Date of Award





Earth and Space Science - Environmental Science Track

Degree Type



Earth & Space Science

First Advisor

Troy Keller

Second Advisor

Chester Figiel

Third Advisor

Clifton Ruehl


Cambarus harti is a state-listed endangered, endemic crayfish found only in three counties in mid-west Georgia. Several studies have attempted to characterize the biology and ecology of this crayfish, however data regarding the distribution of this rare, endemic crayfish remains limited. The International Union for Conservation of nature stated that in order to create an effective conservation plan, the known distribution must be expanded. Species distribution models are a cost-effective way to identify locations that have similar habitat characteristics to those with known populations. One species distribution model, Maximum Entropy (MaxEnt), is the preferred approach when modeling species, like C. harti, that only have a few known locations. I used MaxEnt to create a predictive, spatial model for C. harti. The MaxEnt model was developed using 14 C. harti occurrence locations and five environmental layers (distance to water, soils, geology, landcover, and slope) for six counties in West Central Georgia. Using a 2km buffer for background points the model produced a receiver operating characteristic curve (ROC) with an area under the curve (AUC) value of 0.97. The high AUC value correlates with the high discriminatory power of the model. The five environmental layers were weighted differently starting with the most important; distance to water (35.4%), soil (29.1%), landcover (14.8%), geology (14.3%), and slope (6.3%). The model's results covered 6110 km2 in Georgia with probabilities of C. harti occurrence ranging from: 0%-100% [(0%-10%) 4432 km2, (10%-20%) 622km2, (20%-30%) 371 km2, (30%-40%) 214 km2, (40%-50%) 150 km2, (50%-60%) 137 km2, (60%- 70%) 122 km2, (70%-80%) 30 km2, (80%-90%) 30 km2, (90%-100%) 2 km2]. The MaxEnt model was evaluated through two different ground truthing methods. The first approach examined the model’s overall accuracy by randomly sampling for crayfish at 30 sites across three model predicted probability ranges (0%-20%, 40%-60%, 80%-100%). The second method evaluated the model output at finer resolutions by comparing probabilities of known C. harti locations to sites within 183m of known locations but without crayfish. The first approach yielded no verified C. harti locations within any of the sampling brackets. The second method confirmed that the model was ineffective at identifying C. harti habitat on large spatial scales (i.e. locally).

Review of the environmental data layers used to create the model uncovered errors in the underlying data. For example, the USGS National Hydrology Dataset was a large source of error, with many streams improperly mapped. This data set was used to create the distance to water grid. It is clear that data resolution, accuracy and resolution have not advanced to the point where these models can justifiably be used to map the potential habitat of this endemic burrowing crayfish. Cambarus harti is likely just one of many species this model is inadequate for; models for amphibians and other species that rely on ground water or surface water depicted by the USGS National Hydrology Dataset would lack adequate data. A high resolution (10m) groundwater layer needs to be obtained in order to more accurately model burrowing crayfish habitat.