Tree mortality in an agricultural landscape of Southwestern Panama assessed using remote sensing and field data
Data files
Apr 03, 2025 version files 554.23 KB
-
field.csv
1.56 KB
-
full_dataset.csv
477.16 KB
-
README.md
4.57 KB
-
remote.csv
34.27 KB
-
y.csv
36.66 KB
Abstract
Agricultural tree cover is declining globally, including the loss of large, scattered trees that function as keystone structures. Understanding the drivers of agricultural tree loss could help prevent further declines. However, the drivers of agricultural tree mortality vary across scales, from individual trees to landscapes, complicating efforts to quantify mortality risk. We applied high-resolution remote sensing and multi-method occupancy models to test hypotheses of drivers of tree mortality in a pastoral landscape of Southwestern Panama. Our approach enabled us to identify individual tree mortality across a >20,000 ha area, encompassing a wide range of land use intensity. Neighboring tree cover was the strongest predictor of mortality, with a higher probability of death for isolated trees relative to trees with many neighbors. Landscape-level covariates also predicted mortality risk, including higher mortality closer to roads and in parcels with larger areas. These results implicate land use intensity as a primary driver of agricultural tree loss in our study area. At the individual tree level, we found that larger trees were more likely to die than smaller trees. Our study suggests that the trees with high ecosystem service value in a fragmented landscape—large, isolated trees—also face the highest mortality risk. Supporting agricultural practices that maintain trees in pastures is likely to decrease tree mortality in our study site, broadly representative of cattle ranching landscapes across Latin America. Our workflow could be implemented in other landscapes globally to prioritize agricultural tree conservation, paving the way for increased tree survival and improved ecosystem services.
https://doi.org/10.5061/dryad.gxd2547xt
Description of the data and file structure
These data represent field and remotely sensed measurements of tree mortality, along with landscape- and individual-level covariates for tree mortality. The file titled “y.csv” contains two columns. The first column represents the 269 trees whose mortality was measured directly in the field. The second column represents mortality estimated from remotely sensed data for 6,154 trees. Each row corresponds to a single tree. A value of 1 indicates tree survival, while a value of 0 indicates tree death. Values of -1 indicate trees that were measured using remotely sensed data but not field data.
The file titled “field.csv” includes a single column with an index indicating which trees were measured in the field (out of 6,154 total trees). Similarly, the file titled “remote.csv” includes a single column with an index indicating which trees were measured only using remotely sensed data (out of 6,154 total trees). The file titled “full_dataset.csv” includes covariates, detection, and mortality data, with each row corresponding to a single tree. The R file, titled “Run_Bestmodel.R,” includes code to upload and process the .csv files and run the hierarchical Bayesian model for tree mortality. The Stan file, titled “Bestmodel_Int_twolevelsofdetection_yhat.stan,” contains the code for the mortality model, which accounts for imperfect detection.
Files and variables
File: y.csv
Description: This file contains tree mortality data, including field-measured and remotely sensed estimates.
Variables:
- y_1: Mortality data for 269 trees measured directly in the field (1 = survival, 0 = death).
- y_2: Mortality data estimated from remotely sensed data for 6,154 trees (1 = survival, 0 = death, -1 = trees measured only with remote sensing and not field data).
File: field.csv
Description: An index for which trees were measured with field data (out of 6,154 total trees).
Variables
- x, an index
File: remote.csv
Description: An index for which trees were measured only using remotely sensed data (out of 6,154 total trees).
Variables
- x, an index
File: full_dataset.csv
Description: This file contains tree-level covariates related to mortality.
Variables:
- sp: An index indicating which of the five species the tree belongs to.
- height: The average height of each tree in meters.
- ca: The crown area of each tree in square meters.
- siteExp: An index that combines slope and aspect to estimate exposure for each tree. This variable is unitless.
- elev: The elevation of each tree in meters.
- propid: An index indicating which property each tree belongs to. This property index was created randomly for the purpose of this study and cannot be linked to any identifying information on property identity.
