skip to main content


This content will become publicly available on April 18, 2025

Title: PalmProbNet: A Probabilistic Approach to Understanding Palm Distributions in Ecuadorian Tropical Forest via Transfer Learning
Palms play an outsized role in tropical forests and are important resources for humans and wildlife. A central question in tropical ecosystems is understanding palm distribution and abundance. However, accurately identifying and localizing palms in geospatial imagery presents significant challenges due to dense vegetation, overlapping canopies, and variable lighting conditions in mixed-forest landscapes. Addressing this, we introduce PalmProbNet, a probabilistic approach utilizing transfer learning to analyze high-resolution UAV-derived orthomosaic imagery, enabling the detection of palm trees within the dense canopy of the Ecuadorian Rainforest. This approach represents a substantial advancement in automated palm detection, effectively pinpointing palm presence and locality in mixed tropical rainforests. Our process begins by generating an orthomosaic image from UAV images, from which we extract and label palm and non-palm image patches in two distinct sizes. These patches are then used to train models with an identical architecture, consisting of an unaltered pre-trained ResNet-18 and a Multilayer Perceptron (MLP) with specifically trained parameters. Subsequently, PalmProbNet employs a sliding window technique on the landscape orthomosaic, using both small and large window sizes to generate a probability heatmap. This heatmap effectively visualizes the distribution of palms, showcasing the scalability and adaptability of our approach in various forest densities. Despite the challenging terrain, our method demonstrated remarkable performance, achieving an accuracy of 97.32% and a Cohen's κ of 94.59% in testing.  more » « less
Award ID(s):
2039850
NSF-PAR ID:
10503319
Author(s) / Creator(s):
; ; ; ; ; ; ; ; ; ;
Publisher / Repository:
ACM
Date Published:
Journal Name:
ACM SE '24: Proceedings of the 2024 ACM Southeast Conference
Page Range / eLocation ID:
272 to 277
Format(s):
Medium: X
Location:
Marietta GA USA
Sponsoring Org:
National Science Foundation
More Like this
  1. This study aims to understand the spatiotemporal changes in patterns of tropical crop cultivation in Eastern Thailand, encompassing the periods before, during, and after the COVID-19 pandemic. Our approach involved assessing the efficacy of high-resolution (10 m) Sentinel-2 dense image time series for mapping smallholder farmlands. We integrated harmonic regression and random forest to map a diverse array of tropical crop types between summer 2017 and summer 2023, including durian, rice, rubber, eucalyptus, oil palm, pineapple, sugarcane, cassava, mangosteen, coconut, and other crops. The results revealed an overall mapping accuracy of 85.6%, with several crop types exceeding 90%. High-resolution imagery demonstrated particular effectiveness in situations involving intercropping, a popular practice of simultaneously growing two or more plant species in the same patch of land. However, we observed overestimation in the majority of the studied cash crops, primarily those located in young plantations with open tree canopies and grass-covered ground surfaces. The adverse effects of the COVID-19 pandemic were observed in specific labor-intensive crops, including rubber and durian, but were limited to the short term. No discernible impact was noted across the entirety of the study timeframe. In comparison, financial gain and climate change appeared to be more pivotal in influencing farmers’ decisions regarding crop cultivation. Traditionally dominant crops such as rice and oil palm have witnessed a discernible decline in cultivation, reflecting a decade-long trend of price drops preceding the pandemic. Conversely, Thai durian has seen a significant upswing even over the pandemic, which ironically served as a catalyst prompting Thai farmers to adopt e-commerce to meet the surging demand, particularly from China.

     
    more » « less
  2. Abstract

    Through frugivory and seed dispersal, vertebrates influence plant demography and forest regeneration. Variation in local habitat surrounding fruiting plants can influence frugivore foraging decisions, thereby creating intraspecific variation in seed dispersal services. However, we have little knowledge of drivers of local variation in frugivory. Here, we investigate factors that may influence frugivore diversity and fruit removal at the level of individual plants. We focus on a common understory palm within a continuous Chocó forest with mixed land‐use histories in Ecuador. The density of pioneer tree species in the genusCecropiaaround focal palms was negatively related to fruit removal and the diversity of frugivores visiting palms. This may relate to the fact that the presence and abundance ofCecropiaspecies often indicate the existence and severity of past disturbances. LocalCecropiadensity was also related to an overall shift in the frugivore community that corresponded with an increase in fruit removal by lower‐quality seed dispersers (rodents). We also found that the local density of fruiting conspecifics was positively related to frugivore diversity, but not fruit removal. Our results provide information on drivers of intraspecific inequalities in plant populations across tropical forest landscapes. The reduction in fruit removal and frugivore diversity associated with localCecropiaabundance suggests that seed dispersal services can be sensitive to fine‐scale variation in habitat structure. Furthermore, becauseCecropiaare often indicative of past disturbances, this indicates that even small‐scale habitat degradation by humans can have lasting effects by creating localized pockets of forest unfavored by frugivores.

