimation of the above-ground biomass (AGB) of vegetation stands are made possible by the simulation of the extinction and scattering components from the canopy layer using vector radiative transfer (VRT) theory-based scattering models. With the use of such a model, this study aims to evaluate and compare the potential of dual-pol, multi-frequency SAR data for estimating above-ground biomass. The data selected for this work are L-band dual polarized (HH/HV) ALOS-2 data, S-band dual polarized (HH/HV) NovaSAR data, and C-band dual polarized (VV/VH) Sentinel-1 data. The two key biophysical parameters, tree height, and trunk radius are retrieved using the proposed methodology, applying the frequencies independently. A general allometric equation with vegetation-specific coefficients is used to estimate the AGB from the retrieved biophysical parameters. The retrieval results are validated using ground truth measurements collected from the study area. The L-band, with the coefficient of determination ([Formula: see text]) of 0.73 and the root mean square error (RMSE) of 35.90 t/ha, has the best correlation between the modeled and field AGBs, followed by the S-band with an [Formula: see text] of 0.37 and an RMSE of 63.37 t/ha, and the C-band with an [Formula: see text] of 0.25 and an RMSE of 72.32 t/ha. The L-band has yielded improved estimates of AGB in regression analysis as well, with an [Formula: see text] of 0.48 and an RMSE of 50.02 t/ha, compared to the S- and C-bands, which have the [Formula: see text] of 0.12 and 0.03 and the RMSE of 70.98 t/ha and 80.84 t/ha, respectively.

Potential aboveground biomass increase in Brazilian Atlantic Forest fragments with climate change

Fragmented tropical forest landscapes preserve much of the remaining biodiversity and carbon stocks. Climate change is expected to intensify droughts and increase fire hazard and fire intensities, thereby causing habitat deterioration, and losses of biodiversity and carbon stock losses. Understanding the trajectories that these landscapes may follow under increased climate pressure is imperative for establishing strategies for conservation of biodiversity and ecosystem services. Here, we used a quantitative predictive modelling approach to project the spatial distribution of the aboveground biomass density (AGB) by the end of the 21st century across the Brazilian Atlantic Forest (AF) domain. To develop the models, we used the maximum entropy method with projected climate data to 2100, based on the Intergovernmental Panel on Climate Change Representative Concentration Pathway (RCP) 4.5 from the fifth Assessment Report. Our AGB models had a satisfactory performance (area under the curve > 0.75 and p value < .05). The models projected a significant increase of 8.5% in the total carbon stock. Overall, the projections indicated that 76.9% of the AF domain would have suitable climatic conditions for increasing biomass by 2100 considering the RCP 4.5 scenario, in the absence of deforestation. Of the existing forest fragments, 34.7% are projected to increase their AGB, while 2.6% are projected to have their AGB reduced by 2100. The regions likely to lose most AGB-up to 40% compared to the baseline-are found between latitudes 13° and 20° south. Overall, although climate change effects on AGB vary latitudinally for the 2071-2100 period under the RCP 4.5 scenario, our model indicates that AGB stocks can potentially increase across a large fraction of the AF. The patterns found here are recommended to be taken into consideration during the planning of restoration efforts, as part of climate change mitigation strategies in the AF and elsewhere in Brazil.

How drought events during the last century have impacted biomass carbon in Amazonian rainforests

During the last two decades, inventory data show that droughts have reduced biomass carbon sink of the Amazon forest by causing mortality to exceed growth. However, process-based models have struggled to include drought-induced responses of growth and mortality and have not been evaluated against plot data. A process-based model, ORCHIDEE-CAN-NHA, including forest demography with tree cohorts, plant hydraulic architecture and drought-induced tree mortality, was applied over Amazonia rainforests forced by gridded climate fields and rising CO₂ from 1901 to 2019. The model reproduced the decelerating signal of net carbon sink and drought sensitivity of aboveground biomass (AGB) growth and mortality observed at forest plots across selected Amazon intact forests for 2005 and 2010. We predicted a larger mortality rate and a more negative sensitivity of the net carbon sink during the 2015/16 El Niño compared with the former droughts. 2015/16 was indeed the most severe drought since 1901 regarding both AGB loss and area experiencing a severe carbon loss. We found that even if climate change did increase mortality, elevated CO₂ contributed to balance the biomass mortality, since CO₂ -induced stomatal closure reduces transpiration, thus, offsets increased transpiration from CO₂ -induced higher foliage area.

Relationships between species richness and ecosystem services in Amazonian forests strongly influenced by biogeographical strata and forest types

Despite increasing attention for relationships between species richness and ecosystem services, for tropical forests such relationships are still under discussion. Contradicting relationships have been reported concerning carbon stock, while little is known about relationships concerning timber stock and the abundance of non-timber forest product producing plant species (NTFP abundance). Using 151 1-ha plots, we related tree and arborescent palm species richness to carbon stock, timber stock and NTFP abundance across the Guiana Shield, and using 283 1-ha plots, to carbon stock across all of Amazonia. We analysed how environmental heterogeneity influenced these relationships, assessing differences across and within multiple forest types, biogeographic regions and subregions. Species richness showed significant relationships with all three ecosystem services, but relationships differed between forest types and among biogeographical strata. We found that species richness was positively associated to carbon stock in all biogeographical strata. This association became obscured by variation across biogeographical regions at the scale of Amazonia, resembling a Simpson’s paradox. By contrast, species richness was weakly or not significantly related to timber stock and NTFP abundance, suggesting that species richness is not a good predictor for these ecosystem services. Our findings illustrate the importance of environmental stratification in analysing biodiversity-ecosystem services relationships.

Integrating Multi-Source Remote Sensing to Assess Forest Aboveground Biomass in the Khingan Mountains of North-Eastern China Using Machine-Learning Algorithms

Forest aboveground biomass (AGB) is of great significance since it represents large carbon storage and may reduce global climate change. However, there are still considerable uncertainties in forest AGB estimates, especially in rugged regions, due to the lack of effective algorithms to remove the effects of topography and the lack of comprehensive comparisons of methods used for estimation. Here, we systematically compare the performance of three sources of remote sensing data used in forest AGB estimation, along with three machine-learning algorithms using extensive field measurements (N = 1058) made in the Khingan Mountains of north-eastern China in 2008. The datasets used were obtained from the LiDAR-based Geoscience Laser Altimeter System onboard the Ice, Cloud, and land Elevation satellite (ICESat/GLAS), the optical-based Moderate Resolution Imaging Spectroradiometer (MODIS), and the SAR-based Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR). We show that terrain correction is effective for this mountainous study region and that the combination of terrain-corrected GLAS and PALSAR features with Random Forest regression produces the best results at the plot scale. Including further MODIS-based features added little power for prediction. Based upon the parsimonious data source combination, we created a map of AGB circa 2008 and its uncertainty, which yields a coefficient of determination (R2) of 0.82 and a root mean squared error of 16.84 Mg ha−1 when validated with field data. Forest AGB values in our study area were within the range 79.81 ± 16.00 Mg ha−1, ~25% larger than a previous, SAR-based, analysis. Our result provides a historic benchmark for regional carbon budget estimation.