- fc2012: The summed area of neighboring tree cover within a 30 m radius around each tree in square meters.
- roadP: The distance from each tree to the nearest road in meters.
- propSize: The area of each property in hectares.
Code/software
Two open source software packages are required to run the code. R version 4.4.1 (2024-06-14 ucrt) and Stan version “2.32.2”. The only package required to run the code is the “rstan” package.
The R file, titled “Run_Bestmodel.R,” includes code to upload and process the .csv files and run the hierarchical Bayesian model for tree mortality. The Stan file, titled “Bestmodel_Int_twolevelsofdetection_yhat.stan,” contains the code for the mortality model, which accounts for imperfectly detected mortality. Using the rstan package, the entire model can be run from R.
Access information
Other publicly accessible locations of the data:
- None
Data was derived from the following sources:
- Road network data was downloaded from the Smithsonian Tropical Research Institution’s GIS Data\
Portal (https://stridata-si.opendata.arcgis.com/). - Parcel data, representing landowner properties, is from Panama’s National Authority for the Administration of Land.
- Elevation, solar exposure, tree species identity, neighborhood crown area, and tree height are derived from aerial lidar and hyperspectral data, acquired by the Global Airborne Observatory in 2012.
The study was conducted in Southwestern Panama, covering a 23,000-hectare area in Los Santos province. This region experiences a dry season from December to March, with most of its 1,700 mm of annual rainfall occurring between April and November. Historically, the landscape was dominated by dry tropical forests, but extensive cattle ranching has reduced forest cover to a small fraction of its original extent. The current landscape consists of active pastures, riparian corridors, and second-growth forests, with ongoing agricultural de-intensification and reforestation efforts.
To assess tree mortality between 2012 and 2019, data were integrated from three sources: field measurements of individual trees, aerial hyperspectral-lidar imagery, and high-resolution satellite imagery. Field data were collected during initial surveys in 2012 and 2013, when individual trees were identified, georeferenced, and classified by species. These trees were revisited in 2019 to determine survival, with dead trees identified based on the presence of stumps or the complete absence of tree remains. Aerial hyperspectral-lidar imagery was collected in 2012, allowing for the segmentation of individual tree crowns across the landscape. In 2019, high-resolution satellite imagery was used to assess tree presence or absence by overlaying the 2012 tree crown segments onto the newer imagery.
The field survey in 2012-2013 recorded 1,140 tree crowns, identified to species using a GPS-enabled tablet. Five species were selected for analysis based on their classification accuracy in the hyperspectral imagery. In 2019, a subset of 269 trees from these species was revisited to confirm their status. A tree was classified as dead if no remains were found or if only a stump was present. Hyperspectral and lidar data from 2012 were used to generate a canopy height model and delineate tree crowns. A classification model applied to these data enabled species identification. From this dataset, a random selection of 10,000 tree crowns belonging to the five focal species was used for mortality assessment.
Tree mortality was evaluated by comparing the 2012 tree crowns with the high-resolution satellite imagery from 2019. Researchers visually assessed whether tree crowns remained present or had disappeared. These assessments were conducted in Google Earth, with trained technicians manually reviewing tree locations from both time points. A validation process ensured consistency in classification decisions.
Several covariates were included in the analysis to explore potential drivers of tree mortality. Tree size was quantified using measurements of tree height and crown area. Landscape features such as elevation, slope, and solar exposure were derived from a high-resolution digital elevation model. Distance to roads was measured as an indicator of accessibility and potential land-use pressure. Tree cover was calculated as the total area of neighboring tree crowns within a 30-meter radius of each focal tree. Additionally, trees were geolocated within land parcels using cadastral data to examine whether parcel size influenced mortality risk.
A statistical model was used to estimate tree mortality while accounting for detection errors in the remotely sensed data. This model integrated field measurements with remotely sensed observations to improve mortality estimates. A Bayesian framework was used to quantify mortality probabilities, providing a robust approach for assessing the effects of tree size and landscape features on tree survival.