Tracts Of Forest Research Articles

The effect of illicit crops on forest cover in Colombia

ABSTRACT Many armed conflicts worldwide occur in biodiversity hotspots and nearly 50% of those conflicts occur in forested regions. In Colombia, the armed conflict has implied the clearing of large forest tracts for the establishment of illicit crops. The aim of this study was to assess the role that illicit crops played in the deforestation dynamics in Colombia between 2001 and 2014. We established a database with the annual deforestation rates and nine predictors for 1120 municipalities and built fixed effects models that take spatial autocorrelation into account. Model selection with AIC suggested that the area cultivated with coca crops was the best predictor of annual rates of deforestation, whereas coca crop removal was associated with increasing forest cover. According to our results, coca crops promoted deforestation in Colombia between 2001 and 2014 through indirect (spilling-over to nearby areas), immediate and temporally-lagged mechanisms.

Home range patterns of Helmeted Woodpecker (Celeus galeatus), Lineated Woodpecker (Dryocopus lineatus), and Robust Woodpecker (Campephilus robustus) in Misiones, Argentina, in a global perspective

Home range mapping studies of birds inform about area requirements and responses to land management as reflected by home range sizes and by resource selection within home ranges. Tracking studies of woodpeckers (<em>Picidae</em>) so far have been concentrated in temperate regions. In the subtropical Atlantic Forest of northeast Argentina, we assessed interspecific differences in home range sizes in old-growth forests and selectively logged forests of Helmeted Woodpecker (<em>Celeus galeatus</em>, a globally threatened species), Lineated Woodpecker (<em>Dryocopus lineatus</em>), and Robust Woodpecker (<em>Campephilus robustus</em>). Helmeted Woodpecker had larger breeding home ranges in selectively logged forests, averaging 105 ± 39 ha for pairs, versus 60 ± 13 ha in old-growth forests. Lineated Woodpecker breeding home ranges of pairs averaged 56 ± 22 ha, and those of Robust Woodpecker 43 ± 22 ha, with no differences between forest types. Helmeted Woodpeckers had an unusual separation between the home ranges of males and females in breeding pairs, with a mean area overlap of only 8% ± 9% near the nest tree, resulting in large home ranges for pairs. Helmeted Woodpecker and Robust Woodpecker individuals that were followed into the post-breeding stage had marked expansions of their home range sizes relative to breeding home ranges. To place our findings with Atlantic Forest woodpeckers in perspective we reviewed whether woodpecker home range sizes increase with latitude and body mass globally. For 29 populations of 22 woodpecker species, a power regression model with these factors explained 24.1% of variation in breeding home range sizes, with 17.1% of variation explained by latitude alone. Woodpecker species with larger home ranges than predicted values were three species of North American woodpeckers of coniferous forests, as well as the Helmeted Woodpecker. Our results of smaller home ranges in old-growth forests for the Helmeted Woodpecker affirm an association of this species with such forests. We urge the conservation of the few remaining tracts of old-growth Atlantic Forest and more restoration of logged forests to mature conditions.

GENETIC DIVERSITY OF NATURAL AND RESTORED API-API PUTIH (AVICENNIA ALBA) POPULATIONS IN THE WEST COAST OF PENINSULAR MALAYSIA

A large tract of mangrove forest in Malaysia has been lost due to increased anthropogenic activities. Restorative practices of mangrove forest have been adopted nationwide to re-establish ecosystem services in combating coastal erosion. However, genetic considerations in local mangrove restoration practices are still far lacking despite the vast literature on their genetic diversity. To understand whether the restored mangroves can impact the genetic diversity distribution among natural populations, we used eight microsatellite markers to assess the genetic diversity of the Api-api putih (Avicennia alba) between the natural and restored populations along the west coast of Peninsular Malaysia. We found no difference in terms of genetic diversity between these populations. Two genetic clusters were detected among A. alba along the west coast based on Bayesian clustering and discriminant analysis of principal components (DAPC). The southwest monsoon current circulation that coincides the timing of seed dispersal of A. alba may explain such pattern of genetic differentiation. Despite the minimal genetic structure, our results suggest that seed sourcing from either population is viable for the local mangrove restoration programs in the future.