A 3D approach to model the taper of irregular tree stems: making plots biomass estimates comparable in tropical forests

In tropical forests, the high proportion of trees showing irregularities at the stem base complicates forest monitoring. For example, in the presence of buttresses, the height of the point of measurement (H_POM ) of the stem diameter (D_POM ) is raised from 1.3 m, the standard breast height, up to a regular part of the stem. While D_POM is the most important predictor for tree aboveground biomass (AGB) estimates, the lack of harmonized H_POM for irregular trees in forest inventory increases the uncertainty in plot-level AGB stock and stock change estimates. In this study, we gathered an original non-destructive three-dimensional (3D) data set collected with terrestrial laser scanning and close range terrestrial photogrammetry tools in three sites in central Africa. For the 228 irregularly shaped stems sampled, we developed a set of taper models to harmonize H_POM by predicting the equivalent diameter at breast height (DBH') from a D_POM measured at any height. We analyzed the effect of using DBH' on tree-level and plot-level AGB estimates. To do so, we used destructive AGB data for 140 trees and forest inventory data from eight 1-ha plots in the Republic of Congo. Our results showed that our best simple taper model predicts DBH' with a relative mean absolute error of 3.7% (R²  = 0.98) over a wide D_POM range of 17-249 cm. Based on destructive AGB data, we found that the AGB allometric model calibrated with harmonized H_POM data was more accurate than the conventional local and pantropical models. At the plot level, the comparison of AGB stock estimates with and without H_POM harmonization showed an increasing divergence with the increasing share of irregular stems (up to -15%). The harmonization procedure developed in this study could be implemented as a standard practice for AGB monitoring in tropical forests as no additional forest inventory measurements is required. This would probably lead to important revisions of the AGB stock estimates in regions having a large number of irregular tree stems and increase their carbon sink estimates. The growing use of three-dimensional (3D) data offers new opportunities to extend our approach and further develop general taper models in other tropical regions.

Assessing Black Locust Biomass Accumulation in Restoration Plantations

Forests (either natural or planted) play a key role in climate change mitigation due to their huge carbon-storing potential. In the 1980s, the Hellenic Public Power Corporation (HPPC) started the rehabilitation of lignite post-mining areas in Northwest Greece by planting mainly black locust (Robinia pseudoacacia L.). Today, these plantations occupy about 2570 ha, but the accumulation of Above Ground Biomass (AGB) and deadwood has not been assessed to date. Therefore, we aimed at estimating these biomass pools by calibrating an allometric model for AGB, performing an inventory for both pools and predicting the spatial distribution of AGB. 214 sample plots of 100 m2 each were set up through systematic sampling in a grid dimension of 500 × 500 m and tree dbh and height were recorded. AGB was estimated using an exponential allometric model and performing inventory measurements and was on average 57.6 t ha−1. Kriging analysis reliably estimated mean AGB, but produced errors in the prediction of high and low biomass values, related to the high fragmentation and heterogeneity of the studied area. Mean estimated AGB was low compared with European biomass yield tables for black locust. Similarly, standing deadwood was low (6–10%) and decay degrees were mostly 1 and 2, indicating recent deadwood formation. The overall low biomass accumulation in the studied black locust restoration plantations may be partially attributed to their young age (5–30 years old), but is comparable to that reported in black locust restoration plantation in extremely degraded sites. Thus, black locust successfully adapted to the studied depositions of former mines and its accumulated biomass has the potential to improve the carbon footprint of the region. However, the invasiveness of the species should be considered for future management planning of these restoration plantations.

Plant growth: the What, the How, and the Why

Growth is a widely used term in plant science and ecology, but it can have different meanings depending on the context and the spatiotemporal scale of analysis. At the meristem level, growth is associated with the production of cells and initiation of new organs. At the organ or plant scale and over short time periods, growth is often used synonymously with tissue expansion, while over longer time periods the increase in biomass is a common metric. At even larger temporal and spatial scales, growth is mostly described as net primary production. Here, we first address the question 'what is growth?'. We propose a general framework to distinguish between the different facets of growth, and the corresponding physiological processes, environmental drivers and mathematical formalisms. Based on these different definitions, we then review how plant growth can be measured and analysed at different organisational, spatial and temporal scales. We conclude by discussing why gaining a better understanding of the different facets of plant growth is essential to disentangle genetic and environmental effects on the phenotype, and to uncover the causalities around source or sink limitations of plant growth.

Changes in global terrestrial live biomass over the 21st century

Live woody vegetation is the largest reservoir of biomass carbon, with its restoration considered one of the most effective natural climate solutions. However, terrestrial carbon fluxes remain the largest uncertainty in the global carbon cycle. Here, we develop spatially explicit estimates of carbon stock changes of live woody biomass from 2000 to 2019 using measurements from ground, air, and space. We show that live biomass has removed 4.9 to 5.5 PgC year^-1 from the atmosphere, offsetting 4.6 ± 0.1 PgC year^-1 of gross emissions from disturbances and adding substantially (0.23 to 0.88 PgC year^-1) to the global carbon stocks. Gross emissions and removals in the tropics were four times larger than temperate and boreal ecosystems combined. Although live biomass is responsible for more than 80% of gross terrestrial fluxes, soil, dead organic matter, and lateral transport may play important roles in terrestrial carbon sink.

Worldwide Research on Land Use and Land Cover in the Amazon Region

Land cover is an important descriptor of the earth’s terrestrial surface. It is also crucial to determine the biophysical processes in global environmental change. Land-use change showcases the management of the land while revealing what motivated the alteration of the land cover. The type of land use can represent local economic and social benefits, framed towards regional sustainable development. The Amazon stands out for being the largest tropical forest globally, with the most extraordinary biodiversity, and plays an essential role in climate regulation. The present work proposes to carry out a bibliometric analysis of 1590 articles indexed in the Scopus database. It uses both Microsoft Excel and VOSviewer software for the evaluation of author keywords, authors, and countries. The method encompasses (i) search criteria, (ii) search and document compilation, (iii) software selection and data extraction, and (iv) data analysis. The results classify the main research fields into nine main topics with increasing relevance: ‘Amazon’, ‘deforestation’, ‘remote sensing’, ‘land use and land cover change’, and ‘land use’. In conclusion, the cocitation authors’ network reveals the development of such areas and the interest they present due to their worldwide importance.

Unveiling African rainforest composition and vulnerability to global change

Africa is forecasted to experience large and rapid climate change¹ and population growth² during the twenty-first century, which threatens the world's second largest rainforest. Protecting and sustainably managing these African forests requires an increased understanding of their compositional heterogeneity, the environmental drivers of forest composition and their vulnerability to ongoing changes. Here, using a very large dataset of 6 million trees in more than 180,000 field plots, we jointly model the distribution in abundance of the most dominant tree taxa in central Africa, and produce continuous maps of the floristic and functional composition of central African forests. Our results show that the uncertainty in taxon-specific distributions averages out at the community level, and reveal highly deterministic assemblages. We uncover contrasting floristic and functional compositions across climates, soil types and anthropogenic gradients, with functional convergence among types of forest that are floristically dissimilar. Combining these spatial predictions with scenarios of climatic and anthropogenic global change suggests a high vulnerability of the northern and southern forest margins, the Atlantic forests and most forests in the Democratic Republic of the Congo, where both climate and anthropogenic threats are expected to increase sharply by 2085. These results constitute key quantitative benchmarks for scientists and policymakers to shape transnational conservation and management strategies that aim to provide a sustainable future for central African forests.

