Search Results (35)

Hurricanes are major drivers of forest structure in the Caribbean. In 2017, Hurricane Maria caused substantial damage to Puerto Rico’s forests. We studied forest structure variation across 75 sites at Las Casas de la Selva, a sustainable forest plantation in Patillas, Puerto Rico, seven years after Hurricane Maria hit the property. At each site we analyzed 360° photos in a 3D VR headset to quantify the vertical structure and transformed them into hemispherical images to quantify canopy closure and ground cover. We also computed the Vertical Habitat Diversity Index (VHDI) from the amount of foliage in four strata: herbaceous, shrub, understory, and canopy. Using the Local Bivariate Relationship tool in ArcGIS Pro, we analyzed the relationship between forest recovery (vertical structure, canopy closure, and ground cover) and damage. Likewise, we analyzed the effects of elevation, slope, and aspect, on damage, canopy closure, and vertical forest structure. We found that canopy closure decreases with increasing elevation and increases with the amount of damage. Higher elevations show a greater amount of damage even seven years post hurricane. We conclude that trees in the mixed tabonuco/plantation forest are more susceptible to hurricanes at higher elevations. The results have implications for plantation forest management under climate-change-driven higher intensity hurricane regimes. Full article

Soil temperature and soil moisture are significant interactive parameters that influence many ecological and hydrological processes within forest ecosystems. Furthermore, they are affected by the above canopy characteristics, which determine the amount of sunlight penetration. These canopy characteristics spatially vary within isolated or narrowed forest patches, which include interior parts and edges. On the other hand, forest patches display different effects on the soil temperature and moisture than agricultural meadows within rural landscapes. This study aimed to analyze and evaluate the influences of interior–edge canopies and meadow cover on soil temperature and moisture. Hence, the mutual responses of canopy phenology and physiology, along with the soil temperature and moisture beneath, were analyzed and determined on a temporal basis throughout one year. For this purpose, the air–soil temperature and precipitation data of close meteorological stations were utilized. In addition, soil temperature and moisture parameters were analyzed using an on-site measuring device. Furthermore, canopy parameters—namely LAI, LT, CO, and GF—were determined using a hemispherical photographing procedure and image processing–analysis methodology. Moreover, the LAI of the meadow cover was determined using an on-site analysis device. The maximum LAI, with mean values of 3.69 m² m⁻² and 2.54 m² m⁻², occurred in late May (DOY: 142) within the forest canopies of the interior parts and the patch edge, respectively. On the other hand, the maximum LAI with a mean value of 2.77 m² m⁻² occurred again in late May within the meadow field. On the contrary, during the same period, the lowest percentages were observed for LT and CO, each at 5%, and for GF with 0.5% within the interior parts of the forest patch. However, their lowest percentages were 23% and 16%, respectively, within the forest patch edge. For that late May period, the mean soil temperatures were 17.2, 26.0, and 21.0 °C under the forest canopies of the interior parts, the patch edge, and the meadow field, respectively. Meanwhile, their mean soil moistures were 56.4%, 51.6%, and 32.9% when the mean air temperature was 16.2 °C. Definite correlation did not exist between the canopy–vegetation parameters and the soil temperature–moisture values for all the interior parts, for the edge of the multi-story grove forest patch, and for the meadow field. Based on the overall results of this study, there were apparent differences amongst the interior parts, the edge of the forest patch, and the meadow field in terms of both the canopy–vegetation parameters and the soil temperature–moisture values. The multi-story structure of the interior parts and the edge of the forest patch determined the temporal patterns of their canopy–vegetation parameters. This study elucidated ecology, hydrology, and therefore management of narrow forest patches between agricultural areas within rural landscapes. Full article

Leaf Area Index (LAI) is one of the key parameters for characterizing leaf density, vegetation growth status, and canopy structure. Rapid, objective, and accurate acquisition of forest LAI is of great significance for studying forest ecosystems and forestry production. This study focuses on the core issue of accurately segmenting leaf elements from background elements in hemispherical photography used for forest LAI measurement, with a particular focus on meeting the real-time requirements of embedded platforms. The differences in grayscale values and frequency characteristics between leaf regions, trunk regions, and sky regions in vegetation canopy images were leveraged to decompose, process, and reconstruct such images using a 9/7 wavelet-based transformation method, achieving efficient and precise segmentation of leaf regions. Effectively addresses the issue of LAI overestimation caused by trunk regions in traditional threshold-based segmentation methods. Through the extraction of canopy gap fraction, rapid LAI measurement was enabled. Comparative experimental results showed that the proposed inversion method exhibited a high correlation with the LAI-2200C measurement results (r = 0.897, RMSE = 0.431), fully verifying its accuracy across different forest ecological environments. This study provides strong support for the development of portable, high-precision LAI measurement devices and holds practical application value and broad application prospects. Full article

