Copula modeling is a second-order statistical analysis that builds upon several prerequisite analyses. The initial release of Copula Analysis in Neptune enables users to fit six copula families to observed data, compare model performance using both a quantitative metric (Akaike Information Criterion, AIC) and graphical diagnostics, select one or more appropriate copula models, and visualize the resulting dependence structure. The HEC Statistical Hydrology Team is actively developing a more streamlined, end-to-end workflow for future Neptune releases.
Before creating a Copula Analysis in Neptune, users should complete the following prerequisite steps:
Step
Task
Suggested Tools
1
Compile time series data
HEC-SSP or HEC-Neptune
2
Extract extreme events (annual maximum series or partial duration series) from each variable
HEC-SSP Data Filtering and/or Peaks Over Threshold Analysis
3
Determine ordinal association between paired events
R
A sample R script that uses the MultiHazard package to compute ordinal association between two time series over a range of lags is available here.
4
Fit marginal probability distributions to the resulting samples
HEC-SSP Distribution Fitting Analysis and/or Peaks Over Threshold Analysis