Combination of Feature Selection and CatBoost for Prediction: The First Application to the Estimation of Aboveground Biomass

Increasing numbers of explanatory variables tend to result in information redundancy and “dimensional disaster” in the quantitative remote sensing of forest aboveground biomass (AGB). Feature selection of model factors is an effective method for improving the accuracy of AGB estimates. Machine learning algorithms are also widely used in AGB estimation, although little research has addressed the use of the categorical boosting algorithm (CatBoost) for AGB estimation. Both feature selection and regression for AGB estimation models are typically performed with the same machine learning algorithm, but there is no evidence to suggest that this is the best method. Therefore, the present study focuses on evaluating the performance of the CatBoost algorithm for AGB estimation and comparing the performance of different combinations of feature selection methods and machine learning algorithms. AGB estimation models of four forest types were developed based on Landsat OLI data using three feature selection methods (recursive feature elimination (RFE), variable selection using random forests (VSURF), and least absolute shrinkage and selection operator (LASSO)) and three machine learning algorithms (random forest regression (RFR), extreme gradient boosting (XGBoost), and categorical boosting (CatBoost)). Feature selection had a significant influence on AGB estimation. RFE preserved the most informative features for AGB estimation and was superior to VSURF and LASSO. In addition, CatBoost improved the accuracy of the AGB estimation models compared with RFR and XGBoost. AGB estimation models using RFE for feature selection and CatBoost as the regression algorithm achieved the highest accuracy, with root mean square errors (RMSEs) of 26.54 Mg/ha for coniferous forest, 24.67 Mg/ha for broad-leaved forest, 22.62 Mg/ha for mixed forests, and 25.77 Mg/ha for all forests. The combination of RFE and CatBoost had better performance than the VSURF–RFR combination in which random forests were used for both feature selection and regression, indicating that feature selection and regression performed by a single machine learning algorithm may not always ensure optimal AGB estimation. It is promising to extending the application of new machine learning algorithms and feature selection methods to improve the accuracy of AGB estimates.

New insights into large tropical tree mass and structure from direct harvest and terrestrial lidar

A large portion of the terrestrial vegetation carbon stock is stored in the above-ground biomass (AGB) of tropical forests, but the exact amount remains uncertain, partly owing to the lack of measurements. To date, accessible peer-reviewed data are available for just 10 large tropical trees in the Amazon that have been harvested and directly measured entirely via weighing. Here, we harvested four large tropical rainforest trees (stem diameter: 0.6-1.2 m, height: 30-46 m, AGB: 3960-18 584 kg) in intact old-growth forest in East Amazonia, and measured above-ground green mass, moisture content and woody tissue density. We first present rare ecological insights provided by these data, including unsystematic intra-tree variations in density, with both height and radius. We also found the majority of AGB was usually found in the crown, but varied from 42 to 62%. We then compare non-destructive approaches for estimating the AGB of these trees, using both classical allometry and new lidar-based methods. Terrestrial lidar point clouds were collected pre-harvest, on which we fitted cylinders to model woody structure, enabling retrieval of volume-derived AGB. Estimates from this approach were more accurate than allometric counterparts (mean tree-scale relative error: 3% versus 15%), and error decreased when up-scaling to the cumulative AGB of the four trees (1% versus 15%). Furthermore, while allometric error increased fourfold with tree size over the diameter range, lidar error remained constant. This suggests error in these lidar-derived estimates is random and additive. Were these results transferable across forest scenes, terrestrial lidar methods would reduce uncertainty in stand-scale AGB estimates, and therefore advance our understanding of the role of tropical forests in the global carbon cycle.

Model-Based Estimation of Amazonian Forests Recovery Time after Drought and Fire Events

In recent decades, droughts, deforestation and wildfires have become recurring phenomena that have heavily affected both human activities and natural ecosystems in Amazonia. The time needed for an ecosystem to recover from carbon losses is a crucial metric to evaluate disturbance impacts on forests. However, little is known about the impacts of these disturbances, alone and synergistically, on forest recovery time and the resulting spatiotemporal patterns at the regional scale. In this study, we combined the 3-PG forest growth model, remote sensing and field derived equations, to map the Amazonia-wide (3 km of spatial resolution) impact and recovery time of aboveground biomass (AGB) after drought, fire and a combination of logging and fire. Our results indicate that AGB decreases by 4%, 19% and 46% in forests affected by drought, fire and logging + fire, respectively, with an average AGB recovery time of 27 years for drought, 44 years for burned and 63 years for logged + burned areas and with maximum values reaching 184 years in areas of high fire intensity. Our findings provide two major insights in the spatial and temporal patterns of drought and wildfire in the Amazon: (1) the recovery time of the forests takes longer in the southeastern part of the basin, and, (2) as droughts and wildfires become more frequent—since the intervals between the disturbances are getting shorter than the rate of forest regeneration—the long lasting damage they cause potentially results in a permanent and increasing carbon losses from these fragile ecosystems.

An Evaluation of Eight Machine Learning Regression Algorithms for Forest Aboveground Biomass Estimation from Multiple Satellite Data Products

This study provided a comprehensive evaluation of eight machine learning regression algorithms for forest aboveground biomass (AGB) estimation from satellite data based on leaf area index, canopy height, net primary production, and tree cover data, as well as climatic and topographical data. Some of these algorithms have not been commonly used for forest AGB estimation such as the extremely randomized trees, stochastic gradient boosting, and categorical boosting (CatBoost) regression. For each algorithm, its hyperparameters were optimized using grid search with cross-validation, and the optimal AGB model was developed using the training dataset (80%) and AGB was predicted on the test dataset (20%). Performance metrics, feature importance as well as overestimation and underestimation were considered as indicators for evaluating the performance of an algorithm. To reduce the impacts of the random training-test data split and sampling method on the performance, the above procedures were repeated 50 times for each algorithm under the random sampling, the stratified sampling, and separate modeling scenarios. The results showed that five tree-based ensemble algorithms performed better than the three nonensemble algorithms (multivariate adaptive regression splines, support vector regression, and multilayer perceptron), and the CatBoost algorithm outperformed the other algorithms for AGB estimation. Compared with the random sampling scenario, the stratified sampling scenario and separate modeling did not significantly improve the AGB estimates, but modeling AGB for each forest type separately provided stable results in terms of the contributions of the predictor variables to the AGB estimates. All the algorithms showed forest AGB were underestimated when the AGB values were larger than 210 Mg/ha and overestimated when the AGB values were less than 120 Mg/ha. This study highlighted the capability of ensemble algorithms to improve AGB estimates and the necessity of improving AGB estimates for high and low AGB levels in future studies.

