An unexpectedly large count of trees in the West African Sahara and Sahel

Mapping Temperate Savanna in Northeastern China Through Integrating UAV and Satellite Imagery

Temperate savannas are globally distributed ecosystems that play a crucial role in regulating the global carbon cycle and significantly contribute to human livelihoods. This study aims to develop a novel method for identifying temperate savannas and to map their distribution in Northeastern China. To achieve this objective, Unmanned Aerial Vehicle (UAV) imagery was integrated with Sentinel-2 and Sentinel-1 satellite imagery using Random Forest  (RF) regression and Classification and Regression Tree (CART) algorithms. The training and validation datasets were derived from UAV imagery covering a ground area of 5 × 107m2. The proposed method achieved an overall accuracy of 0.82 in identifying temperate savanna in Northeastern China, covering a total area of 1.7 × 1011 m2. The resulting map significantly improves understanding of the spatial distribution and extent of temperate savannas. The developed methodology establishes a framework for assessing regional and global savanna distributions in future studies.

Multipurpose trees on farms can improve nutrition in Malawi

In low- and middle-income countries, there is growing evidence that trees in landscapes can support healthy diets. Yet, the bulk of this evidence is based on broad-scale associations and thus fails to tease apart the contributions of different types of trees. Here, we examine how the use of on-farm trees for food, income, and fuel relates to micronutrient adequacy (vitamin A, zinc, iron, and folate) and food sourcing patterns in rural Malawi. We used data from socioeconomic, land use, and dietary surveys conducted with 460 women in both the dry and wet seasons. Our results illustrate that, compared to other uses, the use of on-farm trees for food is the most significant determinant of women's micronutrient adequacy across seasons. While this study does not find consistent dietary benefits from using on-farm trees for only fuel and income, our results suggest that multipurpose on-farm trees can support adequate intake of all measured micronutrients.

Fragmentation in patchy ecosystems: a call for a functional approach

Habitat fragmentation is a major threat to biodiversity, but existing literature largely ignores naturally patchy ecosystems in favor of forests, where deforestation creates spatially distinct fragments. Here, we use savannas to highlight the problems with applying forest fragmentation principles to spatially patchy ecosystems. Identifying fragmentation using landscape functionality, specifically connectivity, enables better understanding of ecosystem dynamics. Tools and concepts from connectivity research are well suited to identifying barriers other than vegetation structure contributing to fragmentation. Opportunities exist to improve fragmentation mapping by combining remote-sensing data with field measurements related to connectivity to empirically test whether landscapes are functionally fragmented. Advancements in deep learning and increasingly accessible data open many possibilities for comprehensive maps of fragmentation.

Tracking tree demography and forest dynamics at scale using remote sensing

Capturing how tree growth and survival vary through space and time is critical to understanding the structure and dynamics of tree-dominated ecosystems. However, characterising demographic processes at scale is inherently challenging, as trees are slow-growing, long-lived and cover vast expanses of land. We used repeat airborne laser scanning data acquired across 25 km2 of semi-arid, old-growth temperate woodland in Western Australia to track the height growth, crown expansion and mortality of 42 213 individual trees over 9 yr. We found that demographic rates are constrained by a combination of tree size, competition and topography. After initially investing in height growth, trees progressively shifted to crown expansion as they grew larger, while mortality risk decreased considerably with size. Across the landscape, both tree growth and survival increased with topographic wetness, resulting in vegetation patterns that are strongly spatially structured. Moreover, biomass gains from woody growth generally outpaced losses from mortality, suggesting these old-growth woodlands remain a net carbon sink in the absence of wildfires. Our study sheds new light on the processes that shape the dynamics and spatial structure of semi-arid woody ecosystems and provides a roadmap for using emerging remote sensing technologies to track tree demography at scale.

Carbon sequestration potential of tree planting in China

China's large-scale tree planting programs are critical for achieving its carbon neutrality by 2060, but determining where and how to plant trees for maximum carbon sequestration has not been rigorously assessed. Here, we developed a comprehensive machine learning framework that integrates diverse environmental variables to quantify tree growth suitability and its relationship with tree numbers. Then, their correlations with biomass carbon stocks were robustly established. Carbon sink potentials were mapped in distinct tree-planting scenarios. Under one of them aligned with China's ecosystem management policy, 44.7 billion trees could be planted, increasing forest stock by 9.6 ± 0.8 billion m³ and sequestering 5.9 ± 0.5 PgC equivalent to double China's 2020 industrial CO2 emissions. We found that tree densification within existing forests is an economically viable and effective strategy and so it should be a priority in future large-scale planting programs.

Facing the challenge of NDVI dataset consistency for improved characterization of vegetation response to climate variability

Non-linear trend detection in Earth observation time series has become a standard method to characterize changes in terrestrial ecosystems. However, results are largely dependent on the quality and consistency of the input data, and only few studies have addressed the impact of data artifacts on the interpretation of detected abrupt changes. Here we study non-linear dynamics and turning points (TPs) of temperate grasslands in East Eurasia using two independent state-of-the-art satellite NDVI datasets (CGLS v3 and MODIS C6) and explore the impact of water availability on observed vegetation changes during 2001-2019. By applying the Break For Additive Season and Trend (BFAST01) method, we conducted a classification typology based on vegetation dynamics which was spatially consistent between the datasets for 40.86 % (459,669 km2) of the study area. When considering also the timing of the TPs, 27.09 % of the pixels showed consistent results between datasets, suggesting that careful interpretation was needed for most of the areas of detected vegetation dynamics when applying BFAST to a single dataset. Notably, for these areas showing identical typology we found that interrupted decreases in vegetation productivity were dominant in the transition zone between desert and steppes. Here, a strong link with changes in water availability was found for >80 % of the area, indicating that increasing drought stress had regulated vegetation productivity in recent years. This study shows the necessity of a cautious interpretation of the results when conducting advanced characterization of vegetation response to climate variability, but at the same time also the opportunities of going beyond the use of single dataset in advanced time-series approaches to better understanding dryland vegetation dynamics for improved anthropogenic interventions to combat vegetation productivity decrease.

The importance of different forest management systems for people's dietary quality in Tanzania

Context

A large body of literature has shown that forests provide nutritious foods in many low- and middle-income countries. Yet, there is limited evidence on the contributions from different types of forest and tree systems.

Objectives

Here, we focus on individual trees and smaller forest patches outside established forest reserves as well as different forest management systems.

Methods

We do so by combining novel high-resolution data on tree cover with 24-h dietary recall surveys from 465 women in Tanzania.

