Limits...
Landscape Analysis of Adult Florida Panther Habitat.

Frakes RA, Belden RC, Wood BE, James FE - PLoS ONE (2015)

Bottom Line: Sensitivity analysis showed that the presence of human populations, roads, and agriculture (other than pasture) had strong negative effects on the probability of panther presence.Forest cover and forest edge had strong positive effects.This model should be useful for evaluating the impacts of future development projects, in prioritizing areas for panther conservation, and in evaluating the potential impacts of sea-level rise and changes in hydrology.

View Article: PubMed Central - PubMed

Affiliation: U.S. Fish and Wildlife Service, South Florida Ecological Services Office, 1339 20th Street, Vero Beach, Florida, United States of America.

ABSTRACT
Historically occurring throughout the southeastern United States, the Florida panther is now restricted to less than 5% of its historic range in one breeding population located in southern Florida. Using radio-telemetry data from 87 prime-aged (≥3 years old) adult panthers (35 males and 52 females) during the period 2004 through 2013 (28,720 radio-locations), we analyzed the characteristics of the occupied area and used those attributes in a random forest model to develop a predictive distribution map for resident breeding panthers in southern Florida. Using 10-fold cross validation, the model was 87.5 % accurate in predicting presence or absence of panthers in the 16,678 km2 study area. Analysis of variable importance indicated that the amount of forests and forest edge, hydrology, and human population density were the most important factors determining presence or absence of panthers. Sensitivity analysis showed that the presence of human populations, roads, and agriculture (other than pasture) had strong negative effects on the probability of panther presence. Forest cover and forest edge had strong positive effects. The median model-predicted probability of presence for panther home ranges was 0.81 (0.82 for females and 0.74 for males). The model identified 5579 km2 of suitable breeding habitat remaining in southern Florida; 1399 km2 (25%) of this habitat is in non-protected private ownership. Because there is less panther habitat remaining than previously thought, we recommend that all remaining breeding habitat in south Florida should be maintained, and the current panther range should be expanded into south-central Florida. This model should be useful for evaluating the impacts of future development projects, in prioritizing areas for panther conservation, and in evaluating the potential impacts of sea-level rise and changes in hydrology.

No MeSH data available.


Related in: MedlinePlus

Variable importance.Importance was calculated based on mean decrease in model accuracy (black bars) and mean decrease in Gini index (gray bars). Importance scores were standardized relative to the most important variable by each method. Variables are ranked from highest to lowest importance, based on combined scores from the two methods. Wet_For = wetland forest, Pop_Dens = human population density, For_Edge = forest edge, dry_depth = average dry season water depth, wet_depth = average wet season water depth, Wet_Shrub = wetland shrub, Rd_Dens = road density, FW_Wet = open freshwater wetlands, Ag = agricultural, Up_For = upland forest, Grass = grasslands/dry prairies, Water = open water, Up_Shrub = upland shrub, SW_Wet = saltwater wetland.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4519242&req=5

pone.0133044.g004: Variable importance.Importance was calculated based on mean decrease in model accuracy (black bars) and mean decrease in Gini index (gray bars). Importance scores were standardized relative to the most important variable by each method. Variables are ranked from highest to lowest importance, based on combined scores from the two methods. Wet_For = wetland forest, Pop_Dens = human population density, For_Edge = forest edge, dry_depth = average dry season water depth, wet_depth = average wet season water depth, Wet_Shrub = wetland shrub, Rd_Dens = road density, FW_Wet = open freshwater wetlands, Ag = agricultural, Up_For = upland forest, Grass = grasslands/dry prairies, Water = open water, Up_Shrub = upland shrub, SW_Wet = saltwater wetland.

Mentions: The 15 explanatory variables are ranked from highest to lowest importance in Fig 4. Human population density stood out as the most important variable affecting model accuracy, followed by wetland forest. The amount of wetland forest and forest edge were the most important variables according to the Gini index. The top five variables were the same by both importance measures, although in different order. Using the combined relative importance from the two methods, the order of variable importance was wetland forest > human density > forest edge > dry season water depth > wet season water depth. It is surprising that both water depth variables were included in the top five, even though they were highly collinear. Wetland shrubs, road density, freshwater wetlands, and agricultural use were of medium importance relative to the other variables. The upland cover types (upland forests, grasslands, and upland shrubs) did not score as highly in importance as expected. Along with urban, open water, and saltwater wetlands, these were among the least predictive variables. There was greater variation in importance among the variables based on the Gini index compared with model accuracy. According to the accuracy analysis, all variables contributed somewhat to model accuracy.