Approaches of Satellite Remote Sensing for the Assessment of Above-Ground Biomass across Tropical Forests: Pan-tropical to National Scales

Tropical forests are acknowledged for providing important ecosystem services and are renowned as “the lungs of the planet Earth” due to their role in the exchange of gasses—particularly inhaling CO2 and breathing out O2—within the atmosphere. Overall, the forests provide 50% of the total plant biomass of the Earth, which accounts for 450–650 PgC globally. Understanding and accurate estimates of tropical forest biomass stocks are imperative in ascertaining the contribution of the tropical forests in global carbon dynamics. This article provides a review of remote-sensing-based approaches for the assessment of above-ground biomass (AGB) across the tropical forests (global to national scales), summarizes the current estimate of pan-tropical AGB, and discusses major advancements in remote-sensing-based approaches for AGB mapping. The review is based on the journal papers, books and internet resources during the 1980s to 2020. Over the past 10 years, a myriad of research has been carried out to develop methods of estimating AGB by integrating different remote sensing datasets at varying spatial scales. Relationships of biomass with canopy height and other structural attributes have developed a new paradigm of pan-tropical or global AGB estimation from space-borne satellite remote sensing. Uncertainties in mapping tropical forest cover and/or forest cover change are related to spatial resolution; definition adapted for ‘forest’ classification; the frequency of available images; cloud covers; time steps used to map forest cover change and post-deforestation land cover land use (LCLU)-type mapping. The integration of products derived from recent Synthetic Aperture Radar (SAR) and Light Detection and Ranging (LiDAR) satellite missions with conventional optical satellite images has strong potential to overcome most of these uncertainties for recent or future biomass estimates. However, it will remain a challenging task to map reference biomass stock in the 1980s and 1990s and consequently to accurately quantify the loss or gain in forest cover over the periods. Aside from these limitations, the estimation of biomass and carbon balance can be enhanced by taking account of post-deforestation forest recovery and LCLU type; land-use history; diversity of forest being recovered; variations in physical attributes of plants (e.g., tree height; diameter; and canopy spread); environmental constraints; abundance and mortalities of trees; and the age of secondary forests. New methods should consider peak carbon sink time while developing carbon sequestration models for intact or old-growth tropical forests as well as the carbon sequestration capacity of recovering forest with varying levels of floristic diversity.

Spatial validation reveals poor predictive performance of large-scale ecological mapping models

Mapping aboveground forest biomass is central for assessing the global carbon balance. However, current large-scale maps show strong disparities, despite good validation statistics of their underlying models. Here, we attribute this contradiction to a flaw in the validation methods, which ignore spatial autocorrelation (SAC) in data, leading to overoptimistic assessment of model predictive power. To illustrate this issue, we reproduce the approach of large-scale mapping studies using a massive forest inventory dataset of 11.8 million trees in central Africa to train and validate a random forest model based on multispectral and environmental variables. A standard nonspatial validation method suggests that the model predicts more than half of the forest biomass variation, while spatial validation methods accounting for SAC reveal quasi-null predictive power. This study underscores how a common practice in big data mapping studies shows an apparent high predictive power, even when predictors have poor relationships with the ecological variable of interest, thus possibly leading to erroneous maps and interpretations.

Mapping ecological variables using machine-learning algorithms based on remote-sensing data has become a widespread practice in ecology. Here, the authors use forest biomass mapping as a study case to show that the most common model validation approach, which ignores data spatial structure, leads to overoptimistic assessment of model predictive power.

Fusion of Multiple Gridded Biomass Datasets for Generating a Global Forest Aboveground Biomass Map

Many advanced satellite estimation methods have been developed, but global forest aboveground biomass (AGB) products remain largely uncertain. In this study, we explored data fusion techniques to generate a global forest AGB map for the 2000s at 0.01-degree resolution with improved accuracy by integrating ten existing local or global maps. The error removal and simple averaging algorithm, which is efficient and makes no assumption about the data and associated errors, was proposed to integrate these ten forest AGB maps. We first compiled the global reference AGB from in situ measurements and high-resolution AGB data that were originally derived from field data and airborne lidar data and determined the errors of each forest AGB map at the pixels with corresponding reference AGB values. Based on the errors determined from reference AGB data, the pixel-by-pixel errors associated with each of the ten AGB datasets were estimated from multiple predictors (e.g., leaf area index, forest canopy height, forest cover, land surface elevation, slope, temperature, and precipitation) using the random forest algorithm. The estimated pixel-by-pixel errors were then removed from the corresponding forest AGB datasets, and finally, global forest AGB maps were generated by combining the calibrated existing forest AGB datasets using the simple averaging algorithm. Cross-validation using reference AGB data showed that the accuracy of the fused global forest AGB map had an R-squared of 0.61 and a root mean square error (RMSE) of 53.68 Mg/ha, which is better than the reported accuracies (R-squared of 0.56 and RMSE larger than 80 Mg/ha) in the literature. Intercomparison with previous studies also suggested that the fused AGB estimates were much closer to the reference AGB values. This study attempted to integrate existing forest AGB datasets for generating a global forest AGB map with better accuracy and moved one step forward for our understanding of the global terrestrial carbon cycle by providing improved benchmarks of global forest carbon stocks.

Regional Mapping and Spatial Distribution Analysis of Canopy Palms in an Amazon Forest Using Deep Learning and VHR Images

Mapping plant species at the regional scale to provide information for ecologists and forest managers is a challenge for the remote sensing community. Here, we use a deep learning algorithm called U-net and very high-resolution multispectral images (0.5 m) from GeoEye satellite to identify, segment and map canopy palms over ∼3000 km 2 of Amazonian forest. The map was used to analyse the spatial distribution of canopy palm trees and its relation to human disturbance and edaphic conditions. The overall accuracy of the map was 95.5% and the F1-score was 0.7. Canopy palm trees covered 6.4% of the forest canopy and were distributed in more than two million patches that can represent one or more individuals. The density of canopy palms is affected by human disturbance. The post-disturbance density in secondary forests seems to be related to the type of disturbance, being higher in abandoned pasture areas and lower in forests that have been cut once and abandoned. Additionally, analysis of palm trees’ distribution shows that their abundance is controlled naturally by local soil water content, avoiding both flooded and waterlogged areas near rivers and dry areas on the top of the hills. They show two preferential habitats, in the low elevation above the large rivers, and in the slope directly below the hill tops. Overall, their distribution over the region indicates a relatively pristine landscape, albeit within a forest that is critically endangered because of its location between two deforestation fronts and because of illegal cutting. New tree species distribution data, such as the map of all adult canopy palms produced in this work, are urgently needed to support Amazon species inventory and to understand their distribution and diversity.

