timing of geese Mitra Shariati Najafabadi Skidmore, A.K., - - PowerPoint PPT Presentation

Green wave indices for predicting spring migration timing of geese

Mitra Shariati Najafabadi Skidmore, A.K., Darvishzadeh, R., Kölzsch, A., Vrieling, A., Nolet, B.A., Exo, K.M., Meratnia, N., Havinga, P., Stahl, J., Toxopeus, A.G

• Fertility & Mortality Decrease
• Life Expectancy Increase

 Green wave hypothesis  Arctic nesting geese

• Partially capital breeders
• Timing of optimal breeding condition

 Climate change  An accurate understanding of the timing of the spring migration

INTRODUCTION

1

 Satellite-derived NDVI time series

• Near-infrared (NIR) and red reflectance
• Reasonable estimates of biomass
• Photosynthetically active radiation
• Direct measure of plant phenology

 Growing degree days (GDD)

• Sum of mean daily temperature
• Indirect measure of plant development

INTRODUCTION

2

INTRODUCTION

 Hypothesis: timing of geese migration, with respect to the green wave phenology, would be predicted more accurately by NDVI than GDD

3

Satellite-derived green wave index (GWI)

• MODIS 16-day composite NDVI dataset with a 1-km spatial

resolution

• GWIt= (NDVIt-NDVImin)/ (NDVImax-NDVImin) ×100
• Date of 50% GWI

4

MATERIALS AND METHODS – green wave indices

 The rate of change in temperature acceleration (GDDjerk)

• GLDAS on a fixed grid of 0.25° and at a 3-hour temporal

resolution

• GDD = (TAVG,k − TBASE)
• GDDjerk: the third derivative of the sigmoid function through

the data points that plot the day k against GDDk

• Date of first peak in GDDjerk

5

MATERIALS AND METHODS – green wave indices

MATERIALS AND METHODS – satellite tracking data

 12 female barnacle geese, Branta leucopsis  30 g solar GPS/ARGOS transmitters  Tracked from 2008-2011

6

 Stopover: birds stopped for longer than 48 hours within a radius of 30 km  Breeding site: sites used for 7 to 26 days within a radius of 30 km in the second half of June/ end of June  64 stopover sites and 30 breeding sites

7

MATERIALS AND METHODS – delineation of stopover, and

breeding sites

• Fertility & Mortality Decrease
• Life Expectancy Increase

THE WORLD

5 10 15 20 25 30 % 2009 2050 11.0 22.0

 Linear mixed-effect model  Cross-validation with the leave-one-out procedure  Bland-Altman plot with the 95% limits of agreement

8

MATERIALS AND METHODS – statistical analysis

• Fertility & Mortality Decrease
• Life Expectancy Increase

THE WORLD

5 10 15 20 25 30 % 2009 2050 11.0 22.0

Results – arrival date at the stopover sites

9

Model GWI Fixed effect Parameter ±SE t-value p-value

• 95% CI

+95% CI Intercept 21±9.76 2.15 <0.05 1.49 40.52 GWI 0.83±0.06 11.99 <0.001 0.70 0.98 GDDjerk Fixed effect Parameter ±SE t-value p-value

• 95% CI

+95% CI Intercept 111.94±3.67 19.07 <0.001 103.88 119.44 GDDjerk 0.21±0.02 9.46 <0.001 0.17 0.26 Random effect Variance

χ2

p-value ID 10.9 13.83 <0.01 Year 18.88 7.84 <0.001

• Fertility & Mortality Decrease
• Life Expectancy Increase

THE WORLD

5 10 15 20 25 30 % 2009 2050 11.0 22.0

Results – arrival date at the stopover sites

10

11

Results – arrival date at the stopover sites

• Pearson's correlation analysis between the absolute residuals

values and distance to the breeding site

• GWI model: Pearson correlation coefficient=-0.01, p=0.89
• GDDjerk model: Pearson correlation coefficient=0.32, p<0.01

12

Results – arrival date at the stopover sites

• Fertility & Mortality Decrease
• Life Expectancy Increase

THE WORLD

5 10 15 20 25 30 % 2009 2050 11.0 22.0

Results – arrival date at the stopover sites

13

14

Results – arrival date at the breeding sites

Model GWI Fixed effect Parameter ±SE t-value p-value

• 95% CI

+95% CI Intercept 74.98±11.96 6.26 <0.000 50.36 99.90 GWI 0.50±0.07 6.77 <0.000 0.35 0.66 Random effect Variance

χ2

p-value ID 4.96 5.45 <0.05 Year 7.09 11.64 <0.000 GDDjerk Fixed effect Parameter ±SE t-value p-value

• 95% CI

+95% CI Intercept 103.35±16.37 6.31 <0.000 70.16 137.06 GDDjerk 0.34±0.10 3.21 <0.01 0.12 0.55 Random effect Variance

χ2

p-value ID 8.88 6.35 <0.05 Year 24.78 17.40 <0.000 Model GWI Fixed effect Parameter ±SE t-value p-value

• 95% CI

+95% CI Intercept 74.98±11.96 6.26 <0.000 50.36 99.90 GWI 0.50±0.07 6.77 <0.000 0.35 0.66 Random effect Variance

χ2

p-value ID 4.96 5.45 <0.05 Year 7.09 11.64 <0.000 GDDjerk Fixed effect Parameter ±SE t-value p-value

• 95% CI

+95% CI Intercept 103.35±16.37 6.31 <0.000 70.16 137.06 GDDjerk 0.34±0.10 3.21 <0.01 0.12 0.55 Random effect Variance

χ2

p-value ID 8.88 6.35 <0.05 Year 24.78 17.40 <0.000

15

Results – arrival date at the breeding sites

16

Results – arrival date at the breeding sites

Discussion

17

 GWI is a more reliable index to predict barnacle goose arrival time at both stopover and breeding sites than the GDDjerk  Unlike the GWI model, the GDDjerk model was sensitive to latitude  Difference between the RMSDcv of the two models became smaller in breeding sites  In the high Arctic environment, growing season is short and the plant growth is more rapid in relation to favorable temperatures

In collaboration with The Netherlands Institute of Ecology (NIOO-KNAW) Sovon Dutch Centre for Field Ornithology Institute of Avian Research, “Vogelwarte Helgoland”

19

Thank you for your attention

Vigorous Moderate Light

20