Schemes designing for transport development of the forest tract by using information and intelligent systems

A model of computer-aided design of the transport network of timber highways, which allows improving the road design system is considered The features of the preparation of automated initial data are adduced and their advantages for solving a linear model of the timber transport network are shown through an experiment. Keywords logging road, design of a transport network, problem of a transport linear programming

Forest fragments, primary and secondary forests harbour similar arthropod assemblages after 40 years of landscape regeneration in the Central Amazon

Abstract Large‐scale deforestation leads to habitat loss and fragmentation with major consequences for biodiversity. However, the impacts of these environmental disturbances can be mitigated by forest regeneration. We evaluated the effects of forest fragmentation and vegetation recovery on arthropod assemblages in human‐modified landscapes in the Amazon. Arthropods were sampled in 10‐ha forest fragments, continuous primary forests, and secondary forests using Malaise traps. We identified arthropods in multiple taxonomic resolutions from order to species. We collected 25,230 arthropods belonging to 17 orders. Coleoptera and Hemiptera were more abundant in secondary forests compared to other habitats. A dipteran family (Clusiidae) and a horse fly species (Dichelacera cervicornis—Tabanidae) were identified as indicators of forest fragments and continuous forests, respectively. Similar levels of taxonomic diversity between habitats were documented. The taxonomic composition also was similar between studied habitats in all taxonomic resolutions. Although the studied landscapes suffered strong environmental disturbances four decades ago, the vegetation has been in continuous regeneration since then, favouring the dispersal of arthropods throughout habitats. Forest fragments and old secondary forests contribute to the maintenance of a rich and diverse arthropod assemblage in landscapes composed of large tracts of continuous primary forests.

A method of sequentially generating a set of components of a multidimensional random variable using a nonparametric pattern recognition algorithm

We study in which way a priori information on the independence of random variables affects the approximation accuracy of a nonparametric estimate of the Rosenblatt–Parzen probability density. A new technique for generating sets of independent components of a multidimensional random variable is proposed. The methodology is based on testing the hypotheses of the independence of combinations of the multidimensional random variable components using a two-alternative nonparametric kernel algorithm for pattern recognition corresponding to the maximum likelihood criterion. Classes correspond to the domains of definition of the probability densities of sets of independent and dependent components of the multidimensional random variable. Nonparametric statistics of the kernel type are used to estimate the probability densities. The choice of the bandwidths of the kernel estimates of the probability densities is made from the condition of the minimum root-mean-square criterion. The sequential procedure for generating a set of independent components begins with the analysis of paired combinations of components of a multidimensional random variable. For each pair of components, the probability of an error in recognizing classes corresponding to the assumptions of independence and dependence of the considered components is estimated. A pair of components with the maximum difference between these errors is determined. If the errors obtained do not differ significantly, then there are no independent components in the considered multivariate random variable. If there is a significant difference in the probability estimates of class recognition errors, a pair of independent components is established. These components are included in a three-component set of a multidimensional random variable. The analysis of their combinations is carried out in the same way, following the above-described procedure. The process of generating the set of independent components is stopped when no reliable difference occurs any more between the probabilities of errors in recognizing situations belonging to the accepted classes. In this case, the previous set of independent components is the desired result. In contrast to the traditional methodology based on the Pearson criterion, the proposed approach allows us to bypass a problem of the decomposition of the range of values of random variables into multidimensional intervals. The method of generating a set of independent components of a multidimensional random variable is illustrated by the results of the analysis of spectral features of remote sensing data of forest tracts using space imagery from the Landsat-8 satellite.