Results

We show that people with more unclassified tree cover (i.e., individual trees and small forest patches) in their nearby surroundings have more adequate protein, iron, zinc, and vitamin A intakes. We also find that having a nearby forest under Participatory Forest Management (PFM) system is associated with higher adequacy levels of energy, iron, zinc and vitamin A. By contrast, tree cover within other types of forest (e.g., Government Forest Reserves and Government Forest Plantations) is not positively associated with people's dietary quality.

Conclusions

Our key finding is that having individual trees, smaller forest patches and/or forest under PFM in close proximity is more beneficial for people's diets than other types of established forests. Our results highlight the nutritional importance of trees outside established forests and question the often-assumed benefits of forests if these are made inaccessible by social barriers (e.g., legislation). Finally, our results emphasize the need to distinguish between different forest management systems when studying forest-diet linkages.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10980-024-01961-6.

Artificial intelligence for geoscience: Progress, challenges, and perspectives

This paper explores the evolution of geoscientific inquiry, tracing the progression from traditional physics-based models to modern data-driven approaches facilitated by significant advancements in artificial intelligence (AI) and data collection techniques. Traditional models, which are grounded in physical and numerical frameworks, provide robust explanations by explicitly reconstructing underlying physical processes. However, their limitations in comprehensively capturing Earth's complexities and uncertainties pose challenges in optimization and real-world applicability. In contrast, contemporary data-driven models, particularly those utilizing machine learning (ML) and deep learning (DL), leverage extensive geoscience data to glean insights without requiring exhaustive theoretical knowledge. ML techniques have shown promise in addressing Earth science-related questions. Nevertheless, challenges such as data scarcity, computational demands, data privacy concerns, and the "black-box" nature of AI models hinder their seamless integration into geoscience. The integration of physics-based and data-driven methodologies into hybrid models presents an alternative paradigm. These models, which incorporate domain knowledge to guide AI methodologies, demonstrate enhanced efficiency and performance with reduced training data requirements. This review provides a comprehensive overview of geoscientific research paradigms, emphasizing untapped opportunities at the intersection of advanced AI techniques and geoscience. It examines major methodologies, showcases advances in large-scale models, and discusses the challenges and prospects that will shape the future landscape of AI in geoscience. The paper outlines a dynamic field ripe with possibilities, poised to unlock new understandings of Earth's complexities and further advance geoscience exploration.

Multi-model assessment of potential natural vegetation to support ecological restoration

Ecological restoration is imperative for controlling desertification. Potential natural vegetation (PNV), the theoretical vegetation succession state, can guides near-natural restoration. Although a rising transition from traditional statistical methods to advanced machine learning and deep learning is observed in PNV simulation, a comprehensive comparison of their performance is still unexplored. Therefore, we overview the performance of PNV mapping in terms of 12 commonly used methods with varying spatial scales and sample sizes. Our findings indicate that the methodology should be carefully selected due to the variation in performance of different model types, with Area Under the Curve (AUC) values ranging from 0.65 to 0.95 for models with sample sizes up to 80% of the total sample size. Specifically, semi-supervised learning performs best with small sample sizes (i.e., 10 to 200), while Random Forest, XGBoost, and artificial neural networks perform better with large sample sizes (i.e., over 500). Further, the performance of all models tends to improve significantly as the sample size increases and the grain size of the crystals becomes smaller. Take the downstream Tarim River Basin, a hyper-arid region undergoing ecological restoration, as a case study. We showed that its potential restored areas were overestimated by 2-3 fold as the spatial scale became coarser, revealing the caution needed while planning restoration projects at coarse resolution. These findings enhance the application of PNV in the design of restoration programs to prevent desertification.

Mapping every adult baobab (Adansonia digitata L.) across the Sahel and relationships to rural livelihoods

The baobab tree (Adansonia digitata L.) is an integral part of rural livelihoods throughout the African continent. However, the combined effects of climate change and increasing global demand for baobab products are currently exerting pressure on the sustainable utilization of these resources. Here we use sub-metre-resolution satellite imagery to identify the presence of nearly 2.8 million (underestimation bias 27.1%) baobab trees in the Sahel, a dryland region of 2.4 million km2. This achievement is considered an essential step towards an improved management and monitoring system of valuable woody species. Using Senegal as a case country, we find that 94% of rural buildings have at least one baobab tree in their immediate surroundings and that the abundance of baobabs is associated with a higher likelihood of people consuming a highly nutritious food group: dark green leafy vegetables. The generated database showcases the feasibility of mapping the location of single tree species at a sub-continental scale, providing vital information in times when deforestation and climate change cause the extinction of numerous tree species.

Regional uniqueness of tree species composition and response to forest loss and climate change

The conservation and restoration of forest ecosystems require detailed knowledge of the native plant compositions. Here, we map global forest tree composition and assess the impacts of historical forest cover loss and climate change on trees. The global occupancy of 10,590 tree species reveals complex taxonomic and phylogenetic gradients determining a local signature of tree lineage assembly. Species occupancy analyses indicate that historical forest loss has significantly restricted the potential suitable range of tree species in all forest biomes. Nevertheless, tropical moist and boreal forest biomes display the lowest level of range restriction and harbor extremely large ranged tree species, albeit with a stark contrast in richness and composition. Climate change simulations indicate that forest biomes are projected to differ in their response to climate change, with the highest predicted species loss in tropical dry and Mediterranean ecoregions. Our findings highlight the need for preserving the remaining large forest biomes while regenerating degraded forests in a way that provides resilience against climate change.

Shrub leaf area and leaf vein trait trade-offs in response to the light environment in a vegetation transitional zone

The leaf is an important site for energy acquisition and material transformation in plants. Leaf functional traits and their trade-off mechanisms reflect the resource utilisation efficiency and habitat adaptation strategies of plants, and contribute to our understanding of the mechanism by which the distribution pattern of plant populations in arid and semi-arid areas influences the evolution of vegetation structure and function. We selected two natural environments, the tree-shrub community canopy area and the shrub-grass community open area in the transition zone between the Qinghai-Tibet Plateau and the Loess Plateau. We studied the trade-off relationships of leaf area with leaf midvein diameter and leaf vein density in Cotoneaster multiflorus using the standardised major axis (SMA) method. The results show that the growth pattern of C. multiflorus , which has small leaves of high density and extremely small vein diameters, in the open area. The water use efficiency and net photosynthetic rate of plants in the open area were significantly greater than those of plants growing in the canopy area. The adaptability of C. multiflorus to environments with high light and low soil water content reflects its spatial colonisation potential in arid and semiarid mountains.

