Species distribution modeling of
Hibiscus brackenridgei on Oahu

Abstract

Anthropogenic activities threaten the existence of native species within various ecosystems. In Hawaiian dry forest, 45% of all tropical dry forest trees are federally endangered (Pau et al., 2009). As a result, there is a need for understanding the current range of endangered species, the factors that affect their ranges, and ultimately, possible areas for the most efficient and successful restoration. Through species distribution modeling, climate and environmental factors along with presence data are inputted into a program in order to generate a model showing the predicted distribution for a species, and the probability of its occurrence for a given area. Additionally, the model identifies the variable with the highest percent contribution to the predicted niche. This research utilizes species distribution modeling in order to evaluate the possible range for Hibiscus brackenridgei, which is both the state flower and a federally endangered species within the dry forest, on Oahu. Based on the results, the predicted occurrence for the species is in the northwest region, and therefore, it is suggested that restoration for the species be implemented within this region. The variable with the highest percent contribution to the distribution was slope.

Introduction

Tropical dry forests are among the most endangered ecosystems worldwide. More than 25% of the endangered plant taxa in the entire Hawaiian flora are from dry forest or dry scrub ecosystems and approximately 20% of taxa from these habitats are known to be extinct (Sakai et al. 2002). Hawaiian dry forests were initially deforested during the Polynesian period. Large-scale ranching and commercial agriculture further intensified the impact on dry forest regions after European arrival. Forestry practices in the early 20th century prescribed widespread planting of non-native tree species in an attempt to protect the quality of watersheds. Many of these species were invasive, transforming the remaining fragments by replacing native species that were not as competitive due to evolutionary or ecological patterns (Gillespie et al. 2008; Pau et al. 2009). Land development, fire, non-native ungulate grazing, and invasions by non-native plant species have been identified as the most serious threats to Hawaiian dry forests. Tropical dry forests in Hawaii may be the most endangered forest type in the world with 45% of the trees and shrubs in the dry forest region on the federal endangered species list (Pau et al. 2009). However, there is currently no information on the current distribution of the dry forest as well as potential sites for endangered species restoration.

Species distribution modeling is one method for evaluating the potential range of a species and has been increasingly used to address problems across a variety of fields including biogeography, ecology, conservation biology, and climate change science (Guisan and Thuiller 2005). Species distribution models require two sets of inputs in order to create distribution maps of the targeted species within landscapes, regions, and continents. These include presence data from museum vouchers or field surveys and environmental predictors’ data in a geographic information system format (Guisan and Thuiller, 2005). The species distribution modeling creates a probability map of the distribution of a target species based on the known location of the species and environmental predictors such as climate data. This modeling approach has been shown to be a powerful tool in improving research in conservation and restoration (Araujo and Williams, 2000). Species distribution modeling is especially useful in tropical regions with deficient geographic locality data resulting from small sample sizes or imprecise locations of specimens (Graham et al. 2004A).

This research aims to identify the potential distribution for Hibiscus brackenridgei on Oahu as well as the climate and environmental factors that have the greatest effect on its distribution. An additional goal is to compare human population size across Oahu with the species' predicted range. This will be assessed by comparing the generated predictive model with a classified night light image of Oahu.

Methods

Study Site

The research considered 49 geographic point locations for Hibiscus brackenridgei on the island of Oahu. These locations are the only known specimens of the species on the island and were collected by the US Military.

Data Collection

Climate data was downloaded from the WorldClim data set version 1.4 (Hijmans et al., 2005; www.worldclim.org), which includes 19 bioclimatic variables. These metrics have a 1 km pixel resolution, are derived from monthly temperature and rainfall climatologies, and represent biologically meaningful variables for characterizing species range. Additionally, elevation data from WorldClim was used to calculate slope and aspect using ArcGIS 9.4. A boundary shapefile for Oahu was used to format the 19 bioclimatic variables, slope, and aspect to the same extent and then, resample them to a North American Datum 1989 UTM Zone 4N projection.

Maxent

The program, Maxent, was used to model the species' distribution. The 19 bioclimatic variables, slope, aspect, and the species' geographic point locations were inputted into the program which generated a map showing the probability of the species' distribution based on these environmental variables. Default parameters were used for the modeling. In addition to producing a probability distribution map, the program also generated values for each variable to show its overall contribution to the predicted species' range.

Night Light

In addition to the species distribution model for Hibiscus brackenridgei, a night light analysis was performed for the island of Oahu. A world light image was downloaded from NOAA DMSP and analyzed in ENVI 4.3. A density slice was taken to show areas on the island emitting light, and brightness was classified into 3 categories of reflectance. This image was then compared to the predicted distribution model created for Hibiscus brackenridgei.

Results

Maxent

Based on the slope, aspect, 19 bioclimatic variables, and the geographic point locations for Hibiscus brackenridgei, a model was created showing the predicted distribution of the species. The values generated range from zero (blue) to one (red), with one being the highest probability of the species occurring in a specific place. The model displays the highest probability for species occurrence on the north-western region of Oahu and isolated areas along the western coast. The highest probability value is .92 (orange) while the lowest probability numeric is 0, meaning that there is no probability of the species occurring in that area based on the variables and point locations.