Variability and uncertainty in forest biomass estimates from the tree to landscape scale: the role of allometric equations

BACKGROUND

Biomass maps are valuable tools for estimating forest carbon and forest planning. Individual-tree biomass estimates made using allometric equations are the foundation for these maps, yet the potentially-high uncertainty and bias associated with individual-tree estimates is commonly ignored in biomass map error. We developed allometric equations for lodgepole pine (Pinus contorta), ponderosa pine (P. ponderosa), and Douglas-fir (Pseudotsuga menziesii) in northern Colorado. Plot-level biomass estimates were combined with Landsat imagery and geomorphometric and climate layers to map aboveground tree biomass. We compared biomass estimates for individual trees, plots, and at the landscape-scale using our locally-developed allometric equations, nationwide equations applied across the U.S., and the Forest Inventory and Analysis Component Ratio Method (FIA-CRM). Total biomass map uncertainty was calculated by propagating errors from allometric equations and remote sensing model predictions. Two evaluation methods for the allometric equations were compared in the error propagation-errors calculated from the equation fit (equation-derived) and errors from an independent dataset of destructively-sampled trees (n = 285).

RESULTS

Tree-scale error and bias of allometric equations varied dramatically between species, but local equations were generally most accurate. Depending on allometric equation and evaluation method, allometric uncertainty contributed 30-75% of total uncertainty, while remote sensing model prediction uncertainty contributed 25-70%. When using equation-derived allometric error, local equations had the lowest total uncertainty (root mean square error percent of the mean [% RMSE] = 50%). This is likely due to low-sample size (10-20 trees sampled per species) allometric equations and evaluation not representing true variability in tree growth forms. When independently evaluated, allometric uncertainty outsized remote sensing model prediction uncertainty. Biomass across the 1.56 million ha study area and uncertainties were similar for local (2.1 billion Mg; % RMSE = 97%) and nationwide (2.2 billion Mg;  % RMSE = 94%) equations, while FIA-CRM estimates were lower and more uncertain (1.5 billion Mg;  % RMSE = 165%).

CONCLUSIONS

Allometric equations should be selected carefully since they drive substantial differences in bias and uncertainty. Biomass quantification efforts should consider contributions of allometric uncertainty to total uncertainty, at a minimum, and independently evaluate allometric equations when suitable data are available.

Tree Crown Delineation Algorithm Based on a Convolutional Neural Network

Tropical forests concentrate the largest diversity of species on the planet and play a key role in maintaining environmental processes. Due to the importance of those forests, there is growing interest in mapping their components and getting information at an individual tree level to conduct reliable satellite-based forest inventory for biomass and species distribution qualification. Individual tree crown information could be manually gathered from high resolution satellite images; however, to achieve this task at large-scale, an algorithm to identify and delineate each tree crown individually, with high accuracy, is a prerequisite. In this study, we propose the application of a convolutional neural network—Mask R-CNN algorithm—to perform the tree crown detection and delineation. The algorithm uses very high-resolution satellite images from tropical forests. The results obtained are promising—the R e c a l l , P r e c i s i o n , and F 1 score values obtained were were 0.81 , 0.91 , and 0.86 , respectively. In the study site, the total of tree crowns delineated was 59,062 . These results suggest that this algorithm can be used to assist the planning and conduction of forest inventories. As the algorithm is based on a Deep Learning approach, it can be systematically trained and used for other regions.

Estimating intertidal seaweed biomass at larger scales from quadrat surveys

The amount of macroalgal biomass is an important ecosystem variable. Estimates can be made for a sampled area or values can be extrapolated to represent biomass over a larger region. Typically biomass is scaled-up using the area multiplied by the mean: a non-spatial method. Where algal biomass is patchy or shows gradients, non-spatial estimates for an area may be improved by spatial interpolation. A separate issue with scaling-up biomass estimates is that conventional confidence intervals based on the standard error (SE) of the sample may not be appropriate. The issues around interpolation and confidence intervals were examined for three fucoid species using data from 40 × 0.25 m-2 quadrats thrown in a 0.717 ha sampling plot on the shore of Galway Bay. Despite evidence of spatial autocorrelation, interpolation did not appear to improve estimates of the total plot biomass of Fucus serratus and F. vesiculosus. In contrast, interpolated estimates for Ascophyllum nodosum had less error than those based on the non-spatial method. Bootstrapped confidence intervals had several benefits over those based on the SE. These benefits include the avoidance of negative confidence limits at low sample sizes and no assumptions of normality in the data. If there is reason to expect strong patchiness or a gradient of biomass in the area of interest, interpolation is likely to produce more accurate estimates of biomass than non-spatial methods. Development of methodologies for biomass would benefit from more definition of local and regional gradients in biomass and their associated covariates.

Carbon and Nitrogen Stocks in Three Types of Larix gmelinii Forests in Daxing’an Mountains, Northeast China

Studying carbon and nitrogen stocks in different types of larch forest ecosystems is of great significance for assessing the carbon sink capacity and nitrogen level in larch forests. To evaluate the effects of the differences of forest type on the carbon and nitrogen stock capacity of the larch forest ecosystem, we selected three typical types of larch forest ecosystems in the northern part of Daxing’an Mountains, which were the Rhododendron simsii-Larix gmelinii forest (RL), Ledum palustre-Larix gmelinii forest (LL) and Sphagnum-Bryum-Ledum palustre-Larix gmelinii forest (SLL), to determine the carbon and nitrogen stocks in the vegetation (trees and understories), litter and soil. Results showed that there were significant differences in carbon and nitrogen stocks among the three types of larch forest ecosystems, showing a sequence of SLL (288.01 Mg·ha−1 and 25.19 Mg·ha−1) > LL (176.52 Mg·ha−1 and 14.85 Mg·ha−1) > RL (153.93 Mg·ha−1 and 10.00 Mg·ha−1) (P < 0.05). The largest proportions of carbon and nitrogen stocks were found in soils, accounting for 83.20%, 72.89% and 64.61% of carbon stocks and 98.61%, 97.58% and 96.00% of nitrogen stocks in the SLL, LL and RL, respectively. Also, it was found that significant differences among the three types of larch forest ecosystems in terms of soil carbon and nitrogen stocks (SLL > LL > RL) (P < 0.05) were the primary reasons for the differences in the ecosystem carbon and nitrogen stocks. More than 79% of soil carbon and 51% of soil nitrogen at a depth of 0–100 cm were stored in the upper 50 cm of the soil pool. In the vegetation layer, due to the similar tree biomass carbon and nitrogen stocks, there were no significant differences in carbon and nitrogen stocks among the three types of larch forest ecosystems. The litter carbon stock in the SLL was significantly higher than that in the LL and RL (P < 0.05), but no significant differences in nitrogen stock were found among them (P > 0.05). These findings suggest that different forest types with the same tree layer and different understory vegetation can greatly affect the carbon and nitrogen stock capacity of the forest ecosystem. This indicates that understory vegetation may have significant effects on the carbon and nitrogen stocks in soil and litter, which highlights the need to consider the effects of understory in future research into the carbon and nitrogen stock capacity of forest ecosystems.