    Abstract in Spanish is available with online material.

     
    more » « less
  3. Abstract

    Information about the spatial distribution of species lies at the heart of many important questions in ecology. Logistical limitations and collection biases, however, limit the availability of such data at ecologically relevant scales. Remotely sensed information can alleviate some of these concerns, but presents challenges associated with accurate species identification and limited availability of field data for validation, especially in high diversity ecosystems such as tropical forests.

    Recent advances in machine learning offer a promising and cost‐efficient approach for gathering a large amount of species distribution data from aerial photographs. Here, we propose a novel machine learning framework, artificial perceptual learning (APL), to tackle the problem of weakly supervised pixel‐level mapping of tree species in forests. Challenges arise from limited availability of ground labels for tree species, lack of precise segmentation of tree canopies and misalignment between visible canopies in the aerial images and stem locations associated with ground labels. The proposed APL framework addresses these challenges by constructing a workflow using state‐of‐the‐art machine learning algorithms.

    We develop and illustrate the proposed framework by implementing a fine‐grain mapping of three species, the palmPrestoea acuminataand the tree speciesCecropia schreberianaandManilkara bidentata, over a 5,000‐ha area of El Yunque National Forest in Puerto Rico. These large‐scale maps are based on unlabelled high‐resolution aerial images of unsegmented tree canopies. Misaligned ground‐based labels, available for <1% of these images, serve as the only weak supervision. APL performance is evaluated using ground‐based labels and high‐quality human segmentation using Amazon Mechanical Turk, and compared to a basic workflow that relies solely on labelled images.

    Receiver operating characteristic (ROC) curves and Intersection over Union (IoU) metrics demonstrate that APL substantially outperforms the basic workflow and attains human‐level cognitive economy, with 50‐fold time savings. For the palm andC. schreberiana, the APL framework has high pixelwise accuracy and IoU with reference to human segmentations. ForM.bidentata, APL predictions are congruent with ground‐based labels. Our approach shows great potential for leveraging existing data from global forest plot networks coupled with aerial imagery to map tree species at ecologically meaningful spatial scales.

     
    more » « less
  4. Abstract

    Phytoliths preserved in soils and sediments can be used to provide unique insights into past vegetation dynamics in response to human and climate change. Phytoliths can reconstruct local vegetation in terrestrial soils where pollen grains typically decay, providing a range of markers (or lack thereof) that document past human activities. The ca. 6 million km2of Amazonian forests have relatively few baseline datasets documenting changes in phytolith representation across gradients of human disturbances. Here we show that phytolith assemblages vary on local scales across a gradient of (modern) human disturbance in tropical rainforests of Suriname. Detrended correspondence analysis showed that the phytolith assemblages found in managed landscapes (shifting cultivation and a garden), unmanaged forests, and abandoned reforesting sites were clearly distinguishable from intact forests and from each other. Our results highlight the sensitivity and potential of phytoliths to be used in reconstructing successional trajectories after site usage and abandonment. Percentages of specific phytolith morphotypes were also positively correlated with local palm abundances derived from UAV data, and with biomass estimated from MODIS satellite imagery. This baseline dataset provides an index of likely changes that can be observed at other sites that indicate past human activities and long-term forest recovery in Amazonia.

     
    more » « less
  5. Cloud detection is an inextricable pre-processing step in remote sensing image analysis workflows. Most of the traditional rule-based and machine-learning-based algorithms utilize low-level features of the clouds and classify individual cloud pixels based on their spectral signatures. Cloud detection using such approaches can be challenging due to a multitude of factors including harsh lighting conditions, the presence of thin clouds, the context of surrounding pixels, and complex spatial patterns. In recent studies, deep convolutional neural networks (CNNs) have shown outstanding results in the computer vision domain. These methods are practiced for better capturing the texture, shape as well as context of images. In this study, we propose a deep learning CNN approach to detect cloud pixels from medium-resolution satellite imagery. The proposed CNN accounts for both the low-level features, such as color and texture information as well as high-level features extracted from successive convolutions of the input image. We prepared a cloud-pixel dataset of approximately 7273 randomly sampled 320 by 320 pixels image patches taken from a total of 121 Landsat-8 (30m) and Sentinel-2 (20m) image scenes. These satellite images come with cloud masks. From the available data channels, only blue, green, red, and NIR bands are fed into the model. The CNN model was trained on 5300 image patches and validated on 1973 independent image patches. As the final output from our model, we extract a binary mask of cloud pixels and non-cloud pixels. The results are benchmarked against established cloud detection methods using standard accuracy metrics.

     
    more » « less