Automatedly identify dryland threatened species at large scale by using deep learning

Dryland biodiversity is decreasing at an alarming rate. Advanced intelligent tools are urgently needed to rapidly, automatedly, and precisely detect dryland threatened species on a large scale for biological conservation. Here, we explored the performance of three deep convolutional neural networks (Deeplabv3+, Unet, and Pspnet models) on the intelligent recognition of rare species based on high-resolution (0.3 m) satellite images taken by an unmanned aerial vehicle (UAV). We focused on a threatened species, Populus euphratica, in the Tarim River Basin (China), where there has been a severe population decline in the 1970s and restoration has been carried out since 2000. The testing results showed that Unet outperforms Deeplabv3+ and Pspnet when the training samples are lower, while Deeplabv3+ performs best as the dataset increases. Overall, when training samples are 80, Deeplabv3+ had the best overall performance for Populus euphratica identification, with mean pixel accuracy (MPA) between 87.31 % and 90.2 %, which, on average is 3.74 % and 11.29 % higher than Unet and Pspnet, respectively. Deeplabv3+ can accurately detect the boundaries of Populus euphratica even in areas of dense vegetation, with lower identification uncertainty for each pixel than other models. This study developed a UAV imagery-based identification framework using deep learning with high resolution in large-scale regions. This approach can accurately capture the variation in dryland threatened species, especially those in inaccessible areas, thereby fostering rapid and efficient conservation actions.

Fine-grained urban blue-green-gray landscape dataset for 36 Chinese cities based on deep learning network

Detailed and accurate urban landscape mapping, especially for urban blue-green-gray (UBGG) continuum, is the fundamental first step to understanding human-nature coupled urban systems. Nevertheless, the intricate spatial heterogeneity of urban landscapes within cities and across urban agglomerations presents challenges for large-scale and fine-grained mapping. In this study, we generated a 3 m high-resolution UBGG landscape dataset (UBGG-3m) for 36 Chinese metropolises using a transferable multi-scale high-resolution convolutional neural network and 336 Planet images. To train the network for generalization, we also created a large-volume UBGG landscape sample dataset (UBGGset) covering 2,272 km2 of urban landscape samples at 3 m resolution. The classification results for five cities across diverse geographic regions substantiate the superior accuracy of UBGG-3m in both visual interpretation and quantitative evaluation (with an overall accuracy of 91.2% and FWIoU of 83.9%). Comparative analyses with existing datasets underscore the UBGG-3m's great capability to depict urban landscape heterogeneity, providing a wealth of new data and valuable insights into the complex and dynamic urban environments in Chinese metropolises.

Density affects plant size in the Gobi Desert

Plant size is a crucial functional trait with substantial implications in agronomy and forestry. Understanding the factors influencing plant size is essential for ecosystem management and restoration efforts. Various environmental factors and plant density play significant roles in plant size. However, how plant size responds to mean annual precipitation (MAP), mean annual temperature (MAT), and density in the arid areas remains incomplete. To address this knowledge gap, we conducted comprehensive vegetation surveys in the Gobi Desert in northwestern China with a MAP below 250 mm. We also collected climate data to disentangle the respective influences of climate and density on the community-weighted plant height, crown length, and crown width. Our observations revealed that the community-weighted mean plant height, crown length, and width demonstrated a positive association with MAT but negative relationships with both MAP and density. These patterns can be attributed to the predominance of shrubs over herbs in arid regions, as shrubs tend to be larger in size. The proportion of shrubs increases with MAT, while it decreases with MAP and density, resulting in higher plant height and larger crown dimensions. Although both MAP and MAT affect plant size in the Gobi Desert, our findings highlight the stronger role of plant density in regulating plant size, indicating that the surrounding plant community and competition among individuals are crucial drivers of plant size patterns. Our findings provide valuable guidance for nature-based solutions for vegetation restoration and ecosystem management, highlighting the importance of considering plant density as a key factor when designing and implementing restoration strategies in arid areas.

Silvopastoral management for lowering trade-offs between beef production and carbon storage in tropical dry woodlands

Tropical dry woodlands and savannas harbour high levels of biodiversity and carbon, but are also important regions for agricultural production. This generates trade-offs between agriculture and the environment, as agricultural expansion and intensification typically involve the removal of natural woody vegetation. Cattle ranching is an expanding land use in many of these regions, but how different forms of ranching mediate the production/environment trade-off remains weakly understood. Here, we focus on the Argentine Chaco, to evaluate trade-offs between beef production and carbon storage in grazing systems with different levels of woody cover (n = 27). We measured beef productivity and carbon storage during 2018/19 and used a regression framework to quantify the trade-off between both, and to analyze which agroclimatic and management variables explain the observed trade-off. Our main finding was that silvopastures had the lowest trade-off between beef production and carbon storage, as management in these systems seeks to increase herbaceous forage by removing shrubs, while maintaining most of the bigger trees that contain most above-ground carbon. The most important variable explaining the beef production/carbon storage trade-off was pasture management, specifically the number of shrub encroachment control interventions, with a lower trade-off for higher numbers of interventions. Unfortunately, more interventions can also result in woody cover degradation over time, and shrub encroachment management must therefore be improved to become sustainable. Overall, our study highlights the strong environmental trade-offs associated with beef production in dry woodlands and savanna, but also the key role of good management practices in lowering this trade-off. Specifically, silvopastoral systems can increase beef production as much as converting woodlands to tree-less pastures, but silvopastures retain much more carbon in aboveground vegetation. Silvopastoral systems thus represent a promising land-use option to lower production/environment trade-offs in the Dry Chaco and likely many other tropical dry woodlands and savannas.

Even low levels of tree cover improve dietary quality in West Africa

Forests are attracting attention as a promising avenue to provide nutritious and "free" food without damaging the environment. Yet, we lack knowledge on the extent to which this holds in areas with sparse tree cover, such as in West Africa. This is largely due to the fact that existing methods are poorly designed to quantify tree cover in drylands. In this study, we estimate how various levels of tree cover across West Africa affect children's (aged 12-59 months) consumption of vitamin A-rich foods. We do so by combining detailed tree cover estimates based on PlanetScope imagery (3 m resolution) with Demographic Health Survey data from >15,000 households. We find that the probability of consuming vitamin A-rich foods increases from 0.45 to 0.53 with an increase in tree cover from the median value of 8.8 to 16.8% (which is the tree cover level at which the predicted probability of consuming vitamin A-rich foods is the highest). Moreover, we observe that the effects of tree cover vary across poverty levels and ecoregions. The poor are more likely than the non-poor to consume vitamin A-rich foods at low levels of tree cover in the lowland forest-savanna ecoregions, whereas the difference between poor and non-poor is less pronounced in the Sahel-Sudan. These results highlight the importance of trees and forests in sustainable food system transformation, even in areas with sparse tree cover.

