Analysis Traces

Trace Analysis is used for large data-sets generated by modeling scenarios

The Trace Analysis feature can summarize data-sets with numerous time series traces; data-sets with tens to several hundreds of traces. This is helpful when trying to analyze and define all the possible combinations and/or probabilistic occurrences of certain model outputs to support decision making. This analysis allows users to comparatively and statistically analyze traces/scenarios which usually result from iterative model runs using varying model inputs, assumptions, and conditions.

This analysis allows the user to:

1. Isolate certain traces by input hydrologic condition (wet, dry, or average) or some other threshold and analyze their outputs
2. Determine the per-time-step exceedance probabilities of the entire data-set based on the traces
3. Perform comparative analysis between each of the traces within a data-set

As an example, the graphic below plots the 107 traces of Lake Mead elevations which result from simulating 107 traces of different head-water and hydrologic flow conditions through a water resources model. This particular model generated results similar to what is shown on this graphic for several other modeling nodes defined in the model.

Users may filter and sort their desired traces/scenarios via the scenario selection interface. Using the menu on the lower left corner of the screen below, traces may be sorted by the desired metric. The example below shows traces sorted by the Lake Powell Outflow CY sum for the first year. Run14 on the table shows that it has the largest value for this metric. Selecting only the traces that have high values for this metric allows the user to analyze the impacts to other model outputs that result from high Lake Powell outflows. In short, this is useful in sorting/filtering analysis to just traces that meet a desired condition.

The 2 images below show this in action. The first image is the 10-50-90 exceedance percentiles for Lake Mead elevations using the 107 traces of modeled outputs. The second image shows the same exceedance percentiles after filtering the traces and only using the lowest 20% of traces in terms of Lake Powell outflows.

In Development:

1. Addition of an embedded window that would allow users to dynamically filter traces based on defined thresholds, and to sort traces based on some surrogate (flow volume, value at a specific time-step, etc) - DONE!
2. Canned reports to support Colorado River operations for the USBR LC and UC regions
3. Let us know if there are other products or tools that you think may be useful for this feature