CE 531
Group: Clint Merrell, Chris Hoggan, and Kyle Robe
Project 1
Zona Sur in Zacatecas Mexico
Introduction
Our group project is located in a watershed to the southeast side of the city of Zacatecas, Mexico. We received our DEM data, soil type, and land use from our counterparts that were also working on this project in Mexico. We made a site visit to this area and learned that there were problems in this area with flooding. In an attempt to remediate the flooding, the city built two large collectors, one for storm water drainage and one for wastewater. While in Mexico, we had little time to do a proper hydrologic analysis of the watershed and were forced to run only a simple simulation using the SCS method for loss and transformation. We decided that it would be interesting to model the Zona Sur basin using some of the other models we have learned thus far and compare them with the HMS SCS method model.
Model Development
Our goal was to develop 5 separate models but to make the parameters for these models as similar as possible. By using the Clark unit hydrograph as the transformation method in the HMS models, we hoped to be able to compare the results with the MODClark model. Ideally, we should get similar results between the models. A summary of the methods used in each model us shown below in Table.1
Table 1. Methods Used in Each Model
ModelMethod / 1 (SCS method) / 2 (Green and Ampt) / 3 (MODClark) / 4 (GSSHA no Streams) / 5 (GSSHA w/ Streams)
Transformation / Clark / Clark / MODClark / Physical equations / Physical Equations
Loss / SCS CN / Green and Ampt / Gridded SCS CN / Green and Ampt / Green and Ampt
After visiting the site in Mexico, we realized that the soil is predominantly fine grained in the watershed and thus has relatively low permeability. In determining the curve number for the basin, the basin was designated as more than 90% hydrologic soil type C, with the rest being soil type D. This resulted in a curve number of around 80 when combined with the land use information. For the Green and Ampt parameters in HMS, values for sandy clay on pg 43 of the HMS reference manual were used. To determine the moisture deficit, we assumed the soil was 30% saturated before the precipitation event. For the Green and Ampt parameters in the GSSHA models, values for sandy clay were taken from pg 94 of the GSSHA reference manual. In addition to fine grained soil, approximately 30% of the basin consists of urban land use.
One problem we had in developing the model was that the DEM resolution was not fine enough to accurately designate the stream channel near the outlet of the watershed. This was a result of the flat topography near the outlet as well. In order to model the outlet in the correct location, using the aerial photographs, we created a stream arc where the channel was supposed to be, and then offset all of the elevations along that arc 10 meters downward. When we ran TOPAZ again, the flow accumulations were in the correct location. In the GSSHA model with streams, the streams were all defined as trapezoidal and the cross sectional measurements were estimated from our field visits. The trunk of the drainage network had a bottom width of 3m, a depth of 2m, and a side slope of 0.5:1. A roughness of 0.02 was assigned to all stream channels. None of the other models considered river routing or stream geometry.
The meteorological parameters were determined using intensity-duration-frequency tables developed by the students in Zacatecas. For lack of reliable data, only one climatological station was used. The total rainfall depth of 1.63 inches for a 20-yr, 24-hr duration storm was used to create a 3-hour storm in an attempt to better represent a typical storm in that area. The 3-hour distribution is shown below in Figure 1.
Figure 1. Incremental distribution of the 3-hour design storm.
Results and Discussion
Hydrographs were obtained from all five simulations and were placed on the same set of axes. The hydrographs can be seen below in Figure 2.
Figure 2. Hydrographs from all 5 simulation runs.
Table 2 below gives a summary of the peak flow, precipitation depth, loss volume, and excess volume. The volumes of precipitation, loss and excess for the GSSHA model were determined by hand, using the output volume given in cubic meters. The volume of precipitation should be the same in the GSSHA models as in the HMS models but ends up slightly less, probably due to the numerical methods utilized in GSSHA.
Table 2. Summary of Results
HMS SCS CN / HMS Green and Ampt / HMS MODClark / GSSHA w/ streams / GSSHA no StreamsPeak Flow (cfs) / 1161.3 / 783.8 / 533.4 / 4967.6 / 489.8
Total Precip (in.) / 1.63 / 1.63 / 1 / 1.56 / 1.56
Total Loss (in.) / 1.16 / 1.32 / 0.82 / 0.92 / 1.17
Total Excess (in.) / 0.47 / 0.28 / 0.18 / 0.64 / 0.40
Total Excess (Ac-ft) / 431.8 / 258.1 / 165 / 580.6 / 363.9
From Figure 2 and Table 2 we can make several interesting observations. We predicted that the distributive nature of the GSSHA models would lead to the highest infiltration values. This prediction was wrong as the HMS Green and Ampt model had the highest infiltration volume. This disparity could be due to different soil properties listed for HMS and GSSHA or to differences in the method by which HMS and GSSHA calculate infiltration values.
Although the peaks and volumes are different between the HMS models, the temporal distribution of the outflow hydrographs are very similar. This was expected because Clark unit hydrograph transformation method was used for all three models. The GSSHA model distributions are different because rather than using a transformation method, flow was calculated from cell to cell individually until the runoff reached the outlet.
Although comparisons between the GSSHA models and HMS models are hard to make and may or may not have any real meaning, we can note the differences between the two GSSHA models. The GSSHA Model without stream flow simulation shows a much more attenuated, lagged and distributed outflow hydrograph compared to the model with stream flow simulated. Also, without streams, much more infiltration occurred. Without streams, runoff is modeled flowing from cell to cell from one end of the basin to the other. With streams, the water is channelized and moves much more quickly to the outlet. The faster travel time with streams leads to a faster peak and less time for the water to infiltrate. This is a significant difference as the peak outflow of the stream model is about 100 times higher than the no stream model.
The results for the MODClark method are not reliable as problems were encountered trying to use a 3-hour hyetograph. The most precipitation we could get simulated in the MODClark model was 1 inch. Dr. Nelson was unable to determine the cause of this error as well.
Conclusions
By running these 5 different models, we were able to understand that we really don’t understand anything. That is, we don’t understand what a reasonable result for this watershed is. With a total lack of observed data, we really cannot determine how to best model the response of the Zona Sur basin. Many things in this analysis are approximated and it is difficult to make a good approximation, or to fine tune an approximation without some firsthand knowledge of the response of the watershed. For example, if we had an idea of what depth occurs in the stream channels during a typical storm, using measured or surveyed cross sections of the stream channels, we could use Manning’s equation to determine what the resulting flow rate would be, and then compare this with the flow rates determined by our model. A more thorough survey of the soil type in the watershed would help. Instead of distributing the Green and Ampt Parameters based on soil type alone, land use should probably be considered as well, because urban areas in the Green and Ampt models are basically unaccounted for. From this study, the need for model calibration is made quite apparent.