Understanding the growth dynamics of spore-bearing clonal plant sporophytes and the influence of abiotic and biotic factors is crucial for predicting the persistence of club moss populations and implementing effective habitat management techniques. Despite this, the longevity and development of club-moss populations are rarely studied. This study adopted an integrated approach to assess the probability of repetitive young sporophyte recruitment via sexual propagation in Lycopodium annotinum L. and Lycopodium clavatum L. The size–age problem of clonal spore-bearing forest plants and their niche segregation were addressed. The canopy characteristics, insolation, small-scale disturbance, and genetic polymorphism were studied in temperate semi-natural Scots pine forests in Lithuania. Based on the size of the clones discovered, we hypothesize that initial sporophyte emergence occurred in 20-year-old pine stands, with subsequent sporophyte emergence continuing over time. The emergence was related to small-scale disturbances. High genetic polymorphism indicates that all sporophyte stands studied likely emerged via sexual reproduction. According to Ellenberg values, L. annotinum is related to shady habitats, but our findings show both species coexisting abundantly in the more open habitat, supposedly more suitable for L. clavatum.No significant differences in vegetation relevés and light availability was detected using hemispheric images. Full article

Leaf Area Index (LAI) is a fundamental indicator of plant growth status in agronomy and environmental research. With the rapid development of drone technology, the estimation of crop LAI based on drone imagery and vegetation indices is becoming increasingly popular. However, there is still a lack of detailed research on the feasibility of using image texture to estimate LAI and the impact of combining texture indices with vegetation indices on LAI estimation accuracy. In this study, two key growth stages of winter wheat (i.e., the stages of green-up and jointing) were selected, and LAI was calculated using digital hemispherical photography. The feasibility of predicting winter wheat LAI was explored under three conditions: vegetation index, texture index, and a combination of vegetation index and texture index, at flight heights of 20 m and 40 m. Two feature selection methods (Lasso and recursive feature elimination) were combined with four machine learning regression models (multiple linear regression, random forest, support vector machine, and backpropagation neural network). The results showed that during the vegetative growth stage of winter wheat, the model combining texture information with vegetation indices performed better than the models using vegetation indices alone or texture information alone. Among them, the best prediction result based on vegetation index was RFECV-MLR at a flight height of 40 m (R² = 0.8943, RMSE = 0.4139, RRMSE = 0.1304, RPD = 3.0763); the best prediction result based on texture index was RFECV-RF at a flight height of 40 m (R² = 0.8894, RMSE = 0.4236, RRMSE = 0.1335, RPD = 3.0063); and the best prediction result combining texture and index was RFECV-RF at a flight height of 40 m (R² = 0.9210, RMSE = 0.3579, RRMSE = 0.1128, RPD = 3.5575). The results of this study demonstrate that combining vegetation indices and texture from multispectral drone imagery can improve the accuracy of LAI estimation during the vegetative growth stage of winter wheat. In addition, selecting a flight height of 40 m can improve efficiency in large-scale agricultural field monitoring, as this study showed that drone data at flight heights of 20 m and 40 m did not significantly affect model accuracy. Full article

Quantification of vegetation biophysical variables such as leaf area index (LAI), fractional vegetation cover (fCover), and biomass are among the key factors across hydrological, agricultural, and irrigation management studies. The present study proposes a kernel-based machine learning algorithm capable of performing adaptive and nonlinear data fitting so as to generate a suitable, accurate, and robust algorithm for spatio-temporal estimation of the three mentioned variables using Sentinel-2 images. To this aim, Gaussian process regression (GPR)–particle swarm optimization (PSO), GPR–genetic algorithm (GA), GPR–tabu search (TS), and GPR–simulated annealing (SA) hyperparameter-optimized algorithms were developed and compared against kernel-based machine learning regression algorithms and artificial neural network (ANN) and random forest (RF) algorithms. The accuracy of the proposed algorithms was assessed using digital hemispherical photography (DHP) data and destructive measurements performed during the growing season of silage maize in agricultural fields of Ghale-Nou, southern Tehran, Iran, in the summer of 2019. The results on biophysical variables against validation data showed that the developed GPR-PSO algorithm outperformed other algorithms under study in terms of robustness and accuracy (0.917, 0.931, 0.882 using R² and 0.627, 0.078, and 1.99 using RMSE in LAI, fCover, and biomass of Sentinel-2 20 m, respectively). GPR-PSO also possesses the unique ability to generate pixel-based uncertainty maps (confidence level) for prediction purposes (i.e., estimated uncertainty level <0.7 in LAI, fCover, and biomass, for 96%, 98%, and 71% of the total study area, respectively). Altogether, GPR-PSO appears to be the most suitable option for mapping biophysical variables at the local scale using Sentinel-2 images. Full article