“Human elephant conflict in changing land-use land-cover scenario in and adjoining region of Buxa tiger reserve, India”

The land use land cover changes (LULCC) in and around wildlife habitat are one of the major driving forces behind frequent human elephant conflict (HEC). The present study aims to determine the spatiotemporal LULC changes and its impact on Human – Elephant conflict in and adjoining Buxa Tiger Reserve (BTR). Landsat TM and OLI images of 1990, 2000, 2010, and 2019 were used and supervised classification using maximum likelihood classification algorithm applied to classify the LULC attributes. Each sign of elephant presence was recorded using GPS used for corridor mapping. The recorded sign of conflict was collected, used for settlement based human elephant conflict probability zone identification mapping by areal interpolation method. We found that (1990 - 2019) dense forest, surface water bodies, scrub land and agricultural fallow land decreased by 184.04 sq km, 21.94 sq km, 142.48 sq km, and 130.21 sq km respectively. On the other hand crop land, sparse vegetation, tea garden, open forest, degraded forest and built up areas increased by 119.74 sq km, 90.96 sq km, 87.05 sq km, 57.02 sq km, 19.07 sq km and 32.21 sq km respectively. The visual overlay method was performed between LULC map, and settlement based HEC probability map, therefore impact of spatiotemporal LULC change on HEC was determined. It finds out moderate to very high HE conflict settlement zone lies where nearby forest tract either degraded or large areas of open forest, scrub land replaced into crop fields and settlements. In addition, elephant sign along the broad gauge rail line were collected on foot, recorded using GPS used to prepared accident prone susceptibility map by Inverse Distance Weighting (IDW) interpolation technique to estimate where accidents were more likely to occur due to the concentration of elephant activity. It finds out high to very high accident prone susceptible zone lies in between Rajabhatkhawa station (st) – Damanpur st, Gorapara st- Rajabhatkhawa st, and railway line within Hamiltonganj range. The study wants to draw attention to the concerned authority and policy makers on that there is a need of long term land use planning to save BTR forest and its biodiversity as well as need for proper mitigate method for HE conflict for peaceful coexistences of both the species.

RESTORATION OF FORESTS IN THE TERRITORIES PASSED BY LARGE-FOREST FOREST FIRE IN THE CONDITIONS OF THE STATE ENTERPRISE «OVRUTSKE FORESTRY»

RESTORATION OF FORESTS IN THE TERRITORIES PASSED BY LARGE-FOREST FOREST FIRE IN THE CONDITIONS OF THE STATE ENTERPRISE «OVRUTSKE FORESTRY»

Floristic core of the Russian Plain pine forests along Pinus sylvestris southern borders

Pine forests - forest tracts on the sandy left banks, where the main area is occupied by plant communities with a predominance of Pinus sylvestris. Due to the significant distance from each other, the floras of the pine forests differ. However, the flora of these pine forests also shares common characteristics. In order to establish the floristic unity of the forests at the southern limit of the distribution of Pinus sylvestris, work was carried out to identify their floristic core. For 20 years, floristic surveys have been carried out in the forests of the steppe and forest-steppe zones. A general floristic list was compiled, with the identification of aboriginal and alien fractions, and the floristic core was established. Revealed: the core of the flora is determined by the pine forest flora with the lowest number of species among the comparable pine forests. The core of the flora of the studied pine forests included 30.8% of the species of the general flora. An increase in the number of analyzed floras will bring accuracy to the determination of the floristic core. This will make it possible to form an understanding of the pine forest flora as a single floristic complex.

Tree diversity in the Borra Sacred Groves of Anathagiri in Visakhapatnam District, Andhra Pradesh, India

Sacred groves are tracts of virgin forest with rich biodiversity, as they have been protected for centuries by the local people for their cultural and religious beliefs and taboos. Sacred groves are representatives of climax vegetation and exhibit diversity of species such as trees, climbers and other shade loving herbs. In the present investigation, an attempt was made to study the tree diversity of a sacred grove of Borra, Visakhapatnam District, which comes under the Eastern Ghats of India. A total of 31 species, 28 genera and 20 families were reported in the sacred grove.

What determines the distribution of a threatened species, the brush‐tailed phascogale Phascogale tapoatafa (Marsupialia: Dasyuridae), in a highly modified region?

