Habitat Modeling for Sprague s Pipit in Montana Data and Deductive and Inductive Models for Montana

Transcription

1 Habitat Modeling for Sprague s Pipit in Montana Data and Deductive and Inductive Models for Montana Presentation to USFWS and other Federal and State Agencies April 10 th, 2012 in Helena, Montana Bryce Maxell - Senior Zoologist

2 Data Available For Animal Species Modeling in Montana 1,218,000+ Animal Observation Records 139,000+ Structured Animal Surveys Positive Data Widely Available Negative Data from Systematic Surveys Rare Wide range of dates, but most data is since 1985 Locational Uncertainty Assigned to all Records Records receive 2 to 3 rounds of review for mapping accuracy, precision, and association with appropriate habitats. Limitations on Statewide Environmental Layers

3 Data and Data Access for Sprague s Pipits Montana Field Guide -

4 Data and Data Access for Sprague s Pipits Montana Natural Heritage Tracker -

5 Great Plains Mixedgrass Prairie (Red) Agriculture (Gray) Deductive Model Black points = point count survey Purple points = detections * All Mixedgrass Prairie is not the same!

6 Great Plains Mixedgrass Prairie Density by Township Percent Mixedgrass Prairie = Blue (0 25%), Yellow ( ), Orange ( ), Red ( )

7 Great Plains Mixedgrass Prairie Density by Section PLSS sections with SPPI had mean percent MGP per section of ~ 60% (median = 71%) Mean is probably biased low because some areas along front and Madison Valley contain no MGP Also, MGP can get a little shrubby and is on different soils in the south

8 Surveys Needed! Great Plains Mixedgrass Prairie Public Lands Only

9 Inductive Modeling with Maxent Ideal for positive data only Uses random background points as pseudo-absences Uses different levels of complexity depending on amount of data available Logistic output interpretable as relative habitat suitability Output is generally satisfactory in regions with at least a few observation points Equivalent or superior to other predictive modeling approaches (Elith et al Ecography 29: ) Uses empirical distributions to constrain estimated distributions Iterative machine learning approach with deterministic outcome

10 11 Environmental Layers Layer Identifier Description Aspect CONTEWASP CONTNSASP East to West continuous aspect North to South continuous aspect Bias BIAS Categorical layer representing potential underlying biases inherent in the observation database as a result of proximity to roads and public lands (three categories) Elevation CONTELEV Elevation in meters form the National Elevation Dataset Geology CATSDEGEOL 931 categories of surficial geology Land Cover CATESYS Montana land cover framework with roads removed 27 classes Max Temp CONTTMAX Estimated average maximum daily July temperature in degrees Fahrenheit for Min Temp CONTTMIN Estimated average minimum daily January temperature in degrees Fahrenheit for Precipitation CONTPRECIP Relative effective annual precipitation in 1cm intervals as an indicator of available soil moisture. Slope CONTSLOPE Degrees of slope Soil Temp CATSOILTMP Soil temperature and moisture regime 12 categories Stream Dist CONTSTRMED Euclidean distance from major streams in 1 meter intervals

11 Data Summary for Sprague s Pipit Modeling 2012 (2011) modeling efforts 3,472 (2,696) total records 1753 (1,174) spatially unique breeding records 1,057 (722) spatially independent breeding records 793 (541) training points 264 (181) test points

12 Maxent Algorithm Phillips et al Ecological Modelling 190: Phillips and Dudik 2008 Ecography 31: P(x) = exp(c1 * f1(x) + c2 * f2(x) + c3 * f3(x) ) / Z Constrains Empirical Distribution (Positive Observations) Estimated Distribution (Background Points)

13 Maximum Entropy Continuous Model with Training Data

14 Model Evaluation Absolute Validation Index (AVI) for Suitable = 0.96, AVI for Optimal = 0.50 Deviance = - 2 * 1 - Ln(logistic output) Values range from 0 to ~14 Χ (SD) for test data = 2.0 (2.1)

15 Continuous Model and Data Overview Surveys Needed!

16 Delineating Habitat Suitability Classes Non-suitable Low Suitability (Hirzel et al. 2006) Moderate Suitability Optimal Suitability Habitat Suitability Index Class Number of Observations Ratio of Percent of Observations in class/number to Percent of of Pixels Pixels in Classin Class Habitat Suitability Class

17 Suitable & Optimal Habitat Logistic Cutoff Area in Km 2 % of MT Low , Moderate , Optimal ,

18 Percent of Sites Detected North Valley County Long-term Grassland Bird Monitoring for BLM CommonName Change Per Year Signficant Baird's Sparrow Bobolink Brewer's Blackbird Brewer's Sparrow Brown-headed Cowbird Chestnut-collared Longspur Grasshopper Sparrow Horned Lark Lark Bunting Long-billed Curlew Marbled Godwit McCown's Longspur Northern Harrier Savannah Sparrow Sprague's Pipit Upland Sandpiper Vesper Sparrow Western Meadowlark Willet Wilson's Phalarope

19 North Valley County Long-term Grassland Bird Monitoring for BLM Average Number of Birds Detected on Point Count CommonName Change Per Year Signficant Baird's Sparrow Bobolink Brewer's Blackbird Brewer's Sparrow Brown-headed Cowbird Chestnut-collared Longspur Grasshopper Sparrow Horned Lark Lark Bunting Long-billed Curlew Marbled Godwit McCown's Longspur Northern Harrier Savannah Sparrow Sprague's Pipit Upland Sandpiper Vesper Sparrow Western Meadowlark Willet Wilson's Phalarope

20 North Valley County Long-term Grassland Bird Monitoring for BLM Density per acre* CommonName Change Per Year Signficant Baird's Sparrow Bobolink Brewer's Blackbird Brewer's Sparrow Brown-headed Cowbird Chestnut-collared Longspur Grasshopper Sparrow Horned Lark Lark Bunting Long-billed Curlew Marbled Godwit McCown's Longspur Northern Harrier Savannah Sparrow Sprague's Pipit Upland Sandpiper E-05 - Vesper Sparrow Western Meadowlark Average density equivalent to 26.3 per square kilometer Willet Wilson's Phalarope E-05 -

21 Questions? (406)