Multi-scale modeling of species distributions, hydrology, & gene flow

Multi-scale modeling of species distributions, hydrology, & gene flow Douglas R. Leasure PhD. Candidate University of Arkansas Department of Biological Sciences

BIG data in GIS

http://www.gap.uidaho.edu/portal/datadownload.html http://www.mrlc.gov/index.php

http://mrdata.usgs.gov/geology/

http://www.ccafs-climate.org/ Future Climate

http://ned.usgs.gov/ Future Climate

Future Climate http://www.horizon-systems.com/nhdplus/

http://sedac.ciesin.columbia.edu/data/collection/usgrid Future Climate

Future Climate http://soils.usda.gov/survey/geography/statsgo/

http://www.worldclim.org/current Future Climate

Future Climate http://www.census.gov/geo/www/tiger/tgrshp2011/tgrshp2011.html

The problem of pattern and scale in ecology Simon A. Levin 1992

Spatial Scales Point Watershed Riparian Local Local-watershed Local-riparian Linear Path Stream Path Point-based data from a location s X, Y coordinates.

Spatial Scales Point Watershed Riparian Local Local-watershed Local-riparian Linear Path Stream Path All geographic areas that drain into a userdefined location

Spatial Scales Point Watershed Riparian Local Local-watershed Local-riparian Linear Path Stream Path Areas within a location s watershed AND within a user-defined distance (x) from streams.

Spatial Scales Point Watershed Riparian Local Local-watershed Local-riparian Linear Path Stream Path Areas within a user-defined radius (x) of a user-defined location.

Spatial Scales Point Watershed Riparian Local Local-watershed Local-riparian Linear Path Stream Path Areas within a location s local zone AND within its watershed.

Spatial Scales Point Watershed Riparian Local Local-watershed Local-riparian Linear Path Stream Path Buffered (x) stream path between all pairwise combinations of user-provided locations.

Geodata Crawler Centralized national geodatabase & Automated multi-scale data crawler

Geodata Crawler Centralized national geodatabase & Automated multi-scale data crawler Python scripting with ESRI ArcGIS

1. User-provided Locations

2. User-provided Boundary

3. Variable Selection Human population within 5 km radius Avg. January rainfall in watershed % forest in upstream riparian zone within 100 m of streams

Automated Sample Area Delineation & Data Collection Drainage Area: 926 sq km Forest Cover: 215 sq km Population: 482 people January Rainfall: 6 cm Avg. Terrain Slope: 3.2 degrees

Research Applications Species Distribution Modeling Sulphur Springs diving beetle Hydrological Modeling Mapping natural flow regimes & flow alteration Path analysis Gene flow among bluehead sucker populations in the Colorado River basin

Species Distribution Modeling

Sulphur Springs diving beetle Collaborators: Scott Longing & Pablo Bacon Endemic species of concern Headwater specialist 83 sample locations o Presence/absence Habitat associations at multiple spatial scales Heterosternuta sulphuria Predictive model

Sulphur Springs diving beetle Landscape Characteristics of Interest: Watershed Area % Forest % Canopy Cover % Urban % Impervious Surfaces % Agriculture Watershed Local-watershed (200m-2km radius) Avg./Max. Terrain Slope Avg./Max. Stream Channel Slope Human population density Road density Riparian (50-800m buffer)

Bayesian Information Criteria (BIC) Effects of urbanization at multiple spatial scales Multi-model comparison with logistic regression 108 106 104 102 100 98 0 400 800 1200 1600 2000 Site Radius (meters)

Forested Riparian Buffers Non-parametric Multiplicative Regression McCune 2011, McCune & Medford 2004 Forest Cover in 100 m Riparian Zone Low Med High

Predicted Occurrences Maxent Presence-only modeling Phillips 2006

Predicted Occurrences Maxent Presence-only modeling Phillips 2006 R Package: BIOMOD Thuiller et al. 2009. BIOMOD - a platform for ensemble forecasting of species distributions. Ecography.

Predicted Occurrences Maxent Presence-only modeling Phillips 2006 hydrology? population connectivity?

Hydrological Modeling

Mapping Natural Flow Regimes 7 natural flow regimes in Arkansas region (Leasure et al., in review) 67 reference streams with USGS gauges Collaborators: Dan Magoulick & Scott Longing Predict natural flow at un-gauged and disturbed streams

Landscape-Climate Flow Regime PRECIP SOIL

Predicted Natural Flow Regimes Random Forests Classification Breiman. 2001. 40,000 stream segments Predict natural flow regime Error Rate = 37%

Assessing Hydrologic Alteration Carlisle et al. 2011. Random Forest to predict expected natural conditions (E) Gauge data to measure observed current conditions (O) O/E to measure flow alteration

Path Analysis & Animal Movements

Gene flow among bluehead sucker populations Collaborators: Michael & Marlis Douglas Bluehead Sucker Catostomis discobolus Microsat loci: 16 Specimens: 1092 Locations: 39

Barriers to Gene Flow??

No barrier = Similar allele freq.

Barrier = Dissimilar allele freq.

Fst Dissimilarity Matrix Fst

Map Fst to Stream Segments StreamTree (Kalinowski 2007) Fst Fst Fst Fst Fst Fst

Landscape/segment Fst/segment Fst Fst Fst Fst Fst Fst

Map Gene Flow Resistance Gene flow resistance Red = High Green = Low

Multi-scale Analysis as an Interdisciplinary Bridge

Hydrology Habitat Suitability Population Connectivity

Hydrology Population Connectivity Habitat Suitability

Hydrology Future Climate Future Climate Future Climate Population Connectivity Habitat Suitability

Hydrology Population Connectivity Habitat Suitability

Hydrology Future Climate Population Connectivity Habitat Suitability

Hydrology Population Connectivity Habitat Suitability

Strategies Going Forward Big data from GIS & remote sensing Automated multi-scale data collection Contemporary modeling tools Multi-model inference Machine learning Ensemble forecasting Hierarchical Bayesian models*

Questions?