Landscape-level validation of allometric relationships for carbon stock estimation reveals bias driven by soil type

Mitigation of climate change depends on accurate estimation and mapping of terrestrial carbon stocks, particularly in carbon dense tropical forests. Allometric equations can be used to robustly estimate biomass of tropical trees, but often require tree height, which is frequently unknown. Researchers and practitioners must, therefore, decide whether to directly measure a subset of tree heights to develop diameter : height (D:H) equations or rely on previously published generic equations. To date, studies comparing the two approaches have been spatially restricted and/or not randomly allocated across the landscape of interest, making the implications of deciding whether or not to measure tree heights difficult to determine. To address this issue, we use inventory data from a systematic-random forest inventory across Gabon (102 forest sites; 42,627 trees, including 7,036 height-measured trees). Using plot-specific models of D:H as a benchmark, we compare the performance of a suite of locally fitted and commonly used generic models (parameterized national, georegional, and pantropical equations) across a variety of scales, and assess which abiotic, anthropogenic, and topographical covariates contribute the most to bias in height estimation. We reveal marked spatial structure in the magnitude and direction of bias in tree height estimation using all generic models, due at least in part to soil type, which compounded to substantial error in site-level AGB estimates (of up to 38% or 150 Mg/ha). However, two generic pantropical models (Chave 2014; Feldpausch 2012) converged to within 2.5% of mean AGB at the landscape scale. Our results suggest that some (not all) pantropical equations can extrapolate AGB across large spatial scales with minimal bias in estimated mean AGB. However, extreme caution must be taken when interpreting the AGB estimates from generic models at the site-level as they fail to capture substantial spatial variation in D:H relationships, which could lead to dramatic under- or over-estimation of site-level carbon stocks. Validated allometric models derived at site- or soil-type-levels may be the best way to reduce such biases arising from landscape-level heterogeneity in D:H model fit in the Afrotropics.

From small-scale forest structure to Amazon-wide carbon estimates

Tropical forests play an important role in the global carbon cycle. High-resolution remote sensing techniques, e.g., spaceborne lidar, can measure complex tropical forest structures, but it remains a challenge how to interpret such information for the assessment of forest biomass and productivity. Here, we develop an approach to estimate basal area, aboveground biomass and productivity within Amazonia by matching 770,000 GLAS lidar (ICESat) profiles with forest simulations considering spatial heterogeneous environmental and ecological conditions. This allows for deriving frequency distributions of key forest attributes for the entire Amazon. This detailed interpretation of remote sensing data improves estimates of forest attributes by 20–43% as compared to (conventional) estimates using mean canopy height. The inclusion of forest modeling has a high potential to close a missing link between remote sensing measurements and the 3D structure of forests, and may thereby improve continent-wide estimates of biomass and productivity.

Improving estimates of forest biomass based on remote sensing data is important to assess global carbon cycling. Here the authors develop an approach to use forest gap models to simulate lidar waveforms and compare the outputs with ICESAT-1 GLAS profiles, showing improved estimates across the Amazon basin.

The Forest Observation System, building a global reference dataset for remote sensing of forest biomass

Forest biomass is an essential indicator for monitoring the Earth's ecosystems and climate. It is a critical input to greenhouse gas accounting, estimation of carbon losses and forest degradation, assessment of renewable energy potential, and for developing climate change mitigation policies such as REDD+, among others. Wall-to-wall mapping of aboveground biomass (AGB) is now possible with satellite remote sensing (RS). However, RS methods require extant, up-to-date, reliable, representative and comparable in situ data for calibration and validation. Here, we present the Forest Observation System (FOS) initiative, an international cooperation to establish and maintain a global in situ forest biomass database. AGB and canopy height estimates with their associated uncertainties are derived at a 0.25 ha scale from field measurements made in permanent research plots across the world's forests. All plot estimates are geolocated and have a size that allows for direct comparison with many RS measurements. The FOS offers the potential to improve the accuracy of RS-based biomass products while developing new synergies between the RS and ground-based ecosystem research communities.

Evaluating spatial coverage of data on the aboveground biomass in undisturbed forests in the Brazilian Amazon

BACKGROUND

Brazilian Amazon forests contain a large stock of carbon that could be released into the atmosphere as a result of land use and cover change. To quantify the carbon stocks, Brazil has forest inventory plots from different sources, but they are unstandardized and not always available to the scientific community. Considering the Brazilian Amazon extension, the use of remote sensing, combined with forest inventory plots, is one of the best options to estimate forest aboveground biomass (AGB). Nevertheless, the combination of limited forest inventory data and different remote sensing products has resulted in significant differences in the spatial distribution of AGB estimates. This study evaluates the spatial coverage of AGB data (forest inventory plots, AGB maps and remote sensing products) in undisturbed forests in the Brazilian Amazon. Additionally, we analyze the interconnection between these data and AGB stakeholders producing the information. Specifically, we provide the first benchmark of the existing field plots in terms of their size, frequency, and spatial distribution.

RESULTS

We synthesized the coverage of forest inventory plots, AGB maps and airborne light detection and ranging (LiDAR) transects of the Brazilian Amazon. Although several extensive forest inventories have been implemented, these AGB data cover a small fraction of this region (e.g., central Amazon remains largely uncovered). Although the use of new technology such as airborne LiDAR cover a significant extension of AGB surveys, these data and forest plots represent only 1% of the entire forest area of the Brazilian Amazon.

CONCLUSIONS

Considering that several institutions involved in forest inventories of the Brazilian Amazon have different goals, protocols, and time frames for forest surveys, forest inventory data of the Brazilian Amazon remain unstandardized. Research funding agencies have a very important role in establishing a clear sharing policy to make data free and open as well as in harmonizing the collection procedure. Nevertheless, the use of old and new forest inventory plots combined with airborne LiDAR data and satellite images will likely reduce the uncertainty of the AGB distribution of the Brazilian Amazon.

Species Matter: Wood Density Influences Tropical Forest Biomass at Multiple Scales

The mass of carbon contained in trees is governed by the volume and density of their wood. This represents a challenge to most remote sensing technologies, which typically detect surface structure and parameters related to wood volume but not to its density. Since wood density is largely determined by taxonomic identity this challenge is greatest in tropical forests where there are tens of thousands of tree species. Here, using pan-tropical literature and new analyses in Amazonia with plots with reliable identifications we assess the impact that species-related variation in wood density has on biomass estimates of mature tropical forests. We find impacts of species on forest biomass due to wood density at all scales from the individual tree up to the whole biome: variation in tree species composition regulates how much carbon forests can store. Even local differences in composition can cause variation in forest biomass and carbon density of 20% between subtly different local forest types, while additional large-scale floristic variation leads to variation in mean wood density of 10-30% across Amazonia and the tropics. Further, because species composition varies at all scales and even vertically within a stand, our analysis shows that bias and uncertainty always result if individual identity is ignored. Since sufficient inventory-based evidence based on botanical identification now exists to show that species composition matters biome-wide for biomass, we here assemble and provide mean basal-area-weighted wood density values for different forests across the lowand tropical biome. These range widely, from 0.467 to 0.728 g cm-3 with a pan-tropical mean of 0.619 g cm-3. Our analysis shows that mapping tropical ecosystem carbon always benefits from locally validated measurement of tree-by-tree botanical identity combined with tree-by-tree measurement of dimensions. Therefore whenever possible, efforts to map and monitor tropical forest carbon using remote sensing techniques should be combined with tree-level measurement of species identity by botanists working in inventory plots.

