Statistical guidelines for sampling Statistical guidelines for - - PowerPoint PPT Presentation

Statistical guidelines for sampling marine avian populations Statistical guidelines for sampling marine avian populations

Elise F. Zipkin Brian Kinlan Allison Sussman Mark Wimer Allan F. O’Connell Elise F. Zipkin Brian Kinlan Allison Sussman Mark Wimer Allan F. O’Connell

USGS Patuxent Wildlife Research Center NOAA National Ocean Service

4th International Wildlife Management Conference – July 2012

Seabirds in the Atlantic

Where are the birds?

Not a lot known about the distribution and abundances in the Atlantic

• Difficult to survey
• Rough conditions
• Patchily distributed
• Highly mobile

Off shore wind power garnering lots of interest

• Many states have

implemented a 20% renewable energy by 2020 mandate

• Public perception of oil

spills is poor

Where are the birds?

Wind development

U.S. Bureau of Ocean and Energy Management (BOEM)

• 5km x 5km

lease blocks

• Along the

Outer Continental Shelf of the Atlantic Ocean

All Lease Blocks

Patuxent Wildlife Research Center

Objectives

Develop a framework for assessing: 1) which lease blocks are “hot spots” and “cold spots” 2) the required surveying effort to guide BOEM and industry in determining wind turbine placement

What is a hot/cold spot?

Hot spot = A lease block with an average species specific abundance that is three times the mean of the region Cold spot = A lease block with an average species specific abundance that is one third the mean of the region

The Atlantic Seabird Compendium

• >250,000 seabird observations from U.S.

Atlantic waters

• Collected from 1978 through 2011
• Data collected using a mix of methods

including non‐scientific approaches

The Atlantic Seabird Compendium

• >250,000 seabird observations from U.S.

Atlantic waters

• Collected from 1978 through 2011
• Data collected using a mix of methods

including non‐scientific approaches

We used:

• 32 scientific data sets – 28 ship‐based, 4 aerial
• Transects were standardized to 4.63km
• 44,176 survey transects representing 463 species

Two part approach

1) Determine the best statistical distribution to model the count data for each species in each season 2) Use the best fitting distribution to produce power analyses

The rest of the talk

1) Describe the broad two part approach 2) Integrate an example using Northern Gannets

Two part approach

1) Determine the best statistical distribution to model the count data for each species in each season 2) Use the best fitting distribution to produce power analyses

Part 1: Model the data

Test eight statistical distributions:

Poisson Negative binomial Geometric Logarithmic Discretized lognormal Zeta decay Yule Zeta (power law) Northern Gannet spring count data

Examples of the distributions

1 2 5 10 20 1e-05 1e-03 1e-01

Positive Poisson (simulated)

1 2 5 10 20 50 100 200 1e-05 1e-03 1e-01

Positive neg binomial (simulated)

1 2 5 10 20 50 100 1e-05 1e-03 1e-01

Positive geometric (simulated)

1 2 5 10 20 50 100 200 1e-05 1e-03 1e-01

Logarithmic (simulated)

1 5 10 50 100 500 1000 1e-05 1e-03 1e-01

Discretized lognormal (simulated)

1 100 10000 1e-05 1e-03 1e-01

Zeta (simulated)

1 100 10000 1e-05 1e-03 1e-01

Yule (simulated)

Part 1: Results

Spring Summer Fall Winter Total

Number species with >500 observations 12 10 15 11 48

Part 1: Results

Spring Summer Fall Winter Total

Number species with >500 observations 12 10 15 11 48 Discretized lognormal Yule Negative binomial Logarithmic Zeta decay

Part 1: Results

Spring Summer Fall Winter Total

Number species with >500 observations 12 10 15 11 48 Discretized lognormal 7 (4*) 4 (3*) 8 (3*) 8 (2*) 27 (12*) Yule 1* 3* 1* 1 1 (5*) Negative binomial Logarithmic Zeta decay 3* 0 (3*)

Part 1: Results

Northern Gannet

Discretized lognormal top distribution for fall and spring Discretized lognormal and Yule fit equally well in winter and summer

1 5 10 50 500 1e-04 1e-03 1e-02 1e-01 1e+00

Count (log scale) Probability (log scale)

Discretized lognormal Yule Zeta decay Zeta

Two part approach

1) Determine the best statistical distribution to model the count data for each species in each season 2) Use the best fitting distribution to produce power analyses

Part 2: Power analysis

Part 2: Power analysis for Northern gannets in the spring

*Focusing only on lease blocks where individuals were observed

Part 2: Northern gannet results

Reference mean = 6.9 individuals per lease block conditional on presence

5 10 15 20 25 0.0 0.2 0.4 0.6 0.8 1.0

Number of sampling events Simulated power

Hot spot (3 x mean) Cold spot (0.33 x mean)

Part 2: Northern gannet results

Reference mean = 6.9 individuals per lease block conditional on presence

5 10 15 20 25 0.0 0.2 0.4 0.6 0.8 1.0

Number of sampling events Simulated power

Hot spot (3 x mean) Cold spot (0.33 x mean)

Part 2: Northern gannet results

1 5 10 50 100 500 1000 1e-05 1e-03 1e-01

Counts Frequency

Discretized lognormal

Reference mean = 6.9 individuals per lease block conditional on presence

Part 2: Northern gannet results

Summary of results

• Seabirds tend to be highly aggregated and

require skewed statistical distributions to accurately describe populations

• For many species, we

need a large number of surveys to detect areas with atypical abundances

Implications for wind power

• Intensive sampling in multiple seasons

will be required to determine potential impacts on seabirds

• A possible approach could be to combine

data on functionally similar species or species of high conservation value

Acknowledgments

• The many researchers and their crews who

collected the data used in our analyses

• Emily Silverman, Diana Rypkema
• The Bureau of Ocean, Energy,

Management (BOEM) for funding model development and analysis