AbstractIdentifying factors that affect species’ distributions in highly modified landscapes, especially threatened species, is a key issue for conservation worldwide. Three types of factors commonly are considered important: the extent and pattern of suitable habitat at a landscape scale; the availability of key resources at a local scale; and interactions with other species. The brush‐tailed phascogale (Phascogale tapoatafa) has undergone range contractions and regional extinctions in Australia and falls within the critical weight range of mammal species with high risk of decline. Yet, in some regions, it persists despite marked habitat loss and disturbance. We investigated the distribution of the brush‐tailed phascogale in a highly modified region in central Victoria where it is a ‘flagship’ for community conservation. At 50 sites, stratified geographically and in relation to topography and surrounding forest cover, we carried out surveys using remotely sensed cameras (in parallel with nest box inspections). We used an occupancy modelling framework and information‐theoretic approach to analyse factors influencing detectability and occupancy. The brush‐tailed phascogale was widespread, recorded at 84% (42/50) of sites. The environmental space occupied by sites with camera records matched that of nest boxes containing established nests of this species. Detection probability was influenced by survey effort and site‐level measures of habitat quality (logs, litter, topographic position) likely reflecting population abundance. Dry forested slopes were favoured compared with gullies. The study area, with its dry forests of rough‐barked eucalypts, is a stronghold for this species. The broad distribution of the brush‐tailed phascogale in this highly modified region can be attributed to retention of ˜36% regional forest cover and a relatively high level of structural and functional connectivity. In such modified landscapes, complementary actions by land managers to sustain remaining large tracts of forest and also a mosaic of wooded vegetation across the rural landscape will help such species to persist.

Structure and tree species composition of forest fringe of a forest fragment in an oil palm plantation at Suai, Sarawak, Malaysian Borneo

Abstract. Jana CL, Jusoh I. 2021. Structure and Tree Species Composition of Forest Fringe of A Forest Fragment in an Oil Palm Plantation at Suai, Sarawak, Malaysian Borne. Biodiversitas 22: 3013-3019. Forest fragments in oil palm plantations are remnants of the original forest tract with various sizes and shapes after clearing forested lands for crops and roads. Fragmented forests usually created forest fringe or edge habitats that differ in light intensity, temperature, humidity, heavy rainfall, and strong winds from its core. The study was aimed to determine the structure and tree species composition of the forest fringe of a fragmented forest in an oil palm plantation Suai, Sarawak. This study established a total of 24 sampling plots within the forest fringe. Tree enumeration and species identification were conducted in all sampling plots to all trees with a diameter at breast height 5 cm and higher. A total of 59 families comprised of 274 species and 948 trees individuals were recorded. The aboveground biomass stand at 260 Mg ha?¹. Species Elateriospermum tapos was predominant, and family Dipterocarpaceace dominated the forest fringe. Based on the calculated indices, species diversity and richness were considered high, while the dispersion of individual trees was random but unevenly distributed. Natural regeneration in the forest fringe is sustainable. The forest structure and species composition are intact, and without significant disturbance, the whole forest fragment is self-sustain.

Early successional riparian vegetation is important for western Yellow‐billed Cuckoo nesting habitat

Western Yellow‐billed Cuckoo (Cuckoo; Coccyzus americanus) populations continue to decline in the western United States despite efforts to increase availability of riparian forest. Cuckoos have unique breeding habitat requirements such as large contiguous tracts of riparian forest (&gt;80 ha), large estimated home ranges (20–90 ha), and dense vertical structure around the nest. However, local habitat‐scale features may be missing in landscapes of predominantly mature riparian forest that may need to be specifically managed for nesting. We used historical nest data (n = 95) from the South Fork Kern River Valley, California, from 1985 to 1996 to identify important nest site features that may be missing in current riparian forests. We found that increased canopy cover and vertical structure at all levels in the canopy greatly increased the probability of Cuckoo nesting. With smaller estimated effect sizes, the probability of Cuckoo nesting increased with increasing willows and forbs and smaller mean tree dbh. Cuckoos selected plots with disproportionately high percent willow cover relative to availability plots regardless of whether sites had low or high percent willow available. Counts of fledged young were positively related to willow percentage. No vegetation variable influenced daily survival rate which was 0.991 (LCI = 0.980, UCI = 0.996). Overall 17‐day nest success was likely high (0.86, LCI = 0.71, UCI = 0.93). In the absence of natural processes that create early successional stage forest, specific management for early successional stage forest is needed to increase the probability of Cuckoo nesting and nest productivity.