Accurately diagnosing schizophrenia, a complex psychiatric disorder, is crucial for effectively managing the treatment process and methods. Various types of magnetic resonance (MR) images have the potential to serve as biomarkers for schizophrenia. The aim of this study is to numerically analyze differences in the textural characteristics that may occur in the bilateral amygdala, caudate, pallidum, putamen, and thalamus regions of the brain between individuals with schizophrenia and healthy controls via structural MR images. Towards this aim, Gray Level Co-occurence Matrix (GLCM) features obtained from five regions of the right, left, and bilateral brain were classified using machine learning methods. In addition, it was analyzed in which hemisphere these features were more distinctive and which method among Adaboost, Gradient Boost, eXtreme Gradient Boosting, Random Forest, k-Nearest Neighbors, Linear Discriminant Analysis (LDA), and Naive Bayes had higher classification success. When the results were examined, it was demonstrated that the GLCM features of these five regions in the left hemisphere could be classified as having higher performance in schizophrenia compared to healthy individuals. Using the LDA algorithm, classification success was achieved with a 100% AUC, 94.4% accuracy, 92.31% sensitivity, 100% specificity, and an F1 score of 91.9% in healthy and schizophrenic individuals. Thus, it has been revealed that the textural characteristics of the five predetermined regions, instead of the whole brain, are an important indicator in identifying schizophrenia. Full article

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease whose diagnosis depends on the presence of combined lower motor neuron (LMN) and upper motor neuron (UMN) degeneration. LMN degeneration assessment is aided by electromyography, whereas no equivalent exists to assess UMN dysfunction. Magnetic resonance imaging (MRI) is primarily used to exclude conditions that mimic ALS. We have identified four different clinical/radiological phenotypes of ALS patients. We hypothesize that these ALS phenotypes arise from distinct pathologic processes that result in unique MRI signatures. To our knowledge, no machine learning (ML)-based data analyses have been performed to stratify different ALS phenotypes using MRI measures. During routine clinical evaluation, we obtained T1-, T2-, PD-weighted, diffusion tensor (DT) brain MRI of 15 neurological controls and 91 ALS patients (UMN-predominant ALS with corticospinal tract CST) hyperintensity, n = 21; UMN-predominant ALS without CST hyperintensity, n = 26; classic ALS, n = 23; and ALS patients with frontotemporal dementia, n = 21). From these images, we obtained 101 white matter (WM) attributes (including DT measures, graph theory measures from DT and fractal dimension (FD) measures using T1-weighted), 10 grey matter (GM) attributes (including FD based measures from T1-weighted), and 10 non-imaging attributes (2 demographic and 8 clinical measures of ALS). We employed classification and regression tree, Random Forest (RF) and also artificial neural network for the classifications. RF algorithm provided the best accuracy (70–94%) in classifying four different phenotypes of ALS patients. WM metrics played a dominant role in classifying different phenotypes when compared to GM or clinical measures. Although WM measures from both right and left hemispheres need to be considered to identify ALS phenotypes, they appear to be differentially affected by the degenerative process. Longitudinal studies can confirm and extend our findings. Full article