Scattered tree death contributes to substantial forest loss in California

In recent years, large-scale tree mortality events linked to global change have occurred around the world. Current forest monitoring methods are crucial for identifying mortality hotspots, but systematic assessments of isolated or scattered dead trees over large areas are needed to reduce uncertainty on the actual extent of tree mortality. Here, we mapped individual dead trees in California using sub-meter resolution aerial photographs from 2020 and deep learning-based dead tree detection. We identified 91.4 million dead trees over 27.8 million hectares of vegetated areas (16.7-24.7% underestimation bias when compared to field data). Among these, a total of 19.5 million dead trees appeared isolated, and 60% of all dead trees occurred in small groups ( ≤ 3 dead trees within a 30 × 30 m grid), which is largely undetected by other state-level monitoring methods. The widespread mortality of individual trees impacts the carbon budget and sequestration capacity of California forests and can be considered a threat to forest health and a fuel source for future wildfires.

Mapping the number of mangrove trees in the Guangdong-Hong Kong-Macao Greater Bay Area

Mangroves are vital components of coastal ecosystems. Due to the complex canopy morphology and dense distribution of mangroves, it is challenging to accurately estimate the density based on satellite data. In this study, a density regression-based mangrove mapping network is proposed. The network can capture the multi-scale characteristics of mangroves through the combination of an attention mechanism and a parallel segmentation path, and its performance is better than existing methods. We then apply it to mapping the Greater Bay Area (GBA) the number of mangrove trees. The results show about 2.55 million mangrove trees in the GBA, with an average density of 782 trees per hectare. The tree number of mangroves on the beach is significantly higher than those distributed along the riverbank. This study is the first to achieve mangrove tree count mapping, opening up new prospects for applying satellite-based mangrove monitoring.

The overlooked contribution of trees outside forests to tree cover and woody biomass across Europe

Trees are an integral part in European landscapes, but only forest resources are systematically assessed by national inventories. The contribution of urban and agricultural trees to national-level carbon stocks remains largely unknown. Here we produced canopy cover, height and above-ground biomass maps from 3-meter resolution nanosatellite imagery across Europe. Our biomass estimates have a systematic bias of 7.6% (overestimation; R = 0.98) compared to national inventories of 30 countries, and our dataset is sufficiently highly resolved spatially to support the inclusion of tree biomass outside forests, which we quantify to 0.8 petagrams. Although this represents only 2% of the total tree biomass, large variations between countries are found (10% for UK) and trees in urban areas contribute substantially to national carbon stocks (8% for the Netherlands). The agreement with national inventory data, the scalability, and spatial details across landscapes, including trees outside forests, make our approach attractive for operational implementation to support national carbon stock inventory schemes.

Challenges in studying water fluxes within the soil-plant-atmosphere continuum: A tracer-based perspective on pathways to progress

Tracing and quantifying water fluxes in the hydrological cycle is crucial for understanding the current state of ecohydrological systems and their vulnerability to environmental change. Especially the interface between ecosystems and the atmosphere that is strongly mediated by plants is important to meaningfully describe ecohydrological system functioning. Many of the dynamic interactions generated by water fluxes between soil, plant and the atmosphere are not well understood, which is partly due to a lack of interdisciplinary research. This opinion paper reflects the outcome of a discussion among hydrologists, plant ecophysiologists and soil scientists on open questions and new opportunities for collaborative research on the topic "water fluxes in the soil-plant-atmosphere continuum" especially focusing on environmental and artificial tracers. We emphasize the need for a multi-scale experimental approach, where a hypothesis is tested at multiple spatial scales and under diverse environmental conditions to better describe the small-scale processes (i.e., causes) that lead to large-scale patterns of ecosystem functioning (i.e., consequences). Novel in-situ, high-frequency measurement techniques offer the opportunity to sample data at a high spatial and temporal resolution needed to understand the underlying processes. We advocate for a combination of long-term natural abundance measurements and event-based approaches. Multiple environmental and artificial tracers, such as stable isotopes, and a suite of experimental and analytical approaches should be combined to complement information gained by different methods. Virtual experiments using process-based models should be used to inform sampling campaigns and field experiments, e.g., to improve experimental designs and to simulate experimental outcomes. On the other hand, experimental data are a pre-requisite to improve our currently incomplete models. Interdisciplinary collaboration will help to overcome research gaps that overlap across different earth system science fields and help to generate a more holistic view of water fluxes between soil, plant and atmosphere in diverse ecosystems.

More than one quarter of Africa's tree cover is found outside areas previously classified as forest

The consistent monitoring of trees both inside and outside of forests is key to sustainable land management. Current monitoring systems either ignore trees outside forests or are too expensive to be applied consistently across countries on a repeated basis. Here we use the PlanetScope nanosatellite constellation, which delivers global very high-resolution daily imagery, to map both forest and non-forest tree cover for continental Africa using images from a single year. Our prototype map of 2019 (RMSE = 9.57%, bias = -6.9%). demonstrates that a precise assessment of all tree-based ecosystems is possible at continental scale, and reveals that 29% of tree cover is found outside areas previously classified as tree cover in state-of-the-art maps, such as in croplands and grassland. Such accurate mapping of tree cover down to the level of individual trees and consistent among countries has the potential to redefine land use impacts in non-forest landscapes, move beyond the need for forest definitions, and build the basis for natural climate solutions and tree-related studies.

U-Sleep's resilience to AASM guidelines

AASM guidelines are the result of decades of efforts aiming at standardizing sleep scoring procedure, with the final goal of sharing a worldwide common methodology. The guidelines cover several aspects from the technical/digital specifications, e.g., recommended EEG derivations, to detailed sleep scoring rules accordingly to age. Automated sleep scoring systems have always largely exploited the standards as fundamental guidelines. In this context, deep learning has demonstrated better performance compared to classical machine learning. Our present work shows that a deep learning-based sleep scoring algorithm may not need to fully exploit the clinical knowledge or to strictly adhere to the AASM guidelines. Specifically, we demonstrate that U-Sleep, a state-of-the-art sleep scoring algorithm, can be strong enough to solve the scoring task even using clinically non-recommended or non-conventional derivations, and with no need to exploit information about the chronological age of the subjects. We finally strengthen a well-known finding that using data from multiple data centers always results in a better performing model compared with training on a single cohort. Indeed, we show that this latter statement is still valid even by increasing the size and the heterogeneity of the single data cohort. In all our experiments we used 28528 polysomnography studies from 13 different clinical studies.