Land use and social-ecological legacies of Rio de Janeiro's Atlantic urban forests: from charcoal production to novel ecosystems.

Historical ecology is an important tool in deciphering human–environment interactions imprinted on landscapes throughout time. However, gaps of knowledge still remain regarding the land use legacies hidden in the current Atlantic Forest landscape; and also regarding how this information can help management of the remaining forest cover. The social-ecological systems framework was applied to understand charcoal production in the urban forests of Rio de Janeiro, from the nineteenth to mid-twentieth century, and their current social-ecological legacies. Charcoal production carried out by former enslaved populations, allowed for rapid forest regeneration. Forest thinning instead of forest felling was carried out by small groups in these urban remnant forests, sparing large native trees and facilitating natural regeneration. Currently, more than one thousand former charcoal production sites are accounted for hidden underneath the forest cover. The forest landscape of today is a result of novel forest successional trajectories that recovered structural and functional attributes of the forest ecosystem. However, this came at the cost of social invisibility and marginalization of these populations. The management practices of charcoal production dispersed in the landscape is one of Rio de Janeiro's most important, albeit hidden, land use legacies. Currently, the forested landscape is comprised of regenerated forests, both structurally and functionally sound, though with significant changes in species composition including the introduction of exotic species throughout recent centuries. These urban forests are today a complex mosaic of novel ecosystems, with rich biocultural diversity, and together with managed lands and well conserved forest tracts, provide not only livelihood and sustenance for forest dwelling families, but also important ecosystem services for the entire population of Rio de Janeiro. We believe that these concepts and frameworks can offer practical solutions for urban forest management, taking into account the biocultural diversity of Rio de Janeiro, increasing awareness of sustainability and promoting food security.

Linking demographic rates to local environmental conditions: Empirical data to support climate adaptation strategies for Eleutherodactylus frogs

Conducting managed species translocations and establishing climate change refugia are adaptation strategies to cope with projected consequences of global warming, but successful implementation requires on-the-ground validation of demographic responses to transient climate conditions. Here we estimated the effect of nine abiotic and biotic factors on local occupancy and an index of abundance (few or chorus) for four amphibian species (Eleutherodactylus wightmanae, E. brittoni, E. antillensis, and E. coqui) in Puerto Rico, USA. We also assessed how the same factors influenced reproductive activity of E. coqui and how species responded to hurricane María (20 September 2017). As predicted, occupancy and abundance of E. wightmanae, E. brittoni and E. coqui were positively and strongly influenced by abiotic covariates (e.g., relative humidity) that characterize high elevation, mesic habitats. E. antillensis exhibited the opposite pattern, with highest probabilities (≥0.6) recorded at ≤300 m and with average relative humidity<75%. Biotic covariates (e.g., canopy cover) had a weak influence on both parameters, regardless of species. High probabilities (≥0.9) of detecting an E. coqui chorus and active nests occurred at sites experiencing average relative humidity of>80% and temperature of ≤26 °C. Moderate to high probabilities of detecting a chorus (0.4–0.7) were recorded at sites with average temperatures>26 °C, but no reproductive activity was detected, implying that monitoring abundance alone could misrepresent the capacity of a local population to sustain itself. The possibility underscores the importance of understanding the interplay between local demographic and environmental parameters in the advent of global warming to help guide monitoring and management decisions, especially for high elevation specialists. Hurricanes can inflict marked reductions in population numbers, but impacts vary by location and species. We found that the abundance (chorus) of E. antillensis and E. brittoni increased after the hurricane, but the abundance of the other two species did not differ between years. Lack of impacts was probably mediated by low structural damage to forest tracts (e.g., 9% canopy loss). Our findings help assess habitat suitability in terms of parameters that foster local population growth, which provides a basis for testing spatio-temporal predictions about demographic rates in potential climate refugia and for designing criteria to help guide managed translocations.