Terrestrial LiDAR: a three-dimensional revolution in how we look at trees

Contents Summary I. Introduction II. Terrestrial laser scanning III. Turning points into trees IV. Current and future applications of TLS V. Conclusions Acknowledgements References SUMMARY: Terrestrial laser scanning (TLS) is providing new, very detailed three-dimensional (3D) measurements of forest canopy structure. The information that TLS measurements can provide in describing detailed, accurate 3D canopy architecture offers fascinating new insights into the variety of tree form, environmental drivers and constraints, and the relationship between form and function, particularly for tall, hard-to-measure trees. TLS measurements are helping to test fundamental ecological theories and enabling new and better exploitation of other measurements and models that depend on 3D structural information. This Tansley insight introduces the background and capabilities of TLS in forest ecology, discusses some of the barriers to progress, and identifies some of the directions for new work.

A critique of general allometry-inspired models for estimating forest carbon density from airborne LiDAR

There is currently much interest in developing general approaches for mapping forest aboveground carbon density using structural information contained in airborne LiDAR data. The most widely utilized model in tropical forests assumes that aboveground carbon density is a compound power function of top of canopy height (a metric easily derived from LiDAR), basal area and wood density. Here we derive the model in terms of the geometry of individual tree crowns within forest stands, showing how scaling exponents in the aboveground carbon density model arise from the height−diameter (H−D) and projected crown area−diameter (C−D) allometries of individual trees. We show that a power function relationship emerges when the C−D scaling exponent is close to 2, or when tree diameters follow a Weibull distribution (or other specific distributions) and are invariant across the landscape. In addition, basal area must be closely correlated with canopy height for the approach to work. The efficacy of the model was explored for a managed uneven−aged temperate forest in Ontario, Canada within which stands dominated by sugar maple (Acer saccharum Marsh.) and mixed stands were identified. A much poorer goodness−of−fit was obtained than previously reported for tropical forests (R² = 0.29 vs. about 0.83). Explanations for the poor predictive power on the model include: (1) basal area was only weakly correlated with top canopy height; (2) tree size distributions varied considerably across the landscape; (3) the allometry exponents are affected by variation in species composition arising from timber management and soil conditions; and (4) the C-D allometric power function was far from 2 (1.28). We conclude that landscape heterogeneity in forest structure and tree allometry reduces the accuracy of general power-function models for predicting aboveground carbon density in managed forests. More studies in different forest types are needed to understand the situations in which power functions of LiDAR height are appropriate for modelling forest carbon stocks.

Across Date Species Detection Using Airborne Imaging Spectroscopy

Imaging spectroscopy is a promising tool for airborne tree species recognition in hyper-diverse tropical canopies. However, its widespread application is limited by the signal sensitivity to acquisition parameters, which may require new training data in every new area of application. This study explores how various pre-processing steps may improve species discrimination and species recognition under different operational settings. In the first experiment, a classifier was trained and applied on imaging spectroscopy data acquired on a single date, while in a second experiment, the classifier was trained on data from one date and applied to species identification on data from a different date. A radiative transfer model based on atmospheric compensation was applied with special focus on the automatic retrieval of aerosol amounts. The impact of spatial or spectral filtering and normalisation was explored as an alternative to atmospheric correction. A pixel-wise classification was performed with a linear discriminant analysis trained on individual tree crowns identified at the species level. Tree species were then identified at the crown scale based on a majority vote rule. Atmospheric corrections did not outperform simple statistical processing (i.e., filtering and normalisation) when training and testing sets were taken from the same flight date. However, atmospheric corrections became necessary for reliable species recognition when different dates were considered. Shadow masking improved species classification results in all cases. Single date classification rate was 83.9% for 1297 crowns of 20 tropical species. The loss of mean accuracy observed when using training data from one date to identify species at another date in the same area was limited to 10% when atmospheric correction was applied.

Performance of Laser-Based Electronic Devices for Structural Analysis of Amazonian Terra-Firme Forests

Tropical vegetation biomass represents a key component of the carbon stored in global forest ecosystems. Estimates of aboveground biomass commonly rely on measurements of tree size (diameter and height) and then indirectly relate, via allometric relationships and wood density, to biomass sampled from a relatively small number of harvested and weighed trees. Recently, however, novel in situ remote sensing techniques have been proposed, which may provide nondestructive alternative approaches to derive biomass estimates. Nonetheless, we still lack knowledge of the measurement uncertainties, as both the calibration and validation of estimates using different techniques and instruments requires consistent assessment of the underlying errors. To that end, we investigate different approaches estimating the tropical aboveground biomass in situ. We quantify the total and systematic errors among measurements obtained from terrestrial light detection and ranging (LiDAR), hypsometer-based trigonometry, and traditional forest inventory. We show that laser-based estimates of aboveground biomass are in good agreement (<10% measurement uncertainty) with traditional measurements. However, relative uncertainties vary among the allometric equations based on the vegetation parameters used for parameterization. We report the error metrics for measurements of tree diameter and tree height and discuss the consequences for estimated biomass. Despite methodological differences detected in this study, we conclude that laser-based electronic devices could complement conventional measurement techniques, thereby potentially improving estimates of tropical vegetation biomass.

Reforestation of agricultural land in the tropics: The relative contribution of soil, living biomass and debris pools to carbon sequestration

Tropical regions of the world experience high rates of land-use change and this has a major influence on terrestrial carbon (C) pools and the global C cycle. We assessed land-use change from agriculture to reforested plantings (with endemic species), up to 33 years of age, using 10 paired sites in the wet tropics, Australia. We determined the impacts on 0-50 cm below-ground C (soil organic C (SOC), charcoal C, humic organic C, particulate organic C, resistant organic C), C stored in roots (fine and coarse), C stored in living above-ground biomass and debris C pools. Reforested areas accumulated ecosystem C at a rate of 7.4 Mg ha^-1 yr^-1. Reforestation plantings contained, on average, 2.3 times more ecosystem C than agricultural areas (102 Mg ha^-1 and 233 Mg ha^-1, respectively). Most of the C accumulation was in living above-ground and below-ground biomass (60 and 30%, respectively) with a smaller amount in debris pools (16%). Apart from C in roots, soil C accumulation was not obvious across sites ranging from 8 to 33 years since reforestation, relative to the agricultural baseline. Differences in SOC (and associated SOC pools) to a depth of 50 cm, did exist between reforested areas and adjacent agriculture at some sites, however there was not a consistent trend in SOC associated with reforestation. Local site-based factors (e.g. soil texture and mineralogy, land-use history and microbial activity) appear to have a strong influence on the direction of the change in SOC. While reforestation in the tropics has great potential to accumulate C in biomass in living vegetation, and debris pools, it is likely to take approximately 50 years before C stocks of reforested areas resemble natural ecosystems. Accumulation of SOC through reforestation is difficult to achieve, highlighting the need to conserve carbon pools in remnant forests in the tropics.

