Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services Biophysical...

25
Title Date 1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services www.wavespartnership.org Biophysical modeling of ecosystem services: Module 2D: Building Causal Models WAVES Training Module

Transcript of Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services Biophysical...

Page 1: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 1WAVES © 2014

Wealth Accounting and the Valuation of Ecosystem Services www.wavespartnership.org

Biophysical modeling of ecosystem services:Module 2D: Building Causal Models

WAVES Training Module

Page 2: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 2WAVES © 2014

Building causal models

Page 3: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 3WAVES © 2014

Types of causal models

1. Binary lookup tables

2. Qualitative lookup tables

3. Aggregated statistics lookup tables

4. Multiple layer lookup tables

5. Causal relationships

6. Spatial interpolation

7. Environmental regression models

Schröter, M., et al. In press. Lessons learned for spatial modeling of ecosystem services in support of ecosystem accounting. Forthcoming in: Ecosystem Services.

Page 4: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 4WAVES © 2014

Methods for mapping ecosystem services

Schröter et al. in press

Page 5: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 5WAVES © 2014

Causal relationships

Causal relationships (i.e., ecological production functions) can be used to model ecosystem services

Based on relationship between input data and predicted, quantified output

A

B C D?

Hypothesized relationships between three environmental variables (B-D) to predict ecosystem service (A)

Page 6: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 6WAVES © 2014

Spatial interpolation

Predict ecosystem services based on spatial autocorrelation of measured data points, sometimes using additional environmental layers

Kriging, Inverse Distance Weighting (GIS methods; Sumarga and Hein 2014)

Interpolated biodiversity data using ArcGIS: Inverse distance weighting (left); Empirical Bayesian Kriging (right)

Page 7: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 7WAVES © 2014

Environmental regression models

Use environmental layers as independent variables to predict ecosystem service values

Maximum entropy modeling (MaxEnt) software, e.g., Sherrouse et al. 2014, Sumarga and Hein 2014

Page 8: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 8WAVES © 2014

Steps to develop causal models

Page 9: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 9WAVES © 2014

Step 1: Define model elements/variablesWhich variables matter?

Which are most important?

Importance of scientific consensus, theory (or not?)

A

B

C

D

E

Page 10: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 10WAVES © 2014

Step 2: Build the conceptual model

Define the model structure:

What’s the relationship between the variables?

How do values of input variables influence outputs?

Data & math come later

A

B C D?

Hypothesized relationships between three environmental variables (B-D) to predict ecosystem service (A)

Page 11: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 11WAVES © 2014

Step 3: Collect & prepare data to parameterize model

Collect, clean, and otherwise prepare input data

Develop and document model assumptions and proxy data

A model is only as good as your data and your assumptions about model structure

“Garbage in, garbage out”

Page 12: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 12WAVES © 2014

Step 4: Testing, calibration & validation, sensitivity analysis

Calibration: Compare results of your model runs to an existing dataset

Validation: Set aside part of your dataset, develop & run the model on your remaining data, then go back and see how the model performs using the data you held back during model development

Sensitivity analysis: Test to see which variables have the biggest influence on the model outputs

Page 13: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 13WAVES © 2014

Probabilistic vs. deterministic modeling approaches

Probabilistic Explanatory power (e.g., r2) Explanation whyBased on inductive reasoning

Deterministic/mechanistic Explanation why Explanatory power (e.g., r2)Based on deductive reasoning

Good for:• Exploring patterns• Seeing if real-world patterns conform to theory• Incomplete datasets or situations with high uncertainty

Good for:• Testing/understanding why something works the way it does• When you have a strong understanding of how something works

Dangers (among others):• Putting too much faith into patterns found in the data that lack a reasonable theoretical foundation

Dangers (among others):•Sloppy model construction

Page 14: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 14WAVES © 2014

Calculating & communicating uncertainty

The same input data and equations will produce the same results every time unless something changes

Change input parameters, use stochastic inputs, and run repeatedly to generate a distribution of results (Monte Carlo simulation)

Some probabilistic models have built-in uncertainty estimates (e.g., Bayesian models, see Vigerstol & Aukema 2011)

INSERT UNCERTAINTY MAP FROM ARIES

Page 15: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 15WAVES © 2014

Uncertainty in decision making

Clear communication of uncertainty has been challenging to achieve (Ruckelshaus et al. in press)

People generally tend to:

Prefer a sure thing over a gamble (be risk averse) when outcomes are good (gains)

Reject the sure thing and accept the gamble (be risk-seeking) when choosing between multiple negative outcomes (Kahneman, D. 2011. Thinking, fast and slow. Farr, Stras, and Giroux: New York)

Page 16: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 16WAVES © 2014

Scaling up ecosystem services

Page 17: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 17WAVES © 2014

(can add material here if we need to cover scaling up of modeling results)

Page 18: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 18WAVES © 2014

Quantifying and mapping cultural ecosystem services

Page 19: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 19WAVES © 2014

Mixed methods modeling“One model

fits all” an unrealistic paradigm (Vigerstol et al. 2011, Bagstad et al. in press)

Deterministic models

(ARIES, InVEST, other process-based models): good data availability, systems

well-understood

PPGIS/Social values mapping

(incl. SolVES): cultural ecosystem services & non-use

values

Probabilistic approaches (ARIES, other

Bayesian/Monte Carlo approaches)

weaker data availability & systems knowledge, benefit of

carrying explicit uncertainty

Page 20: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 20WAVES © 2014

Cultural ecosystem service mapping

AestheticBiodiversityCulturalEconomicFutureHistoricIntrinsicLearningLife SustainingRecreationSpiritualTherapeutic

Page 21: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 21WAVES © 2014

Cultural ecosystem services surveysCan survey:

Recreational visitors

Residents

Focus groups

Using:

In-person

Mail surveys

Internet-based mapping

Mapping:

Points

Polygons

Page 22: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 22WAVES © 2014

Mapping cultural ecosystem services

MaxEnt or Kernel Density methods – extrapolate values across the landscape using relationship between points and underlying biophysical environment (land cover, landforms, elevation, slope, distance to roads/water/shoreline/infrastructure) – a spatial interpolation method

Social Values for Ecosystem Services – GIS tool (Sherrouse et al. 2014)

Extensive “Public Participatory GIS” work (Greg Brown and colleagues; http://www.landscapevalues.org

Page 23: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 23WAVES © 2014

Mapping cultural ecosystem services

Brown & Brabyn 2012

Page 24: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 24WAVES © 2014

Exercise 5: Building causal models

1. Starting with the ecosystem services and data sources identified in Exercise 1, build a causal model to quantify and map capacity for one ecosystem service.

1. What final output/metric do you want to use for the ecosystem service?

2. What input datasets could you use to help quantify that service, and to calibrate the model?

3. Where are you more and less certain about the elements included in the model, and its overall structure?

4. Who else would you want to review your causal model before moving ahead to test and refine it?

2. Each group presents their causal model to the full group for discussion.

Page 25: Title Date1 WAVES © 2014 Wealth Accounting and the Valuation of Ecosystem Services  Biophysical modeling of ecosystem services:

Title Date 25WAVES © 2014

Could add slides based on Section 2.3.3 of Hein 2014 “Guidance for the biophysical mapping and analysis of ecosystem services in an ecosystem accounting context”

Add a few slides at the end of Level 2 training re: where can you get level 3 training. Also need a few general slides on advanced modeling topics – ABM, Monte Carlo simulation & uncertainty (see if Ioannins has good overview papers on env. Modeling)