1 of 57

Ecological datasets: �using data to understand drivers of change in social-ecological systems

​

Katherine Siegel

University of Colorado, Boulder

ksiegel@ucar.edu

​

Working with Environmental Data

November 3, 2022

​

2 of 57

3 of 57

GUIDING RESEARCH QUESTION:

How can we manage the social-ecological systems of working landscapes to conserve biodiversity and support equitable human well-being?

3

ABOUT ME

4 of 57

BS, Environmental Sciences

4

ABOUT ME

MS, Biodiversity,

Conservation,

& Management

5 of 57

BS, Environmental Sciences

5

ABOUT ME

MS, Biodiversity,

Conservation,

& Management

Siegel et al, 2019

6 of 57

PhD, Environmental Science, Policy, & Management

6

ABOUT ME

7 of 57

NOAA Climate & Global Change Postdoctoral Fellow,

Department of Ecology & Evolutionary Biology, University of Colorado-Boulder

7

ABOUT ME

8 of 57

8

RESEARCH THEMES

Land management

& wildfires

Global environmental change & ecological transformations

Interdisciplinary conservation science

Sustainability in marine social-ecological systems

9 of 57

9

RESEARCH THEMES

Land management

& wildfires

Global environmental change & ecological transformations

Interdisciplinary conservation science

Sustainability in marine social-ecological systems

10 of 57

Laurel Larsen,

UC Berkeley

Connor Stephens,

University of Wisconsin

Van Butsic,

UC Berkeley

Bill Stewart,

UC Berkeley

11 of 57

11

12 of 57

12

13 of 57

13

Resource Systems

Forests of the western US

Resource Units

Harvested species, etc

Governance Systems

Forest management policies and institutions

Actors

Federal agencies,

private landowners

Focal Action Situations

Wildfires

Social, Economic, & Political Settings

Climate Change

Land management

14 of 57

14

Data from Sass et al., 2020,

USFS

15 of 57

Federal vs. private forests

  • Different management objectives
  • Different levels of biomass removal
  • Diverging approaches to fire suppression

15

Wikipedia

OPB

AP

Audubon

16 of 57

  1. Is there a difference in the probability of wildfire in federally managed forests as opposed to privately-owned, unprotected forests in the western US?

​

  • If so, has this difference changed over the past three decades?

​

  • Do this difference and any potential trends over time vary by state?

​

  • How does the magnitude of this difference compare to the effects of climate variables in modifying wildfire probability?

16

17 of 57

CAUSAL INFERENCE

17

18 of 57

AN ECONOMETRIC TOOLKIT FOR CAUSAL INFERENCE IN SOCIAL-ECOLOGICAL SYSTEMS

18

Panel data

​

Pre-regression matching

​

Mixed effects regression models

19 of 57

PANEL DATA

19

Sample point 1

Sample point 2

Sample point 3

Sample point 4

Sample point 5

Sample point 6

GRAZING

NO GRAZING

20 of 57

PRE-REGRESSION MATCHING

20

NO GRAZING

GRAZING

21 of 57

PRE-REGRESSION MATCHING

21

NO GRAZING

GRAZING

22 of 57

22

MIXED EFFECTS REGRESSION MODELS

NO GRAZING

GRAZING

23 of 57

23

Panel data

​

Pre-regression matching

​

Mixed effects models

24 of 57

Response variable: whether or not the site was located within a fire perimeter in each year (1989-2016)

Predictors/covariates:

  • Management type
  • Human population density
  • Geographic variables
    • Topography
    • Distance to roads
    • Latitude and longitude

  • Recent fires
  • Climate variables
    • Wind speeds
    • Maximum and minimum temperatures
    • Total precipitation
    • Lightning strikes

24

Panel data

​

Pre-regression matching

​

Mixed effects models

25 of 57

DATA SOURCES

25

Variable

Data source

Original data format

Software used to prep data

Land ownership

USGS Protected Area Database

Shapefile

R

Land cover type

National Land Cover Database (USGS); Landfire

Raster

R

Topography (elevation, slope, and aspect)

National Elevation Dataset (USGS)

Raster

QGIS, R

Seasonal climate variables

TerraClimate

Raster

Google Earth Engine, R

Recent fire history

Monitoring Trends in Burn Severity

Shapefile

R

Human population density; distance to roads

US Census Data

Shapefile

R

Lightning strikes

NOAA

Table

R

26 of 57

STEPS FOR FINDING & PREPPING DATA

  1. Lit review to identify key covariates and data sources that others have used
  2. Locate publicly available data sources
  3. Download data (or data summaries, in the case of TerraClimate, since I didn’t want to download >8000 rasters of the entire western US)
  4. Generate sampling grid (R)
  5. Extract raster and shapefile values for each covariate layer at each sample point (R)
  6. Calculate lag variables (R)