The Importance of Consistent Global Forest Aboveground Biomass Product Validation

Several upcoming satellite missions have core science requirements to produce data for accurate forest aboveground biomass mapping. Largely because of these mission datasets, the number of available biomass products is expected to greatly increase over the coming decade. Despite the recognized importance of biomass mapping for a wide range of science, policy and management applications, there remains no community accepted standard for satellite-based biomass map validation. The Committee on Earth Observing Satellites (CEOS) is developing a protocol to fill this need in advance of the next generation of biomass-relevant satellites, and this paper presents a review of biomass validation practices from a CEOS perspective. We outline the wide range of anticipated user requirements for product accuracy assessment and provide recommendations for the validation of biomass products. These recommendations include the collection of new, high-quality in situ data and the use of airborne lidar biomass maps as tools toward transparent multi-resolution validation. Adoption of community-vetted validation standards and practices will facilitate the uptake of the next generation of biomass products.

Illegal logging as a disincentive to the establishment of a sustainable forest sector in the Amazon

Brazil recently began granting timber concessions in public forests to promote sustainable forest use. The effectiveness of this strategy hinges on the design and implementation of the concessions themselves as well as their competitive position within the logging sector as a whole. There is, however, a lack of information on the competitive interaction between legal and illegal logging and its effects on concessions profits. We address this knowledge gap by using a spatially explicit simulation model of the Amazon timber industry to examine the potential impact of illegal logging on timber concessions allocation and profits in a 30-year harvest cycle. In a scenario in which illegal logging takes place outside concessions, including private and public “undesignated” lands, concession harvested area would decrease by 59% due to competition with illegal logging. Moreover, 29 out of 39 National Forests (≈74%) would experience a decrease in harvested area. This “leakage” effect could reduce concession net rents by up to USD 1.3 Billion after 30 years. Federal and State “undesignated” lands, if not adequately protected, could have 40% of their total volume illegally harvested in 30 years. Our results reinforce the need to invest in tackling illegal logging, if the government wants the forest concessions program to be successful.

Multifrequency and Full-Polarimetric SAR Assessment for Estimating Above Ground Biomass and Leaf Area Index in the Amazon Várzea Wetlands

The aim of this study is to evaluate the potential of multifrequency and Full-polarimetric Synthetic Aperture Radar (SAR) data for retrieving both Above Ground Biomass (AGB) and Leaf Area Index (LAI) in the Amazon floodplain forest environment. Two specific questions were proposed: (a) Does multifrequency SAR data perform more efficiently than single-frequency data in estimating LAI and AGB of várzea forests?; and (b) Are quad-pol SAR data more efficient than single- and dual-pol SAR data in estimating LAI and AGB of várzea forest? To answer these questions, data from different sources (TerraSAR-X Multi Look Ground Range Detected (MGD), Radarsat-2 Standard Qual-Pol, advanced land observing satellite (ALOS)/ phased-arrayed L-band SAR (PALSAR-1). Fine-beam dual (FDB) and quad Polarimetric mode) were combined in 10 different scenarios to model both LAI and AGB. A R-platform routine was implemented to automatize the selection of the best regression models. Results indicated that ALOS/PALSAR variables provided the best estimates for both LAI and AGB. Single-frequency L-band data was more efficient than multifrequency SAR. PALSAR-FDB HV-dB provided the best LAI estimates during low-water season. The best AGB estimates at high-water season were obtained by PALSAR-1 quad-polarimetric data. The top three features for estimating AGB were proportion of volumetric scattering and both the first and second dominant phase difference between trihedral and dihedral scattering, extracted from Van Zyl and Touzi decomposition, respectively. The models selected for both AGB and LAI were parsimonious. The Root Mean Squared Error (RMSEcv), relative overall RMSEcv (%) and R2 value for LAI were 0.61%, 0.55% and 13%, respectively, and for AGB, they were 74.6 t·ha−1, 0.88% and 46%, respectively. These results indicate that L-band (ALOS/PALSAR-1) has a high potential to provide quantitative and spatial information about structural forest attributes in floodplain forest environments. This potential may be extended not only with PALSAR-2 data but also to forthcoming missions (e.g., NISAR, Global Ecosystems Dynamics Investigation Lidar (GEDI), BIOMASS, Tandem-L) for promoting wall-to-wall AGB mapping with a high level of accuracy in dense tropical forest regions worldwide.

Fine-scale processes shape ecosystem service provision by an Amazonian hyperdominant tree species

Conspecific distance and density-dependence is a key driver of tree diversity in natural forests, but the extent to which this process may influence ecosystem service provision is largely unknown. Drawing on a dataset of >135,000 trees from the Peruvian Amazon, we assessed its manifestation in biomass accumulation and seed production of Brazil nut (Bertholletia excelsa) which plays a keystone role in carbon sequestration and NTFP harvesting in Amazonia. For the first time, we find both negative and positive effects of conspecific proximity on seed production and above ground biomass at small and large nearest neighbour distances, respectively. Plausible explanations for negative effects at small distances are fine-scale genetic structuring and competition for shared resources, whereas positive effects at large distances are likely due to increasing pollen limitation and suboptimal growth conditions. Finally, findings suggest that most field plots in Amazonia used for estimating carbon storage are too small to account for distance and density-dependent effects and hence may be inadequate for measuring species-centric ecosystem services.

Optimizing biomass estimates of savanna woodland at different spatial scales in the Brazilian Cerrado: Re-evaluating allometric equations and environmental influences

Cerrado is the second largest biome in South America and accounted for the second largest contribution to carbon emissions in Brazil for the last 10 years, mainly due to land-use changes. It comprises approximately 2 million km² and is divided into 22 ecoregions, based on environmental conditions and vegetation. The most dominant vegetation type is cerrado sensu stricto (cerrado ss), a savanna woodland. Quantifying variation of biomass density of this vegetation is crucial for climate change mitigation policies. Integrating remote sensing data with adequate allometric equations and field-based data sets can provide large-scale estimates of biomass. We developed individual-tree aboveground biomass (AGB) allometric models to compare different regression techniques and explanatory variables. We applied the model with the strongest fit to a comprehensive ground-based data set (77 sites, 893 plots, and 95,484 trees) to describe AGB density variation of cerrado ss. We also investigated the influence of physiographic and climatological variables on AGB density; this analysis was restricted to 68 sites because eight sites could not be classified into a specific ecoregion, and one site had no soil texture data. In addition, we developed two models to estimate plot AGB density based on plot basal area. Our data show that for individual-tree AGB models a) log-log linear models provided better estimates than nonlinear power models; b) including species as a random effect improved model fit; c) diameter at 30 cm above ground was a reliable predictor for individual-tree AGB, and although height significantly improved model fit, species wood density did not. Mean tree AGB density in cerrado ss was 22.9 tons ha^-1 (95% confidence interval = ± 2.2) and varied widely between ecoregions (8.8 to 42.2 tons ha^-1), within ecoregions (e.g. 4.8 to 39.5 tons ha^-1), and even within sites (24.3 to 69.9 tons ha^-1). Biomass density tended to be higher in sites close to the Amazon. Ecoregion explained 42% of biomass variation between the 68 sites (P < 0.01) and shows strong potential as a parameter for classifying regional biomass variation in the Cerrado.