VanDantzig

Introduction

This document contains the code used to generate the analysis found in Oddo et al. (2017), Deep Uncertainties in Sea-Level Rise and Storm Surge Projections: Implications for Coastal Flood Risk Management.

Full Citation:

Oddo, P. C., Lee, B. S., Garner, G. G., Srikrishnan, V., Reed, P. M., Forest, C. E. and Keller, K. (2017), Deep Uncertainties in Sea-Level Rise and Storm Surge Projections: Implications for Coastal Flood Risk Management. Risk Analysis. doi:10.1111/risa.12888. (View Online)

Analysis Overview

In this analysis we examine the effects of model and parameter uncertainties underlying the decision-analytical model described in Van Dantzig (1956):

Van Dantzig, David. "Economic decision problems for flood prevention." Econometrica: Journal of the Econometric Society (1956): 276-287.

The Van Dantzig (1956) model uses a cost-benefit framework to determine the optimal dike heightening for a single polder. In this model, the costs of investing into flood protection are weighed against the expected damages from a flood event, given a set of economic and geophysical parameters:

Parameter	Value	Description
p_zero_p	0.0038	Initial flood frequency (1/yr) with zero height increase
alpha_p	2.6	Exponential flood frequency constant
V_p	1e+10 * 2	Value of goods protected by dike (includes "factor of 2" for cost of human life)
delta_prime_p	0.02	Discount rate (percent/year)
k_p	4.2e7	Cost of heightening dikes by 1 meter (Guilders)
subs_rate	0.002	Rate of land subsidence (meter/year)
sea_level_rate	0.008	Rate of sea level rise (meter/year)

Our analysis is evaluates the effects of increasing parametric and structural model uncertainties using the Van Dantzig (1956) model as a baseline. The effects of these uncertainties are evaluated for four decision objectives (Minimize Investment Costs, Maximize Reliability, Minimize Expected Damages, Minimize Total Costs). Each component of the analysis is contained in the following folders:

1. Baseline_Model
2. Parametric_Uncertainty
3. Uncertainty_SLR
4. Uncertainty_SLR_GEV

Model Versions

• The Baseline_Model recreates the original anylsis described in Van Dantzig (1956), using the best-guess parameter values provided in Section 6: "The Doubtful Constants."
• The Parameteric_Uncertainty version uses the same structure as the baseline, but introduces uncertainty in the parameter values.
• The Uncertainty_SLR version introduces uncertainty in the model structure but implementing an updated sea-level rise model.
• The Uncertainty_SLR_GEV version introduces storm surge projections generated using a generalized extreme value (GEV) analysis.

Running the Model

Each iteration of the model builds of the version preceding it, beginning with the vanDantzig_baseline.R file in the Baseline_Model folder. For example:

• Parametric_Uncertainty > vanDantzig_Uncertainty.R
• Uncertainty_SLR > vanDantzig_SLR.R
• Uncertainty_SLR_GEV > vanDantzig_SLR_GEV.R

For each model version, the following diagnostic plots are produced:

1. 'Optimal' dike heightening compared to baseline model
2. Pairwise tradeoff curves for the four decision objectives
3. Pairwise scatter plots for model parameters
4. Histograms for model parameters

Sensitivity Analyses

Each version of the model also produces output for two types of sensitivity analysis, local (One-at-a-time (OAT)), and global (Sobol'). Instructions for running these analysis and producing associated figures are either in the scripts or in additional READMEs.

Links to Paper Figures

• Figure 1
• Figure 2
• Figure 3
• Figure 4
• Figure 5
• Figure 6
• Figure 7

R Packages

• zoo
• lubridate
• lhs
• extrafont
• fExtremes
• extRemes
• timeSeries
• compiler
• fields
• KernSmooth
• RColorBrewer
• MASS
• sfsmisc