Sub-continental-scale carbon stocks of individual trees in African drylands

The distribution of dryland trees and their density, cover, size, mass and carbon content are not well known at sub-continental to continental scales1-14. This information is important for ecological protection, carbon accounting, climate mitigation and restoration efforts of dryland ecosystems15-18. We assessed more than 9.9 billion trees derived from more than 300,000 satellite images, covering semi-arid sub-Saharan Africa north of the Equator. We attributed wood, foliage and root carbon to every tree in the 0-1,000 mm year-1 rainfall zone by coupling field data19, machine learning20-22, satellite data and high-performance computing. Average carbon stocks of individual trees ranged from 0.54 Mg C ha-1 and 63 kg C tree-1 in the arid zone to 3.7 Mg C ha-1 and 98 kg tree-1 in the sub-humid zone. Overall, we estimated the total carbon for our study area to be 0.84 (±19.8%) Pg C. Comparisons with 14 previous TRENDY numerical simulation studies23 for our area found that the density and carbon stocks of scattered trees have been underestimated by three models and overestimated by 11 models, respectively. This benchmarking can help understand the carbon cycle and address concerns about land degradation24-29. We make available a linked database of wood mass, foliage mass, root mass and carbon stock of each tree for scientists, policymakers, dryland-restoration practitioners and farmers, who can use it to estimate farmland tree carbon stocks from tablets or laptops.

Nation-wide mapping of tree-level aboveground carbon stocks in Rwanda

Trees sustain livelihoods and mitigate climate change but a predominance of trees outside forests and limited resources make it difficult for many tropical countries to conduct automated nation-wide inventories. Here, we propose an approach to map the carbon stock of each individual overstory tree at the national scale of Rwanda using aerial imagery from 2008 and deep learning. We show that 72% of the mapped trees are located in farmlands and savannas and 17% in plantations, accounting for 48.6% of the national aboveground carbon stocks. Natural forests cover 11% of the total tree count and 51.4% of the national carbon stocks, with an overall carbon stock uncertainty of 16.9%. The mapping of all trees allows partitioning to any landscapes classification and is urgently needed for effective planning and monitoring of restoration activities as well as for optimization of carbon sequestration, biodiversity and economic benefits of trees.

A spatialized assessment of ecosystem service relationships in a multifunctional agroforestry landscape of Senegal

Agroforestry systems are an integral part of Sub-Saharan agricultural landscapes. Studies conducted at tree or plot scales on the supply of ecosystem services (ES) suggest that agroforestry practices are a promising way to build multifunctional agricultural landscapes. However, the current characterization and understanding of how multiple ES are associated across such heterogeneous agricultural landscapes are still limited. This study provides the first characterization of the multiple ESs supplied by a Sahelian Faidherbia albida agroforestry parkland and their relationships. Relying on field data for 11 ES indicators, recent advances in remote sensing-derived information, and blending different ES mapping approaches, we first assessed the spatial heterogeneity of the supply of each ES. We found that the majority of ES indicators remained below ES potential values over the study area by 25 % to 50 %, revealing that there is a considerable scope for increasing the ES supply in the F. albida parkland. Then, using a scoring approach, we analyzed the supply of multiple ESs. We observed a large number of hotspots and a clear effect of the proximity of F. albida trees fostering the supply of multiple ESs in their vicinity. Finally, we mapped and analyzed the dominant relationships - trade-offs, synergies or losses - between ESs from a cooccurrence spatial approach. We showed that significant trade-offs and losses (58 % of the area) between ESs can exist in the F. albida parkland. Interestingly, we also showed that synergies occurred mainly up to 10 m from the F. albida trees, suggesting that synergies need to be increased beyond this threshold. By adopting an original ES valuation framework, we provided basic insights into ESs and their relationships. The different maps and information generated can support public debates and target new policies fostering the multifunctionality of F. albida parklands as well as in various other parklands of West Africa.

Towards the fully automated monitoring of ecological communities

High-resolution monitoring is fundamental to understand ecosystems dynamics in an era of global change and biodiversity declines. While real-time and automated monitoring of abiotic components has been possible for some time, monitoring biotic components-for example, individual behaviours and traits, and species abundance and distribution-is far more challenging. Recent technological advancements offer potential solutions to achieve this through: (i) increasingly affordable high-throughput recording hardware, which can collect rich multidimensional data, and (ii) increasingly accessible artificial intelligence approaches, which can extract ecological knowledge from large datasets. However, automating the monitoring of facets of ecological communities via such technologies has primarily been achieved at low spatiotemporal resolutions within limited steps of the monitoring workflow. Here, we review existing technologies for data recording and processing that enable automated monitoring of ecological communities. We then present novel frameworks that combine such technologies, forming fully automated pipelines to detect, track, classify and count multiple species, and record behavioural and morphological traits, at resolutions which have previously been impossible to achieve. Based on these rapidly developing technologies, we illustrate a solution to one of the greatest challenges in ecology: the ability to rapidly generate high-resolution, multidimensional and standardised data across complex ecologies.

Automated mapping of Portulacaria afra canopies for restoration monitoring with convolutional neural networks and heterogeneous unmanned aerial vehicle imagery

Ecosystem restoration and reforestation often operate at large scales, whereas monitoring practices are usually limited to spatially restricted field measurements that are (i) time- and labour-intensive, and (ii) unable to accurately quantify restoration success over hundreds to thousands of hectares. Recent advances in remote sensing technologies paired with deep learning algorithms provide an unprecedented opportunity for monitoring changes in vegetation cover at spatial and temporal scales. Such data can feed directly into adaptive management practices and provide insights into restoration and regeneration dynamics. Here, we demonstrate that convolutional neural network (CNN) segmentation algorithms can accurately classify the canopy cover of Portulacaria afra Jacq. in imagery acquired using different models of unoccupied aerial vehicles (UAVs) and under variable light intensities. Portulacaria afra is the target species for the restoration of Albany Subtropical Thicket vegetation, endemic to South Africa, where canopy cover is challenging to measure due to the dense, tangled structure of this vegetation. The automated classification strategy presented here is widely transferable to restoration monitoring as its application does not require any knowledge of the CNN model or specialist training, and can be applied to imagery generated by a range of UAV models. This will reduce the sampling effort required to track restoration trajectories in space and time, contributing to more effective management of restoration sites, and promoting collaboration between scientists, practitioners and landowners.

Mapping global lake dynamics reveals the emerging roles of small lakes

Lakes are important natural resources and carbon gas emitters and are undergoing rapid changes worldwide in response to climate change and human activities. A detailed global characterization of lakes and their long-term dynamics does not exist, which is however crucial for evaluating the associated impacts on water availability and carbon emissions. Here, we map 3.4 million lakes on a global scale, including their explicit maximum extents and probability-weighted area changes over the past four decades. From the beginning period (1984–1999) to the end (2010–2019), the lake area increased across all six continents analyzed, with a net change of +46,278 km², and 56% of the expansion was attributed to reservoirs. Interestingly, although small lakes (<1 km²) accounted for just 15% of the global lake area, they dominated the variability in total lake size in half of the global inland lake regions. The identified lake area increase over time led to higher lacustrine carbon emissions, mostly attributed to small lakes. Our findings illustrate the emerging roles of small lakes in regulating not only local inland water variability, but also the global trends of surface water extent and carbon emissions.