Landscape Analysis of Adult Florida Panther Habitat.

Frakes RA, Belden RC, Wood BE, James FE - PLoS ONE (2015)

Variable importance.Importance was calculated based on mean decrease in model accuracy (black bars) and mean decrease in Gini index (gray bars). Importance scores were standardized relative to the most important variable by each method. Variables are ranked from highest to lowest importance, based on combined scores from the two methods. Wet_For = wetland forest, Pop_Dens = human population density, For_Edge = forest edge, dry_depth = average dry season water depth, wet_depth = average wet season water depth, Wet_Shrub = wetland shrub, Rd_Dens = road density, FW_Wet = open freshwater wetlands, Ag = agricultural, Up_For = upland forest, Grass = grasslands/dry prairies, Water = open water, Up_Shrub = upland shrub, SW_Wet = saltwater wetland.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4519242&req=5

pone.0133044.g004: Variable importance.Importance was calculated based on mean decrease in model accuracy (black bars) and mean decrease in Gini index (gray bars). Importance scores were standardized relative to the most important variable by each method. Variables are ranked from highest to lowest importance, based on combined scores from the two methods. Wet_For = wetland forest, Pop_Dens = human population density, For_Edge = forest edge, dry_depth = average dry season water depth, wet_depth = average wet season water depth, Wet_Shrub = wetland shrub, Rd_Dens = road density, FW_Wet = open freshwater wetlands, Ag = agricultural, Up_For = upland forest, Grass = grasslands/dry prairies, Water = open water, Up_Shrub = upland shrub, SW_Wet = saltwater wetland.
Mentions: The 15 explanatory variables are ranked from highest to lowest importance in Fig 4. Human population density stood out as the most important variable affecting model accuracy, followed by wetland forest. The amount of wetland forest and forest edge were the most important variables according to the Gini index. The top five variables were the same by both importance measures, although in different order. Using the combined relative importance from the two methods, the order of variable importance was wetland forest > human density > forest edge > dry season water depth > wet season water depth. It is surprising that both water depth variables were included in the top five, even though they were highly collinear. Wetland shrubs, road density, freshwater wetlands, and agricultural use were of medium importance relative to the other variables. The upland cover types (upland forests, grasslands, and upland shrubs) did not score as highly in importance as expected. Along with urban, open water, and saltwater wetlands, these were among the least predictive variables. There was greater variation in importance among the variables based on the Gini index compared with model accuracy. According to the accuracy analysis, all variables contributed somewhat to model accuracy.

Bottom Line: Sensitivity analysis showed that the presence of human populations, roads, and agriculture (other than pasture) had strong negative effects on the probability of panther presence.Forest cover and forest edge had strong positive effects.This model should be useful for evaluating the impacts of future development projects, in prioritizing areas for panther conservation, and in evaluating the potential impacts of sea-level rise and changes in hydrology.

View Article: PubMed Central - PubMed

Affiliation: U.S. Fish and Wildlife Service, South Florida Ecological Services Office, 1339 20th Street, Vero Beach, Florida, United States of America.

ABSTRACT
Historically occurring throughout the southeastern United States, the Florida panther is now restricted to less than 5% of its historic range in one breeding population located in southern Florida. Using radio-telemetry data from 87 prime-aged (≥3 years old) adult panthers (35 males and 52 females) during the period 2004 through 2013 (28,720 radio-locations), we analyzed the characteristics of the occupied area and used those attributes in a random forest model to develop a predictive distribution map for resident breeding panthers in southern Florida. Using 10-fold cross validation, the model was 87.5 % accurate in predicting presence or absence of panthers in the 16,678 km2 study area. Analysis of variable importance indicated that the amount of forests and forest edge, hydrology, and human population density were the most important factors determining presence or absence of panthers. Sensitivity analysis showed that the presence of human populations, roads, and agriculture (other than pasture) had strong negative effects on the probability of panther presence. Forest cover and forest edge had strong positive effects. The median model-predicted probability of presence for panther home ranges was 0.81 (0.82 for females and 0.74 for males). The model identified 5579 km2 of suitable breeding habitat remaining in southern Florida; 1399 km2 (25%) of this habitat is in non-protected private ownership. Because there is less panther habitat remaining than previously thought, we recommend that all remaining breeding habitat in south Florida should be maintained, and the current panther range should be expanded into south-central Florida. This model should be useful for evaluating the impacts of future development projects, in prioritizing areas for panther conservation, and in evaluating the potential impacts of sea-level rise and changes in hydrology.

No MeSH data available.


Related in: MedlinePlus