The peri-urban leopards of Kathmandu: assessing determinants of presence and predation on domestic animals

AbstractThe conservation of large carnivores in human-dominated landscapes needs to be reconciled with the safety of humans and domestic animals. This is especially true for the leopard Panthera pardus, which occurs extensively in agricultural landscapes and remnant forest tracts embedded within peri-urban areas such as Kathmandu district in Nepal. We carried out interviews in 321 households in this district to determine the extent of leopard habitat use and predation on domestic animals (dogs and goats) during October 2015–April 2016. We used multi-state occupancy models, and estimated probabilities of leopard habitat use (Ψ1) and predation on domestic animals (Ψ2) as a function of covariates, while accounting for imperfect detection. Our findings indicate that the rapidly urbanizing outskirts of Kathmandu city are used extensively by leopards. The estimated probability of fine-scale habitat use in 2 km2 sample units was 0.96 ± SE 0.05 and the probability of predation on domestic animals was 0.76 ± SE 0.15. Leopard attacks occurred in areas with high vegetation cover and abundant goats. Addressing the problem of leopard attacks on domestic animals will require developing a comprehensive mitigation plan that includes educational activities to raise awareness, measures to address grievances of affected local communities, interventions to prevent attacks on livestock, compensation programmes, and rapid response teams to ensure human and animal welfare in conflict-prone areas. Land-use planning in these peri-urban landscapes needs to facilitate the safe sharing of space between people and leopards.

Mitigating the risks of indirect land use change (ILUC) related deforestation from industrial palm oil expansion by sharing land access with displaced crop and cattle farmers

The European Commission (EC) is critical of palm oil production as it has a high indirect land-use change (ILUC) risk from industrial palm oil expansion pressuring landless/displaced farmers to clear tropical forests. Major palm oil-producing countries such as Indonesia and Malaysia are claiming that the EC’s decision will cause unwarranted obstacles to the adoption of sustainable practices in the palm oil industry and obstruct free trade. To date, in responding to the EC, both Indonesia and Malaysia have yet to develop a national strategy that mitigates the risks of ILUC caused by industrial palm oil expansion. Hence, we propose these major palm oil-producing countries reform their land policy by making it compulsory for plantation companies to share production land with landless/displaced farmers. This strategy could prevent farmers from clearing more tracts of tropical forest and encourage them to use immature and mature areas of palm oil stands for crop and cattle farming, respectively. To guide this strategy, we performed Monte Carlo simulations to predict palm oil production land area and then estimate the land which could be allocated for crop and cattle farming at global-, country-, and plantation-scales. Besides mitigating further deforestation in the tropics, the crop and livestock integration strategy can boost food security and reduce the use of chemical herbicides in palm oil-producing countries. The strategy is consistent with palm oil certification schemes in terms of ensuring social and environmental sustainability. Palm oil-producing countries should implement the strategy to address ILUC risk while consumer countries should demand producers use such strategy.

Intact forests of the Hkakabo Razi Landscape are a hotspot of bat diversity in South-east Asia

AbstractThe Hkakabo Razi Landscape, in northern Kachin, Myanmar, is one of the largest remaining tracts of intact forest in South-east Asia. In 2016, we undertook a survey in its southern margins to assess bat diversity, distribution and ecology and evaluate the importance of the area for global bat conservation. Two collecting trips had taken place in the area in 1931 and 1933, with four bat species reported. We recorded 35 species, 18 of which are new for Kachin. One species, Murina hkakaboraziensis, was new to science and three, Megaerops niphanae, Phoniscus jagorii, Murina pluvialis, were new records for Myanmar. Our findings indicate high bat diversity in Hkakabo Razi; although it comprises only 1.7% of Myanmar's land area, it is home to 33.6% of its known bat species. This emphasizes Hkakabo Razi's importance for conserving increasingly threatened, forest-interior bats, especially in the families Kerivoulinae and Murininae. There is also a high diversity of other mammals and birds within the Hkakabo Razi Landscape, which supports its nomination as a World Heritage Site.