Lakes are essential components of the hydrological and biogeochemical cycles. Here, Pi et al develop a global lake dataset called GLAKES via high-resolution satellite images and deep learning to examine global lake changes over four decades.

Seeing the System from Above: The Use and Potential of Remote Sensing for Studying Ecosystem Dynamics

Remote sensing techniques are increasingly used for studying ecosystem dynamics, delivering spatially explicit information on the properties of Earth over large spatial and multi-decadal temporal extents. Yet, there is still a gap between the more technology-driven development of novel remote sensing techniques and their applications for studying ecosystem dynamics. Here, I review the existing literature to explore how addressing these gaps might enable recent methods to overcome longstanding challenges in ecological research. First, I trace the emergence of remote sensing as a major tool for understanding ecosystem dynamics. Second, I examine recent developments in the field of remote sensing that are of particular importance for studying ecosystem dynamics. Third, I consider opportunities and challenges for emerging open data and software policies and suggest that remote sensing is at its most powerful when it is theoretically motivated and rigorously ground-truthed. I close with an outlook on four exciting new research frontiers that will define remote sensing ecology in the upcoming decade.

Individual tree crown delineation in high-resolution remote sensing images based on U-Net

We present a deep learning-based framework for individual tree crown delineation in aerial and satellite images. This is an important task, e.g., for forest yield or carbon stock estimation. In contrast to earlier work, the presented method creates irregular polygons instead of bounding boxes and also provides a tree cover mask for areas that are not separable. Furthermore, it is trainable with low amounts of training data and does not need 3D height information from, e.g., laser sensors. We tested the approach in two scenarios: (1) with 30 cm WorldView-3 satellite imagery from an urban region in Bengaluru, India, and (2) with 5 cm aerial imagery of a densely forested area near Gartow, Germany. The intersection over union between the reference and predicted tree cover mask is 71.2% for the satellite imagery and 81.9% for the aerial images. On the polygon level, the method reaches an accuracy of 46.3% and a recall of 63.7% in the satellite images and an accuracy of 52% and recall of 66.2% in the aerial images, which is comparable to previous works that only predicted bounding boxes. Depending on the image resolution, limitations to separate individual tree crowns occur in situations where trees are hardly separable even for human image interpreters (e.g., homogeneous canopies, very small trees). The results indicate that the presented approach can efficiently delineate individual tree crowns in high-resolution optical images. Given the high availability of such imagery, the framework provides a powerful tool for tree monitoring. The source code and pretrained weights are publicly available at https://github.com/AWF-GAUG/TreeCrownDelineation.

Higher leaf nitrogen content is linked to tighter stomatal regulation of transpiration and more efficient water use across dryland trees

The least-cost economic theory of photosynthesis shows that water and nitrogen are mutually substitutable resources to achieve a given carbon gain. However, vegetation in the Sahel has to cope with the dual challenge imposed by drought and nutrient-poor soils. We addressed how variation in leaf nitrogen per area (N_area ) modulates leaf oxygen and carbon isotopic composition (δ¹⁸ O, δ¹³ C), as proxies of stomatal conductance and water-use efficiency, across 34 Sahelian woody species. Dryland species exhibited diverging leaf δ¹⁸ O and δ¹³ C values, indicating large interspecific variation in time-integrated stomatal conductance and water-use efficiency. Structural equation modeling revealed that leaf N_area is a pivotal trait linked to multiple water-use traits. Leaf N_area was positively linked to both δ¹⁸ O and δ¹³ C, suggesting higher carboxylation capacity and tighter stomatal regulation of transpiration in N-rich species, which allows them to achieve higher water-use efficiency and more conservative water use. These adaptations represent a key physiological advantage of N-rich species, such as legumes, that could contribute to their dominance across many dryland regions. This is the first report of a robust mechanistic link between leaf N_area and δ¹⁸ O in dryland vegetation that is consistent with core principles of plant physiology.

Climate legacies drive the distribution and future restoration potential of dryland forests

Knowing the extent and environmental drivers of forests is key to successfully restore degraded ecosystems, and to mitigate climate change and desertification impacts using tree planting. Water availability is the main limiting factor for the development of forests in drylands, yet the importance of groundwater resources and palaeoclimate as drivers of their current distribution has been neglected. Here we report that mid-Holocene climates and aquifer trends are key predictors of the distribution of dryland forests worldwide. We also updated the global extent of dryland forests to 1,283 million hectares and showed that failing to consider past climates and aquifers has resulted in ignoring or misplacing up to 130 million hectares of forests in drylands. Our findings highlight the importance of a wetter past and well-preserved aquifers to explain the current distribution of dryland forests, and can guide restoration actions by avoiding unsuitable areas for tree establishment in a drier world.

Investigation of Long-Term Forest Dynamics in Protected Areas of Northeast China Using Landsat Data

Forest dynamics, including forest loss and gain, are long-term complex ecological processes affected by nature and human activities. It is particularly important to understand the long-term forest dynamics of protected areas to evaluate their conservation efforts. This study adopted the Landsat tree-canopy cover (TCC) method to derive annual TCC data for the period 1984–2020 for the protected areas of northeast China, where protection policies have been carried out since the end of the 20th century, e.g., the Natural Forest Conversion Program (NFCP). A strong correlation was found between the TCC estimates derived from Landsat and LiDAR observations, suggesting the high accuracy of TCC. Forest loss and gain events were also identified from the time series of TCC estimates. High correlations were reported for both forest loss (Producer’s accuracy = 85.21%; User’s accuracy = 84.26%) and gain (Producer’s accuracy = 87.74%; User’s accuracy = 88.31%), suggesting that the approach can be used for monitoring and evaluating the effectiveness of the NFCP and other forest conservation efforts. The results revealed a fluctuating upward trend of the TCC of the protected area from 1986 to 2018. The increased area of TCC was much larger than the decreased area, accounting for 59.68% and 40.34%, respectively, suggesting the effectiveness of protection policies. Since the NFCP was officially implemented in 1998, deforestation was effectively curbed, the area of forest loss was significantly reduced (slope: −14.24%/year), and the area of forest gain significantly increased (slope: 4.13%/year). We found that regional forest changes were mainly manifested in “forest gain after loss (forest recovery)” and “forest gain with no loss (forest newborn)”, accounting for 40.29% and 37.28% of the total area of forest change, respectively. Moreover, the forest gain area far exceeds the forest loss area, reaching 11.24 million hectares, suggesting a successful forest recovery due to forest protection.