26

27 of 57

27

Wikimedia

NYTimes

Panel data

​

Pre-regression matching

​

Mixed effects models

28 of 57

28

Private, unprotected sample points

Federal

sample points

Elevation

Distance to roads

Full dataset

Matched dataset

Panel data

​

Pre-regression matching

​

Mixed effects models

29 of 57

29

Burnit = B0 + B1*Management + B2*Year +

B3*State + B4-6*Lag Variables +

B7-18*Covariates +

B19-22*Interactions + ui + eit

�

Panel data

​

Pre-regression matching

​

Mixed effects models

30 of 57

30

Burnit = B0 + B1*Management + B2*Year +

B3*State + B4-6*Lag Variables +

B7-18*Covariates +

B19-22*Interactions + ui + eit

�

Panel data

​

Pre-regression matching

​

Mixed effects models

31 of 57

Interactions:

  • Management * State
  • Management * Year
  • State * Year
  • Management * State * Year

31

Burnit = B0 + B1*Management + B2*Year +

B3*State + B4-6*Lag Variables +

B7-18*Covariates +

B19-22*Interactions + ui + eit

�

Panel data

​

Pre-regression matching

​

Mixed effects models

32 of 57

32

Siegel et al. 2022,

Regional Environmental Change

33 of 57

33

National Archives

John Marshall

34 of 57

34

OPB

Sikorsky

35 of 57

35

Halofsky et al. 2018,

Ecosphere

36 of 57

36

Siegel et al. 2022,

Regional Environmental Change

37 of 57

37

Siegel et al. 2022,

Regional Environmental Change

38 of 57

TAKEAWAYS

KEY FINDING

In this social-ecological system, management can have as large an impact as some aspects of climate change

38

BIODIVERSITY SOLUTIONS

Projections of the potential future impacts of changing fire regimes on biodiversity must account for both climate change and management variables

METHODOLOGICAL CONTRIBUTION

This methodological framework can be used for quantitatively comparing drivers of change in complex social-ecological systems

39 of 57

  1. Quantifying the effect of livestock grazing on wildfire risk in California rangelands
  2. Integrating quantitative and qualitative data and methods to model land use change in an Amazonian protected area
  3. Understanding the relationship between land management, wildfire severity, and post-fire ecological functioning and ecosystem services in the western US

39

OTHER RECENT PROJECTS REQUIRING DATA SYNTHESIS

40 of 57

40

Impacts of livestock grazing on wildfire probability

41 of 57

41

Impacts of livestock grazing on wildfire probability

Variable

Data source

Original data format

Software used to prep data

Grazing levels; property boundaries

Survey of land owners

Survey responses

R

Land cover type

National Land Cover Database (USGS)

Raster

R

Topography (elevation, slope, and aspect)

National Elevation Dataset (USGS)

Raster

QGIS, R

Seasonal climate variables

TerraClimate

Raster

Google Earth Engine, R

Recent fire history

Monitoring Trends in Burn Severity

Shapefile

R

Human population density; distance to roads

US Census Data

Shapefile

R

Lightning strikes

NOAA

Table

R

Net primary productivity

MODIS NPP

Raster

Google Earth Engine, R

42 of 57

42

Land use change in an Amazonian protected area

43 of 57

43

Land use change in an Amazonian protected area

Kinnebrew,…, Siegel 2020, Conservation Biology

44 of 57

44

Land use change in an Amazonian protected area

Siegel et al. 2022, Conservation Biology

45 of 57

45

Land use change in an Amazonian protected area

Siegel et al. 2022, Conservation Biology

46 of 57

46

Land use change in an Amazonian protected area

Siegel et al. 2022, Conservation Biology

47 of 57

47

Land use change in an Amazonian protected area

Siegel et al. 2022, Conservation Biology

48 of 57

48

TNC

Land management, wildfire severity, & post-fire ecological trajectories

49 of 57

49

TNC

Land management, wildfire severity, & post-fire ecological trajectories

50 of 57

50

Data: Forest Inventory & Analysis data (.csv)

51 of 57

51

Matched across ownership

Data: FIA (.csv) and PAD-US (.shp)

52 of 57

52

Matched across ownership

Two-step hurdle model

Data: Monitoring Trends in Burn Severity (raster)

53 of 57

53

Matched across ownership

Two-step hurdle model

Logistic regression

54 of 57

54

Logistic regression

Functional trait analysis

Matched across ownership

Two-step hurdle model

Data: trait data assembled from TRY Plant Trait Database, published papers, gray literature, combined with FIA data

55 of 57

55

Functional trait analysis

55

Logistic regression

Ecosystem service modeling

Matched across ownership

Two-step hurdle model

56 of 57

56

Ecosystem service modeling

Overlay with social vulnerability index

Matched across ownership

Two-step hurdle model

Functional trait analysis

Logistic regression

Data: US Census (.csv)

57 of 57

57

Acknowledgements

Get in touch!

ksiegel@ucar.edu

@kjohannetsiegel