Prearchaic land use in Grass Valley, Nevada: A novel statistical implementation of optimal distribution modeling

Using Prearchaic (PA) sites in Grass Valley, NV (Fig. 1), this project investigates (i) environmental factors driving variation in PA settlement and (ii) geomorphological factors driving variation in PA surface visibility. Building on previous research [1,2], we evaluate variables using Ideal Free Distribution [3] and Maximum Entropy (MaxEnt) [4].

Prearchaic Land Use in Grass Valley, Nevada: A Novel Statistical Implementation of Optimal Distribution Modeling ARCHAEOLOGICAL CENTER 1 1 1 2 3 1 Kenneth B. Vernon , Peter M. Yaworsky , Kate E. Magargal , D. Craig Young , David Zeanah , Brian F. Codding 1 University of Utah 2 Far Western Anthropological Research Group 3 California State University, Sacramento Predictions OR (a) GAM ID Grass Valley (b) MaxEnt WY 30 Percent Contribution Permutation Importance 1.0 UT 0.8 AZ 0.0 Methods and Data 0 5 10 km For more information, contact Kenneth B. Vernon, Department of Anthropology, University of Utah ([e-mail]) Prearchaic Land Use in Grass Valley, NV ke NP P EH PH T Acknowledgements Special thanks to Ashley K. Parker (Utah), Michael Weight (Utah), Bob Elston (UNR), James F. O'Connell (Utah), and the UU Archaeological Center Lab Group. This material is based upon work supported by the National Science Foundation under Grant No. (BCS-1632521, -1632522, -1632526). 0.6 0.7 0.8 0.9 We bootstrap both the GA and MaxEnt models through 100 iterations to evaluate the predictive power (AUC median) and robustness (AUC dispersion) of each. Selected References 0.5 Predictor Variables DEM: Digital Elevation Slope (from DEM) MI: Moisture Index NPP: Primary Productivity Soil age model (Fig. 2) I LP: Late Pleistocene I EH: Early Holocene I PHT: P-H Transition Tobler cost distance to: I P. lake shoreline I Springs 1. Elevation and MI contribute the most to PA habitat suitability. 2. Contemporaneous soil age layers predict PA surface visibility. 3. MaxEnt is more powerful and robust with small training sets. Performance AUC GV Watershed Pleistocene Lake Soil Age Model EH LP PHT s Our results show that: Figure 3: Grass Valley Prediction Rasters Response Variables Prearchaic sites (n=18) Figure 5: MaxEnt Discussion We fit a predictive model to the data using a Maximum Entropy approach. For comparison, we also fit a generalized additive model (GAM) using Maximum Likelihood. Figure 2: GV Reference La 0.2 ng Figure 1: GV Overview Sp ri 0.4 LP 300 km Slo pe 75 150 MI 0 DE M Present Day Lake/Playa Pluvial Lakes (max extent) 0.6 For the model fitted by MaxEnt, elevation (DEM), moisture (MI), slope, and Late Pleistocene (LP) soil age contribute most to the model's predictive power. These we interpret in terms of habitat suitability and surface visibility. 0 NV 10 CA Percent 20 Using Prearchaic (PA) sites in Grass Valley, NV (Fig. 1), this project investigates (i) environmental factors driving variation in PA settlement and (ii) geomorphological factors driving variation in PA surface visibility. Building on previous research [1,2], we evaluate variables using Ideal Free Distribution [3] and Maximum Entropy (MaxEnt) [4]. Predictors 40 Introduction GAM MaxEnt Figure 4: AUC Distribution [1] Codding, B. F. et al (2016). "Prearchaic Adaptations in the Central Great Basin: Preliminary findings from a stratified open-air site in Grass Valley, Nevada." Great Basin Anthropological Conference, Reno, Nevada, 2016. URL: https://collections.lib.utah.edu/details?id=1202780. [2] Elston, R.G. et al (2014). "Living Outside the Box: An Updated Perspective on Diet Breadth and Sexual Division of Labor in the Prearchaic Great Basin." Quaternary International. [3] Fretwell and Lucas (1969). "On territorial behavior and other factors influencing habitat distribution in birds." Acta Biotheoretica, 19, 16-36. [4] Phillips et al (2006). "Maximum entropy modeling of species geoographic distribution." Ecological Modelling, 19, 231-259. Department of Anthropology - Archaeological Center