Pleistocene drivers of Northwest African hydroclimate and vegetation

Savanna ecosystems were the landscapes for human evolution and are vital to modern Sub-Saharan African food security, yet the fundamental drivers of climate and ecology in these ecosystems remain unclear. Here we generate plant-wax isotope and dust flux records to explore the mechanistic drivers of the Northwest African monsoon, and to assess ecosystem responses to changes in monsoon rainfall and atmospheric pCO₂. We show that monsoon rainfall is controlled by low-latitude insolation gradients and that while increases in precipitation are associated with expansion of grasslands into desert landscapes, changes in pCO₂ predominantly drive the C₃/C₄ composition of savanna ecosystems.

Leveraging OpenStreetMap and Multimodal Remote Sensing Data with Joint Deep Learning for Wastewater Treatment Plants Detection

Humans rely on clean water for their health, well-being, and various socio-economic activities. During the past few years, the COVID-19 pandemic has been a constant reminder of about the importance of hygiene and sanitation for public health. The most common approach to securing clean water supplies for this purpose is via wastewater treatment. To date, an effective method of detecting wastewater treatment plants (WWTP) accurately and automatically via remote sensing is unavailable. In this paper, we provide a solution to this task by proposing a novel joint deep learning (JDL) method that consists of a fine-tuned object detection network and a multi-task residual attention network (RAN). By leveraging OpenStreetMap (OSM) and multimodal remote sensing (RS) data, our JDL method is able to simultaneously tackle two different tasks: land use land cover (LULC) and WWTP classification. Moreover, JDL exploits the complementary effects between these tasks for a performance gain. We train JDL using 4,187 WWTP features and 4,200 LULC samples and validate the performance of the proposed method over a selected area around Stuttgart with 723 WWTP features and 1,200 LULC samples to generate an LULC classification map and a WWTP detection map. Extensive experiments conducted with different comparative methods demonstrate the effectiveness and efficiency of our JDL method in automatic WWTP detection in comparison with single-modality/single-task or traditional survey methods. Moreover, lessons learned pave the way for future works to simultaneously and effectively address multiple large-scale mapping tasks (e.g., both mapping LULC and detecting WWTP) from multimodal RS data via deep learning.

A Review of General Methods for Quantifying and Estimating Urban Trees and Biomass

Understanding the biomass, characteristics, and carbon sequestration of urban forests is crucial for maintaining and improving the quality of life and ensuring sustainable urban planning. Approaches to urban forest management have been incorporated into interdisciplinary, multifunctional, and technical efforts. In this review, we evaluate recent developments in urban forest research methods, compare the accuracy and efficiency of different methods, and identify emerging themes in urban forest assessment. This review focuses on urban forest biomass estimation and individual tree feature detection, showing that the rapid development of remote sensing technology and applications in recent years has greatly benefited the study of forest dynamics. Included in the review are light detection and ranging-based techniques for estimating urban forest biomass, deep learning algorithms that can extract tree crowns and identify tree species, methods for measuring large canopies using unmanned aerial vehicles to estimate forest structure, and approaches for capturing street tree information using street view images. Conventional methods based on field measurements are highly beneficial for accurately recording species-specific characteristics. There is an urgent need to combine multi-scale and spatiotemporal methods to improve urban forest detection at different scales.

Searching thousands of genomes to classify somatic and novel structural variants using STIX

Structural variants are associated with cancers and developmental disorders, but challenges with estimating population frequency remain a barrier to prioritizing mutations over inherited variants. In particular, variability in variant calling heuristics and filtering limits the use of current structural variant catalogs. We present STIX, a method that, instead of relying on variant calls, indexes and searches the raw alignments from thousands of samples to enable more comprehensive allele frequency estimation.

This work describes a strategy for estimating the population frequency of structural variations by searching the raw alignments of large population sequencing samples using the STIX framework.

Indirect effects of African megaherbivore conservation on bat diversity in the world's oldest desert

In extreme environments, temperature and precipitation are often the main forces responsible for structuring ecological communities and species distributions. The role of biotic interactions is typically thought to be minimal. By clustering around rare and isolated features, like surface water, however, effects of herbivory by desert-dwelling wildlife can be amplified. Understanding how species interact in these environments is critical to safeguarding vulnerable or data-deficient species. We examined whether African elephants (Loxodonta africana), black rhinoceros (Diceros bicornis), and southern giraffe (Giraffa giraffa) modulate insectivorous bat communities around permanent waterholes in the Namib Desert. We estimated megaherbivore use of sites based on dung transects, summarized vegetation productivity from satellite measurements of the normalized difference vegetation index, and surveyed local bat communities acoustically. We used structural equation models to identify relationships among megaherbivores and bat species richness and dry- (November 2016-January 2017) and wet- (February-May 2017) season bat activity. Site-level megaherbivore use in the dry season was positively associated with bat activity-particularly that of open-air foragers-and species richness through indirect pathways. When resources were more abundant (wet season), however, these relationships were weakened. Our results indicate that biotic interactions contribute to species distributions in desert areas and suggest the conservation of megaherbivores in this ecosystem may indirectly benefit insectivorous bat abundance and diversity. Given that how misunderstood and understudied most bats are relative to other mammals, such findings suggest that managers pursue short-term solutions (e.g., community game guard programs, water-point protection near human settlements, and ecotourism) to indirectly promote bat conservation and that research includes megaherbivores' effects on biodiversity at other trophic levels.

The Role of Trees Outside Forests in the Cultural Landscape of the Colline del Prosecco UNESCO Site

The multifunctional role of Trees Outside Forests (TOF) is largely recognized in scientific literature, but they are still rarely considered in forest inventories and planning, with consequent underestimation of their role and amount. In addition, their cultural role has rarely been considered both at scientific and management level as well as in UNESCO sites. TOF characterize many European cultural landscapes, including the one of the Colline del Prosecco, inscribed in 2019 in the UNESCO World Heritage List. One of the reasons of the inclusion, in fact, is the landscape mosaic made of vineyards interspersed with small woodlands and tree rows. This paper focuses on two types of TOF, Small Woods and Linear Tree Formations (TOF NON A/U). Their detailed mapping and the performing of different spatial analysis allowed us to assess their role and to provide data for future monitoring and for local forest planning. Results confirmed that TOF NON A/U are one of the main features of the UNESCO site landscape: despite the limited overall surface (1.95% of the area), 931 different patches have been identified. Spatial analysis highlighted the key landscape and ecological roles, acting as intermediate features between large forest patches, and also an important role for hydrological protection (they can be found also in slopes above 80% of inclination). The study provided a detailed mapping and database of one of the main features of the Colline del Prosecco UNESCO site cultural landscape, verifying the multifunctional role of TOF NON A/U and the necessity to include them into local forest planning, but also suggesting their inclusion in national forest inventories.