In addition to the predicted distribution map, the percent contribution of each variable for the predicted range was calculated. Slope had the greatest effect on the distribution model with a 46.4% contribution. Following slope, bio2 had the next highest percentage contribution of 27.3%. Other variables that affected the predicted distribution and their values were aspect with 15.5%, bio15 with 5.5%, bio13 with 4.4%, and bio3 with 1%. The remaining bioclimatic variables had no effect on the predicted distribution for Hibiscus brackenridgei.


Discussion

Maxent

The highest probability of species occurrence in the northwest region of Oahu is most likely due to slope which contributed 46.4% to the predicted distribution. Slope may have such a high impact on the distribution because it is more difficult for trees to grow on steeper slopes. The second highest probability was bio2, or mean diurnal range, which is calculated by subtracting the mean of the monthly minimum temperature by the mean of the monthly maximum temperature, and had a 27.3% effect on the species distribution. The extremity of temperature during a month might affect the growth of a tree for a given season. The third contributing factor was aspect which affected the generated model by 15.5%. Aspect is the direction a slope faces which affects a species exposure to sunlight. Having a higher exposure to sunlight, would provide a suitable environment for a tree species. Bio15, representing precipitation seasonality, had a contribution value of 5.5% while Bio13, or precipitation of the wettest month, affected the model by 4.3%. Precipitation is essential for the growth of Hibiscus brackenridgei which occurs in the tropical dry forest and therefore, the species will be restricted to ranges that receive rainfall. Lastly, bio3, the identification for isothermality, had a contribution of 1% on the predicted species distribution. The remaining variables had a contribution of zero percent suggesting that they have no discernible effect on the probable range of the species. Based on the generated model, it is suggested that restoration of Hibiscus brackenridgei be implemented in the north-western region of Oahu in order to allow for the most efficient and successful conservation of the species.

Night Light

In order to evaluate human population size as a possible limiting factor of the species distribution, the species distribution model was compared to a night-time light image of Oahu. Analyzing night-time light imagery based on the intensity of light allows for a rough estimate of population size in various regions (Sutton et al., 1997). The red region represents the highest amount of light emitted, followed by green, and then yellow. According to the night-time light image, most of the population occurs in the southwest corner of Oahu. As a result, most of the geographic point locations for the species were found outside of these areas. Compared with the predictive species range map, the probability of a species occurring in these areas is close to zero, except for a few outliers which predicted higher probabilities of species occurrence. Additionally, areas where there is a high probability of species occurrence are correlated with low night-light emission, and therefore, lower human population sizes.

Limitations

Although the model should prove useful when considering conservation for Hibiscus brackenridgei, it should not be used as an absolute identification of the species range. Because the values of distribution are estimated from climate and environmental variables, many other factors, such as landuse classification, are not considered, which will ultimately affect the distribution of the species. Additionally, the model is limited by the small sample size which results from the endangered nature of the species on Oahu. Fewer geographic point locations decreases the accuracy of the probabilities of species occurrence generated by the model.

Future Research

For future research, more point locations should be inputted into the model if it is feasible through increased field survey. Another way to improve the accuracy of the model is to include more variables, especially remote sensing variables, to consider factors such as canopy height. While climate, environmental, and remote sensing variables have a direct effect on the species range, it may also be useful to evaluate human factors such as land use classification to gain a better understanding of the current distribution of the species. Lastly, other endangered species on Oahu should be modeled in order to evaluate overlapping ranges and establish possible areas for economically viable restoration sites.

Figures




















Figure 1: : The species distribution map displays the predicted range for Hibiscus brackenridgei based on environmental and climate variables along with geographic point locations. Each area is given a probability value between zero and one, with one being the highest probability that a species will occur.

























Figure 2: The table shows the contribution of the bioclimatic variables to the predicted distribution.























Figure 3: The table identifies the codes for each environmental and climate variable.




















Figure 4: The species distribution map displays the predicted range for Hibiscus brackenridgei based on environmental and climate variables along with geographic point locations. Each area is given a probability value between zero and one, with one being the highest probability that a species will occur.

References

Araujo and Williams. 2000. Selecting areas for species persistence using occurrence data. Biological Conservation 96.3: 331-345.

Gillespie, T.W., et al. 2008. Measuring and modeling biodiversity from space. Progress in Physical Geography 32, 203-221.

Guisan, A., Thuiller, W., 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8, 993-1009.

Hijmans, R.J., et al. 2005. Very high resolution interpolated climate surfaces for global

land areas. International Journal of Climatology 25, 1965-1978.

Pau, S., T. W. Gillespie, and J. P. Price. 2009. Natural history, biogeography, and endangerment of Hawaiian dry forest trees. Biodiversity and Conservation 18:3167-3182.

Sakai A.K. et al. 2002. Patterns of endangerment in the Hawaiian flora. Systematic Biology 51 :276-302.

Sutton, Paul. 1997. Modeling population density with night-time satellite imagery and GIS. Computer, Environment, and Urban Systems 21.3: 227-224.