Final Project

An Exploration of the Use of GIS Analyses to Test the Results of a System Dynamics Model

Image of a representative section of the Rio Grande, showing sandbars.

Image from S. Platania showing historic and current ranges of the silvery minnow.

Introduction

The 1994 federal Endangered Species Act (ESA) listing of the silvery minnow in New Mexico has brought with it legal obligations to ensure the species’ survival and to protect the well being of individuals insofar as possible. These include a recovery plan and critical habitat designation. The consequences for failure to do so are both criminal and financial, and can apply to individuals and to the state as an entity.

Silvery minnow, an endangered species native to the Rio Grande and Pecos rivers of New Mexico.

The Model

In an effort to understand the silvery minnow, Sandia National Laboratories (SNL) is developing a model of the fish in its native habitat. The model, built using the Powersim system dynamics modeling platform, is shown below. The main purpose of the minnow model is to determine the vulnerability of the silvery minnow to a variety of threats, specifically high and low discharge conditions and water pollution, proxied by agricultural and wastewater treatment plant ammonia concentrations.

The model runs for 31 years (1975-2005) on a monthly time step. Fish live for 2-4 years. Cohorts are eggs and larvae, juveniles, and adults, with spawning beginning at 5 months. Six reaches are modeled: Cochiti – San Felipe – Albuquerque – Bernardo – San Acacia – San Marcial (see image below).

Fish populations change according to births, deaths, and on/off switchable captive release program and migration scenarios. There are several types of mortality – background, density dependent, discharge, and pollution. Results consist of time series by reach of output variables.

Data can be input in a number of ways. Some data is input from Excel spreadsheets as time series or as single values. Other inputs can be selected as graphs, e. g., logistic, exponential, and linear. And still other data is selectable using a slider to choose a value from a range of values. This multitude of input data combinations makes it difficult or impossible to test the model and find sets of optimal values.

The primary difficulty has been in data collection. Some of the variables had time series data for the last 31 years, but for other variables it was spotty. And for several variables, there simply was no information and so they had to be estimated. This reduces confidence in any results, and requires extensive sensitivity testing. However, the complexities, non-linearities, and positive and negative feedbacks in the model make any single combination of input variables of marginal value and possibly non-replicable in the outside world. Thus, independent verification of values or at least trends would be invaluable.

GIS Analyses

ArcGIS, a product of ESRI, is a sophisticated menu-driven tool for investigating data with a spatial or pseudo-spatial component. The images used in this analysis, downloaded from RGIS, were 7.5 minute MrSID orthophoto quarter quads, 1-m resolution, for southwest Albuquerque and the northwest quarter quad for San Acacia (images below, with Albuquerque on the left). However, although the flights were flown between 1996 and 1998, the metadata did not say when in the year the flights were or even whether they were at the same time of the year.

Because the adult silvery minnow likes deep pools and slow-moving water, two ratios were selected as possible indicators of hospitable conditions for each of the two reaches: total area of sandbars divided by the total chosen area of river and length of thalwag (deepest part of the river) divided by the length of the bank. A representative section of each quarter quad was chosen for the analysis. However, the resolution of the image was not good enough to ensure determining all - and only – the sandbars. Further, the deepest part of the river was not discernable. In the San Acacia representative section, it was difficult to determine what was water and what was sand, which also made it impossible to delineate the river bank. See the images below for representative sections, with Albuquerque on the left. Thus, it was not possible to determine the ratios for either reach’s representative section, and so not possible to draw any conclusions about the favorability of the reach for the silvery minnow.

Conclusions

After the ArcGIS analyses, it became apparent that the technique would not be suitable to test the minnow model even if the images had been able to yield a result. This system dynamics model is temporally discrete and spatially aggregated and yields time series output, the exact opposite of GIS data, which is spatially discrete but a snapshot in time. Perhaps if the model reaches were smaller and more homogeneous, and with a yearly time step, a GIS image of better resolution (Lidar?) would be able to discern the relevant features so that ratios could be determined and an indication of reach favorability, and perhaps a comparison between the two reaches’ favorabilities, could be made. However, this model is not that model, so the technique is inappropriate for testing silvery minnow model results.