How Does Sample Labeling and Distribution Affect the Accuracy and Efficiency of a Deep Learning Model for Individual Tree-Crown Detection and Delineation

Monitoring and assessing vegetation using deep learning approaches has shown promise in forestry applications. Sample labeling to represent forest complexity is the main limitation for deep learning approaches for remote sensing vegetation classification applications, and few studies have focused on the impact of sample labeling methods on model performance and model training efficiency. This study is the first-of-its-kind that uses Mask region-based convolutional neural networks (Mask R-CNN) to evaluate the influence of sample labeling methods (including sample size and sample distribution) on individual tree-crown detection and delineation. A flight was conducted over a plantation with Fokienia hodginsii as the main tree species using a Phantom4-Multispectral (P4M) to obtain UAV imagery, and a total of 2061 manually and accurately delineated tree crowns were used for training and validating (1689) and testing (372). First, the model performance of three pre-trained backbones (ResNet-34, ResNet-50, and ResNet-101) was evaluated. Second, random deleting and clumped deleting methods were used to repeatedly delete 10% from the original sample set to reduce the training and validation set, to simulate two different sample distributions (the random sample set and the clumped sample set). Both RGB image and Multi-band images derived from UAV flights were used to evaluate model performance. Each model’s average per-epoch training time was calculated to evaluate the model training efficiency. The results showed that ResNet-50 yielded a more robust network than ResNet-34 and ResNet-101 when the same parameters were used for Mask R-CNN. The sample size determined the influence of sample labeling methods on the model performance. Random sample labeling had lower requirements for sample size compared to clumped sample labeling, and unlabeled trees in random sample labeling had no impact on model training. Additionally, the model with clumped samples provides a shorter average per-epoch training time than the model with random samples. This study demonstrates that random sample labeling can greatly reduce the requirement of sample size, and it is not necessary to accurately label each sample in the image during the sample labeling process.

Mapping Canopy Cover in African Dry Forests from the Combined Use of Sentinel-1 and Sentinel-2 Data: Application to Tanzania for the Year 2018

High-resolution Earth observation data is routinely used to monitor tropical forests. However, the seasonality and openness of the canopy of dry tropical forests remains a challenge for optical sensors. In this study, we demonstrate the potential of combining Sentinel-1 (S1) SAR and Sentinel-2 (S2) optical sensors in order to map the tree cover in East Africa. The overall methodology consists of: (i) the generation of S1 and S2 layers, (ii) the collection of an expert-based training/validation dataset and (iii) the classification of the satellite data. Three different classification workflows, together with different approaches to incorporating the spatial information to train the classifiers, are explored. Two types of maps were derived from these mapping approaches over Tanzania: (i) binary tree cover–no tree cover (TC/NTC) maps, and (ii) maps of the canopy cover classes. The overall accuracy of the maps is >95% for the TC/NTC maps and >85% for the forest types maps. Considering the neighboring pixels for training the classification improved the mapping of the areas that are covered by 1–10% tree cover. The study relied on open data and publicly available tools and can be integrated into national monitoring systems.

A commentary of "Using satellites to map trees": 10 remarkable discoveries from 2020 in Nature

Brandt et al. reported the results of their analysis from high-resolution satellite images, covering more than 1.3 million square kilometers of the Western Sahara and Sahel in West Africa. They mapped the locations and sizes of approximately 1.8 billion trees. Prior to this, scientists had never made such a detailed map of trees in such a large area. Commercial satellites have begun to collect data and can detect small ground objects that are 1 square meter or less in size. Therefore, the field of terrestrial remote sensing may have a significant advance from mainly a comprehensive landscape-scale measurement to mapping the position and canopy size of each tree at a regional or even global scale. This progress will revolutionize how we think, monitor, simulate, and manage the global terrestrial ecosystem.

Multi-Species Individual Tree Segmentation and Identification Based on Improved Mask R-CNN and UAV Imagery in Mixed Forests

High-resolution UAV imagery paired with a convolutional neural network approach offers significant advantages in accurately measuring forestry ecosystems. Despite numerous studies existing for individual tree crown delineation, species classification, and quantity detection, the comprehensive situation in performing the above tasks simultaneously has rarely been explored, especially in mixed forests. In this study, we propose a new method for individual tree segmentation and identification based on the improved Mask R-CNN. For the optimized network, the fusion type in the feature pyramid network is modified from down-top to top-down to shorten the feature acquisition path among the different levels. Meanwhile, a boundary-weighted loss module is introduced to the cross-entropy loss function Lmask to refine the target loss. All geometric parameters (contour, the center of gravity and area) associated with canopies ultimately are extracted from the mask by a boundary segmentation algorithm. The results showed that F1-score and mAP for coniferous species were higher than 90%, and that of broadleaf species were located between 75%–85.44%. The producer’s accuracy of coniferous forests was distributed between 0.8–0.95 and that of broadleaf ranged in 0.87–0.93; user’s accuracy of coniferous was distributed between 0.81–0.84 and that of broadleaf ranged in 0.71–0.76. The total number of trees predicted was 50,041 for the entire study area, with an overall error of 5.11%. The method under study is compared with other networks including U-net and YOLOv3. Results in this study show that the improved Mask R-CNN has more advantages in broadleaf canopy segmentation and number detection.

Mapping Key Indicators of Forest Restoration in the Amazon Using a Low-Cost Drone and Artificial Intelligence

Monitoring the vegetation structure and species composition of forest restoration (FR) in the Brazilian Amazon is critical to ensuring its long-term benefits. Since remotely piloted aircrafts (RPAs) associated with deep learning (DL) are becoming powerful tools for vegetation monitoring, this study aims to use DL to automatically map individual crowns of Vismia (low resilience recovery indicator), Cecropia (fast recovery indicator), and trees in general (this study refers to individual crowns of all trees regardless of species as All Trees). Since All Trees can be accurately mapped, this study also aims to propose a tree crown heterogeneity index (TCHI), which estimates species diversity based on: the heterogeneity attributes/parameters of the RPA image inside the All Trees results; and the Shannon index measured by traditional fieldwork. Regarding the DL methods, this work evaluated the accuracy of the detection of individual objects, the quality of the delineation outlines and the area distribution. Except for Vismia delineation (IoU = 0.2), DL results presented accurate values in general, as F1 and IoU were always greater than 0.7 and 0.55, respectively, while Cecropia presented the most accurate results: F1 = 0.85 and IoU = 0.77. Since All Trees results were accurate, the TCHI was obtained through regression analysis between the canopy height model (CHM) heterogeneity attributes and the field plot data. Although TCHI presented robust parameters, such as p-value < 0.05, its results are considered preliminary because more data are needed to include different FR situations. Thus, the results of this work show that low-cost RPA has great potential for monitoring FR quality in the Amazon, because Vismia, Cecropia, and All Trees can be automatically mapped. Moreover, the TCHI preliminary results showed high potential in estimating species diversity. Future studies must assess domain adaptation methods for the DL results and different FR situations to improve the TCHI range of action.