Occurrence and conservation of the Vulnerable titi monkey Callicebus melanochir in fragmented landscapes of the Atlantic Forest hotspot

AbstractTropical forest hotspots have a high diversity of species but have lost &gt; 70% of their original vegetation cover and are characterized by a multitude of small and isolated fragments. Paradoxically, conservation actions in these areas are still mainly focused on protection of large tracts of forests, a strategy now infeasible because of the small area of forest remnants. Here we use the Vulnerable black-handed titi monkey Callicebus melanochir as a model to study the effects of habitat loss, fragmentation and degradation on arboreal mammals and to provide insights for science-driven conservation in fragmented landscapes in tropical forest hotspots. We surveyed 38 Atlantic Forest fragments in Bahia State, Brazil and assessed the effects of patch area, quality and visibility, and landscape connectivity on the occurrence of our model species. Patch area was the single best model explaining species occurrence. Nonetheless, patch quality and visibility, and landscape connectivity, positively affect occurrence. In addition to patch area, patch quality, patch visibility and landscape connectivity are useful for predicting the occurrence of arboreal mammals in the fragments of tropical forest hotspots. We encourage the assessment of habitat quality (based on remotely sensed vegetation indices) and habitat visibility (based on digital elevation models) to improve discoverability of arboreal mammal populations and selection of fragments for conservation purposes across fragmented landscapes of tropical forest hotspots. Large remnants of tropical forest hotspots are scarce and therefore we require baseline data to support conservation actions and management in small forest fragments.

Accurate mapping of Brazil nut trees (Bertholletia excelsa) in Amazonian forests using WorldView-3 satellite images and convolutional neural networks

The commercialization of Brazil nuts, seeds of Bertholletia excelsa Bonpl. (Lecythidaceae), represents one of the main income-generation activities for local and indigenous people from the Brazilian Amazon region. Because trees of B. excelsa grow and bear fruit almost exclusively in natural forests, information on their spatial distribution is crucial for nut harvest planning. However, this information is difficult to obtain with traditional approaches such as ground-based surveys. Here, we show the potential of convolutional neural networks (CNNs) and WorldView-3 satellite images (pixel size = 30 cm) to map individual tree crowns (ITCs) and groves of B. excelsa in Amazonian forests. First, we manually outlined B. excelsa ITCs in the WorldView-3 images using field-acquired geolocation information. Then, based on ITC boundaries, we sequentially extracted image patches and selected 80% of them for training and 20% for testing. We trained the DeepLabv3+ architecture with three backbones: ResNet-18, ResNet-50, and MobileNetV2. The average producer's accuracy was 93.87 ± 0.85%, 93.89 ± 1.6% and 93.47 ± 3.6% for ResNet-18, ResNet-50 and MobileNetV2, respectively. We then developed a new random patch extraction training strategy and assessed how a reduction in the percentage of training patches impacted the classification accuracy. To illustrate the robustness of the new training strategy, similar F1-scores were achieved whether 80% or 10% of the total number of patches were used to train the CNN model. By analyzing the feature maps derived from ResNet-18, we found that the shadow of emergent B. excelsa trees are important for their discrimination. Geometric distortions in the WorldView-3 images resulting from extreme off-nadir viewing angles compromise the presence of shadows, thus potentially hampering B. excelsa detection. Our results show that ITCs and groves of B. excelsa can be mapped by integrating CNNs and very-high-resolution (VHR) satellite images, paving the way for monitoring this important tree species in large tracts of Amazonian forests.