Wildfires play a critical role in re-shaping boreal ecosystems and climate. It was projected that, owing to the Arctic amplification, boreal wildfires would become more frequent and severe in the coming decades. Although provoking concern, the spatiotemporal changes in boreal wildfires remain unclear, and there are substantial inconsistencies among previous findings. In this study, we performed a comprehensive analysis to determine the spatiotemporal changes in wildfires over Northern Eurasia (NEA) from 2003 to 2020 using a reconstructed Moderate Resolution Imaging Spectroradiometer (MODIS) active fire product. We found that wildfires in NEA exhibited contrasting changes in different latitudinal zones, land cover types, and seasons from 2003 to 2020. Cropland wildfires, mainly distributed at low latitudes (50–60°N), considerably decreased by 81% during the study period. Whereas forest wildfires ignited at high latitudes (north of 60°N) have nearly tripled (increasing at rate of 11~13% per year) during the past two decades. The southwestern and northeastern NEA regions exhibited contrasting patterns of wildfire changes. The active fire counts in the southwestern NEA decreased by 90% at a rate of 0.29(±0.12) × 10⁵ per year, with cropland fires contributing to ~66% of the decrease. However, the fire counts in the northeastern NEA increased by 292% at a rate of 0.23(±0.12) × 10⁵ per year, with boreal forests contributing to ~97% of the increase. It is worth noting that the contrasting changes in wildfires during the past two decades have led to significant structural alternation in the NEA wildfire composition. Forest fires, contributing over 60% of the total fire counts in NEA nowadays, have become the predominant component of the NEA wildfires. The contrasting changes in NEA wildfires imply that more forest fires may emerge in far northern regions of the North Hemisphere as the Arctic becomes progressively warmer in the coming decades. As wildfires continue to increase, more gases and aerosols would be released to the atmosphere and cause considerable feedback to the Arctic climate. The increased wildfire-related climate feedbacks should, therefore, be seriously considered in climate models and projections. Full article

Existing efforts to continuously monitor land cover change using satellite image time series have mostly focused on forested ecosystems in the tropics and the Northern Hemisphere. The notable difference in spectral reflectance that occurs following deforestation allows land cover change to be detected with relative accuracy. Less progress has been made in detecting change in low productivity or disturbance-prone vegetation such as grasslands and shrublands where natural dynamics can be difficult to distinguish from habitat loss. Renosterveld is a hyperdiverse, critically endangered shrubland ecosystem in South Africa with less than 5–10% of its original extent remaining in small, highly fragmented patches. I demonstrate that classification of satellite image time series using neural networks can accurately detect the transformation of Renosterveld within a few days of its occurrence and that trained models are suitable for operational continuous monitoring. A dataset of precisely dated vegetation change events between 2016 and 2021 was obtained from daily, high resolution Planet Labs satellite data. This dataset was then used to train 1D convolutional neural networks and Transformers to continuously detect land cover change events in time series of vegetation activity from Sentinel 2 satellite data. The best model correctly identified 89% of land cover change events at the pixel-level, achieving a f-score of 0.93, a 79% improvement over the f-score of 0.52 achieved using a method designed for forested ecosystems based on trend analysis. Models have been deployed to operational use and are producing updated detections of habitat loss every 10 days. There is great potential for continuous monitoring of habitat loss in non-forest ecosystems with complex natural dynamics. A key limiting step is the development of accurately dated datasets of land cover change events with which to train machine-learning classifiers. Full article

The accurate evaluation of shifts in vegetation phenology is essential for understanding of vegetation responses to climate change. Remote-sensing vegetation index (VI) products with multi-day scales have been widely used for phenology trend estimation. VI composites should be interpolated into a daily scale for extracting phenological metrics, which may not fully capture daily vegetation growth, and how this process affects phenology trend estimation remains unclear. In this study, we chose 120 sites over four vegetation types in the mid-high latitudes of the northern hemisphere, and then a Moderate Resolution Imaging Spectroradiometer (MODIS) MCD43A4 daily surface reflectance data was used to generate a daily normalized difference vegetation index (NDVI) dataset in addition to an 8-day and a 16-day NDVI composite datasets from 2001 to 2019. Five different time interpolation methods (piecewise logistic function, asymmetric Gaussian function, polynomial curve function, linear interpolation, and spline interpolation) and three phenology extraction methods were applied to extract data from the start of the growing season and the end of the growing season. We compared the trends estimated from daily NDVI data with those from NDVI composites among (1) different interpolation methods; (2) different vegetation types; and (3) different combinations of time interpolation methods and phenology extraction methods. We also analyzed the differences between the trends estimated from the 8-day and 16-day composite datasets. Our results indicated that none of the interpolation methods had significant effects on trend estimation over all sites, but the discrepancies caused by time interpolation could not be ignored. Among vegetation types with apparent seasonal changes such as deciduous broadleaf forest, time interpolation had significant effects on phenology trend estimation but almost had no significant effects among vegetation types with weak seasonal changes such as evergreen needleleaf forests. In addition, trends that were estimated based on the same interpolation method but different extraction methods were not consistent in showing significant (insignificant) differences, implying that the selection of extraction methods also affected trend estimation. Compared with other vegetation types, there were generally fewer discrepancies between trends estimated from the 8-day and 16-day dataset in evergreen needleleaf forest and open shrubland, which indicated that the dataset with a lower temporal resolution (16-day) can be applied. These findings could be conducive for analyzing the uncertainties of monitoring vegetation phenology changes. Full article

