Application of LiDAR to assess the habitat selection of an endangered small mammal in an estuarine wetland environment
Data files
Jan 31, 2024 version files 328.25 KB
-
All_Sites_5m.csv
317.42 KB
-
All_Sites_Grids.csv
6.08 KB
-
README.md
4.75 KB
Abstract
Light detection and ranging (lidar) has emerged as a valuable tool for examining the fine-scale characteristics of vegetation. However, lidar is rarely used to examine coastal wetland vegetation or the habitat selection of small mammals. Extensive anthropogenic modification has threatened the endemic species in the estuarine wetlands of the California coast, such as the endangered salt marsh harvest mouse (Reithrodontomys raviventris; SMHM). A better understanding of SMHM habitat selection could help managers better protect this species. We assessed the ability of airborne topographic lidar imagery in measuring the vegetation structure of SMHM habitats in a coastal wetland with a narrow range of vegetation heights. We also aimed to better understand the role of vegetation structure in habitat selection at different spatial scales. Habitat selection was modeled from data compiled from 15 small mammal trapping grids collected in the highly urbanized San Francisco Estuary in California, USA. Analyses were conducted at three spatial scales: microhabitat (25 m2), mesohabitat (2,025 m2), and macrohabitat (10,000 m2). A suite of structural covariates was derived from raw lidar data to examine vegetation complexity. We found that adding structural covariates to conventional habitat selection variables significantly improved our models. At the microhabitat scale in managed wetlands, SMHM preferred areas with denser and shorter vegetation, and selected for proximity to levees and taller vegetation in tidal wetlands. At the mesohabitat scale, SMHM were associated with a lower percentage of bare ground and with pickleweed (Salicornia pacifica) presence. All covariates were insignificant at the macrohabitat scale. Our results suggest that SMHM preferentially selected microhabitats with access to tidal refugia and mesohabitats with consistent food sources. Our findings showed that lidar can contribute to improving our understanding of habitat selection of wildlife in coastal wetlands and help to guide future conservation of an endangered species.
Hagani, J. S.1, J. Y. Takekawa1, S. M. Skalos2,3, M. L. Casazza2, M. K. Riley4, S. A. Estrella4, L. M. Barthman-Thompson5, K. R. Smith6,7, K. J. Buffington8, and K. M. Thorne8*
1Suisun Resource Conservation District, 2544 Grizzly Island Road, Suisun City, CA USA 94585
2USGS Western Ecological Research Center, Dixon Field Station, 800 Business Park Drive, Dixon, CA USA 95620
3California Department of Fish and Wildlife, 1010 Riverside Parkway, West Sacramento, CA USA 95605
4California Department of Fish and Wildlife, 2825 Cordelia Road, Suite 100, Fairfield, CA USA 94534
5California Department of Fish and Wildlife, 2109 Arch Airport Road, Stockton, CA USA 95206
6WRA, Inc., 2169-G Francisco Boulevard East, San Rafael, CA USA 94901
7Department of Wildlife, Fish and Conservation Biology, UC Davis, 1088 Academic Surge, 455 Crocker Lane, University of California Davis, Davis, CA USA 95616
8USGS Davis Field Station, 1 Shields Avenue, University of California Davis, Davis, CA USA 95616
*Corresponding author: jhagani@suisunrcd.org
R Code
· “SMHM Habitat Selection Models.R”
o Code for running generalized linear models (GLMs) to assess the importance of various lidar and/or conventional BASE covariates on the habitat selection of salt marsh harvest mice at any spatial scale.
· “Lidar Processing.R”
o Code for normalizing a raw lidar point cloud using an external DEM and for extracting a suite of vegetation structure covariates from the resulting normalized dataset.
Datasets
· “All_Sites_5m.csv”
o Data at the trap-level (25 m2) spatial scale for every small mammal trap location analyzed in this study. GPS coordinates have been excluded for safety. Includes both BASE and habitat structure covariates.
§ “Grid” = name of trapping grid
§ “Trap” = trap number
§ “RERA_CPUE” = catch per unit effort of SMHM
§ “Type” = habitat type (managed wetland, tidal wetland, or upland area)
§ “zmax” = maximum vegetation height (ft)
§ “zmean” = mean vegetation height (ft)
§ “zsd” = standard deviation of vegetation height (ft)
§ “zskew” = skewness of vegetation height
§ “zkurt” = kurtosis of vegetation height
§ “pzabovezmean” = percentage of returns above mean vegetation height (%)
§ “pzaboveX” = Percentage of returns above X (0.25, 0.50, 0.75) meters (%)
§ “zqX” = The vegetation height at each 5% quantile (zq5, zq10…zq95). (ft)
§ “zpcumX” = Cumulative percentage of return of the Xth bin (zpcum1, zpcum2…zpcum9) (%)
§ “pground” = percentage of lidar returns classified as bare ground (vegetation height = 0) (%)
§ “Elevation” = average ground surface elevation (m)
§ “Levee_Dist” = average distance to the nearest levee (m)
§ “Urban_Dist” = average distance to the nearest urban area (m)
§ “Vegetation” = code for the most dominant vegetation species
· “All_Sites_Grids.csv”
o Data at the grid-level (~10,000 m2) spatial scale for every trapping grid analyzed in this study. Includes both BASE and habitat structure covariates. Height units are in U.S. Feet (ft) unless otherwise stated below.
§ “Grid” = name of the trapping grid
§ “Nights” = number of survey nights trapping occurred
§ “Traps” = number of traps
§ “RERA_Unique” = number of unique SMHM captured
§ “RERA_CPUE” = catch per unit effort of SMHM
§ “Type” = habitat type (managed wetland, tidal wetland, or upland area)
§ “zmax” = maximum vegetation height (ft)
§ “zmean” = mean vegetation height (ft)
§ “zsd” = standard deviation of vegetation height (ft)
§ “zskew” = skewness of vegetation height
§ “zkurt” = kurtosis of vegetation height
§ “pzabovezmean” = percentage of returns above mean vegetation height (%)
§ “pzaboveX” = Percentage of returns above X (0.25, 0.50, 0.75) meters (%)
§ “zqX” = The vegetation height at each 5% quantile (zq5, zq10…zq95) (ft)
§ “zpcumX” = Cumulative percentage of return of the Xth bin (zpcum1, zpcum2…zpcum9) (%)
§ “pground” = percentage of lidar returns classified as bare ground (vegetation height = 0) (%)
§ “Elevation” = average ground surface elevation (m)
§ “Levee_Dist” = average distance to the nearest levee (m)
§ “Urban_Dist” = average distance to the nearest urban area (m)
§ “Vegetation” = code for the most dominant vegetation species