The scope of this work was to evaluate simulated carbon monoxide (CO) and aerosol optical depth (AOD) from the CAM-chem model against observed satellite data and additionally explore the empirical relationship of CO, AOD and fire radiative power (FRP). The simulated seasonal global concentrations of CO and AOD were compared, respectively, with the Measurements of Pollution in the Troposphere (MOPITT) and the Moderate-Resolution Imaging Spectroradiometer (MODIS) satellite products for the period 2010–2014. The CAM-chem simulations were performed with two configurations: (A) tropospheric-only; and (B) tropospheric with stratospheric chemistry. Our results show that the spatial and seasonal distributions of CO and AOD were reasonably reproduced in both model configurations, except over central China, central Africa and equatorial regions of the Atlantic and Western Pacific, where CO was overestimated by 10–50 ppb. In configuration B, the positive CO bias was significantly reduced due to the inclusion of dry deposition, which was not present in the model configuration A. There was greater CO loss due to the chemical reactions, and shorter lifetime of the species with stratospheric chemistry. In summary, the model has difficulty in capturing the exact location of the maxima of the seasonal AOD distributions in both configurations. The AOD was overestimated by 0.1 to 0.25 over desert regions of Africa, the Middle East and Asia in both configurations, but the positive bias was even higher in the version with added stratospheric chemistry. By contrast, the AOD was underestimated over regions associated with anthropogenic activity, such as eastern China and northern India. Concerning the correlations between CO, AOD and FRP, high CO is found during March–April–May (MAM) in the Northern Hemisphere, mainly in China. In the Southern Hemisphere, high CO, AOD, and FRP values were found during August–September–October (ASO) due to fires, mostly in South America and South Africa. In South America, high AOD levels were observed over subtropical Brazil, Paraguay and Bolivia. Sparsely urbanized regions showed higher correlations between CO and FRP (0.7–0.9), particularly in tropical areas, such as the western Amazon region. There was a high correlation between CO and aerosols from biomass burning at the transition between the forest and savanna environments over eastern and central Africa. It was also possible to observe the transport of these pollutants from the African continent to the Brazilian coast. High correlations between CO and AOD were found over southeastern Asian countries, and correlations between FRP and AOD (0.5–0.8) were found over higher latitude regions such as Canada and Siberia as well as in tropical areas. Higher correlations between CO and FRP are observed in Savanna and Tropical forests (South America, Central America, Africa, Australia, and Southeast Asia) than FRP x AOD. In contrast, boreal forests in Russia, particularly in Siberia, show a higher FRP x AOD correlation than FRP x CO. In tropical forests, CO production is likely favored over aerosol, while in temperate forests, aerosol production is more than CO compared to tropical forests. On the east coast of the United States, the eastern border of the USA with Canada, eastern China, on the border between China, Russia, and Mongolia, and the border between North India and China, there is a high correlation of CO x AOD and a low correlation between FRP with both CO and AOD. Therefore, such emissions in these regions are not generated by forest fires but by industries and vehicular emissions since these are densely populated regions. Full article

Satellite-borne passive microwave radiometers provide brightness temperature (TB) measurements in a large spectral range which includes a number of frequency channels and generally two polarizations: horizontal and vertical. These TBs are widely used to retrieve several atmospheric and surface variables and parameters such as precipitation, soil moisture, water vapor, air temperature profile, and land surface emissivity. Since TBs are measured at different microwave frequencies with various instruments and at various incidence angles, spatial resolutions, and radiometric characteristics, a mere direct integration of them from different microwave sensors would not necessarily provide consistency. However, when appropriately harmonized, they can provide a complete dataset to estimate the diurnal cycle. This study first constructs the diurnal cycle of land TBs using the non-sun-synchronous Global Precipitation Measurement (GPM) Microwave Imager (GMI) observations by utilizing a cubic spline fit. The acquisition times of GMI vary from day to day and, therefore, the shape (amplitude and phase) of the diurnal cycle for each month is obtained by merging several days of measurements. This diurnal pattern is used as a point of reference when intercalibrated TBs from other passive microwave sensors with daily fixed acquisition times (e.g., Special Sensor Microwave Imager/Sounder, and Advanced Microwave Scanning Radiometer 2) are used to modify and tune the monthly diurnal cycle to daily diurnal cycle at a global scale. Since the GMI does not cover polar regions, the proposed method estimates a consistent diurnal cycle of land TBs at global scale. Results show that the shape and peak of the constructed TB diurnal cycle is approximately similar to the diurnal cycle of land surface temperature. The diurnal brightness temperature range for different land cover types has also been explored using the derived diurnal cycle of TBs. In general, a large diurnal TB range of more than 15 K has been observed for the grassland, shrubland, and tundra land cover types, whereas it is less than 5K over forests. Furthermore, seasonal variations in the diurnal TB range for different land cover types show a more consistent result over the Southern Hemisphere than over the Northern Hemisphere. The calibrated TB diurnal cycle may then be used to consistently estimate the diurnal cycle of land surface emissivity. Moreover, since changes in land surface emissivity are related to moisture change and freeze–thaw (FT) transitions in high-latitude regions, the results of this study enhance temporal detection of FT state, particularly during the transition times when multiple FT changes may occur within a day. Full article

Research Highlights: This paper proposes a new method for hemispherical forest canopy image segmentation. The method is based on a deep learning methodology and provides a robust and fully automatic technique for the segmentation of forest canopy hemispherical photography (CHP) and gap fraction (GF) calculation. Background and Objectives: CHP is widely used to estimate structural forest variables. The GF is the most important parameter for calculating the leaf area index (LAI), and its calculation requires the binary segmentation result of the CHP. Materials and Methods: Our method consists of three modules, namely, northing correction, valid region extraction, and hemispherical image segmentation. In these steps, a core procedure is hemispherical canopy image segmentation based on the U-Net convolutional neural network. Our method is compared with traditional threshold methods (e.g., the Otsu and Ridler methods), a fuzzy clustering method (FCM), commercial professional software (WinSCANOPY), and the Habitat-Net network method. Results: The experimental results show that the method presented here achieves a Dice similarity coefficient (DSC) of 89.20% and an accuracy of 98.73%. Conclusions: The method presented here outperforms the Habitat-Net and WinSCANOPY methods, along with the FCM, and it is significantly better than the Otsu and Ridler threshold methods. The method takes the original canopy hemisphere image first and then automatically executes the three modules in sequence, and finally outputs the binary segmentation map. The method presented here is a pipelined, end-to-end method. Full article

A reliable estimation and monitoring of tree canopy cover or shade distribution is essential for a sustainable cocoa production via agroforestry systems. Remote sensing (RS) data offer great potential in retrieving and monitoring vegetation status at landscape scales. However, parallel advancements in image processing and analysis are required to appropriately use such data for different targeted applications. This study assessed the potential of Sentinel-1A (S-1A) C-band synthetic aperture radar (SAR) backscatter in estimating canopy cover variability in cocoa agroforestry landscapes. We investigated two landscapes, in Center and South Cameroon, which differ in predominant vegetation: forest-savannah transition and forest landscape, respectively. We estimated canopy cover using in-situ digital hemispherical photographs (DHPs) measures of gap fraction, verified the relationship with SAR backscatter intensity and assessed predictions based on three machine learning approaches: multivariate bootstrap regression, neural networks regression, and random forest regression. Our results showed that about 30% of the variance in canopy gap fraction in the cocoa production landscapes was shared by the used SAR backscatter parameters: a combination of S-1A backscatter intensity, backscatter coefficients, difference, cross ratios, and normalized ratios. Based on the model predictions, the VV (co-polarization) backscatter showed high importance in estimating canopy gap fraction; the VH (cross-polarized) backscatter was less sensitive to the estimated canopy gap. We observed that a combination of different backscatter variables was more reliable at predicting the canopy gap variability in the considered type of vegetation in this study—agroforests. Semi-variogram analysis of canopy gap fraction at the landscape scale revealed higher spatial clustering of canopy gap, based on spatial correlation, at a distance range of 18.95 m in the vegetation transition landscape, compared to a 51.12 m spatial correlation range in the forest landscape. We provide new insight on the spatial variability of canopy gaps in the cocoa landscapes which may be essential for predicting impacts of changing and extreme (drought) weather conditions on farm management and productivity. Our results contribute a proof-of-concept in using current and future SAR images to support management tools or strategies on tree inventorying and decisions regarding incentives for shade tree retention and planting in cocoa landscapes. Full article