A Study of Precipitation and Stream Flow for the
Upper Sevier Hydrologic Basin
 
 

A term project prepared by
Bryan Kimball in conjunction with an assignment for
CEE 5440 : Water Resources GIS
Utah State University
Last Updated December 5, 2002


 

    Introduction
 
As part of the requirements of the Water Resource GIS class taught by Dr. Tarboton of Utah State University and Dr. Maidment of University of Texas, we were required to pick a subject that we were interested in and perform research such that we could obtain and manipulate data to be presented in a written project. The result of that project is presented here on this web page.   For this web page, I will attempt to explain my methods, and the process I went through to obtain the necessary data and what I did to get the data into the final form shown below.
    I chose to do my research in the area of of the Upper Sevier Hydrologic Basin, located in Southern Utah. I chose this area because I am originally from Southern Utah and I am familiar with the area.  I know that area is a particularly dry area of the state as far as rainfall is concerned, and I was interested in observing some data that would relate rainfall and stream flow for that particular watershed.  Given that I have very little experience in the area of hydrology and the use of GIS software, I wanted to pick a project that would challenge me to the point of learning new material about both the software and methods of hydrology, but I also wanted to pick a project that was actually realistic and not too overwhelming.

    Objective
 
My main objective of this project was to observe stream flow and rainfall data for the Upper Sevier Hydrologic Basin, and try to observe any correlation between the two sets of data.  I thought it would be interesting to try and relate historical data of precipitation to stream flow, in such a way that one might be able to roughly predict stream flow levels from precipitation.   This would be accomplished by obtaining stream flow and precipitation data for the hydrologic basin, and manipulating the data in order to establish regional runoff ratios for the basin.

 
Upper Sevier Water Basin
 Picture of Utah Hydrologic Basins

 
      This is a picture of all the Utah Hydrologic Basins.  Originally I considered doing the Cedar/Pa rowan Hydrologic Basin, because I'm actually from Cedar City, but then I  changed my mind to the Upper Sevier Basin instead because it was more in the mountains and I thought it would provide more interesting data as far as precipitation and stream flow.  The Cedar/Parowan area is in the valley west of Cedar City, where it's really dry and a lot of sagebrush and dry dirt.  There's really not very much precipitation that way.
         Upper Sevier Hydrologic Basin in relation to surrounding counties,
                 showing gage stations and river reaches
 
            LOCATION


            COUNTIES


            DRAINAGE AREA  OF STREAM GAGE


            GAGE

Scope of Project
       In order to make this a realistic project, I accounted only for stream flow and precipitation values for the Upper Sevier Basin. This means I neglected all other factors that would affect the precipitation during it's path to the river, including groundwater effects, infiltration rates of the soil, ponding, vegetation, evaporation, run off from snow melt, and all other factors.
Original Method of Attack
       Once I had determined what I wanted to do for my project, and then defined the scope of what my project would involve, I came up with a basic plan of attack.  Of course the first step was to gather information about the Upper Sevier basin.  When I picked this area, I had heard the name only, but I didn't actually have any information about it.  Next, I needed to find data for the stream flow in that area.  In order to do this I would have to find data sources on the internet for stream flow that I could import into Excel for manipulation.  The third step was to get precipitation data.  This data would be obtained in a similar manner as the stream flow data.   Next I would need to combine the data from both precipitation and stream flow in order to come up with regional runoff ratios.  Among other things, I would plot the historical data from both precipitation and stream flow in order to show any trends.  From studying those plots I would be able to come up with a method to develop runoff ratios.  Finally, I would analyze the data and the runoff ratios, and list my observations and conclusions.

Learning about the Upper Sevier Watershed
        In order to find out information about the Upper Sevier Area, I got on the internet and looked up "Upper Sevier" in a search engine.  I was soon able to find several sites relating to this area.  One site that I found particularly helpful was the "Surf Your Watershed" page on the EPA web site.  On this page, I could look up a watershed by geographical area in a given state.  Each watershed has it's own web page, with a handful of links to sites with information and data for that particular watershed.   On this page, I was able to find the exact location of the Upper Sevier Watershed, including Latitude and Longitude, surrounding counties, and surrounding watersheds.  Another good site was the NHD data site on the USGS.  An example of the data available on that page is shown below.

        Once I had learned enough about the Upper Sevier Watershed to begin working on the data collection, I then began to look around for sites where I could download maps to use in the Arcmap GIS software.  I found a pretty cool site where I could find most everything I needed.  That site was the EPA Basins page.  From this site, I downloaded a few zip files which contained a map of the state, all the counties, contours, river reaches, hydrologic basins, along with almost anything else you could possibly want.  There was actually so much information there that it took me quite some time to sort through everything in ArcMap to pick out just the things I wanted for my map.  I found many other sites that I could have used for DEM models, ortho photos, quad maps, raster sheets, etc., but I ended up using the information off the the Basins page for everything I needed.  Check the other  references listed below at the conclusion of this document to see some of the other sites that had good information on them as well.


 

Stream Flow
        Once I had obtained everything I needed for my maps, I was ready to get some data for the streams.  When I first started this project, I really had no idea where to find any data on stream flow or precipitation.  There were a few sources suggested in class, one of which was the USGS web site.  There are many other sites that I found on the internet, but the USGS page seemed to be the most helpful to me because of all the possible data types that are available at that site.  The picture below showing the daily streamflow conditions is one example of the various data available on the USGS web site.  This map shows all the gaging stations that are monitored by the USGS, and each dot on the map represents a gage station.  The corresponding stream flow is distinguished by color based on the percent of normal stream flow levels.  You can click any dot on the page to see actual stream flow data for a determined time frame for that stream gage.
 


                            Daily Stream flow conditions for the state of Utah

         I wanted to find stream gages in the lower part of the watershed that would be influenced by precipitation from  the entire water shed, and not just from a local area further up in the watershed.  The data that I ended up using was a historical record of a few of the downstream stream gages in the Upper Sevier Watershed, near Hatch, Utah.  I ended up taking data over a period of almost 30 years, from 1961 to 1990.  The reason why I chose this particular time frame is basically to be consistent for both stream flow and precipitation.  As I got into comparing  the data for both stream flow and precipitation, I found many holes in the data from a lack of record keeping at various times over the years.  For the gages that I used, I found that from about 1961 on the data was pretty consistent and was available in monthly averages, so that's the time frame I used.
 


     Table of water data for a stream gage stations located in the Upper Sevier Basin

 
        The above picture shows a precipitation table from the USGS web site for a gage in the Upper Sevier Basin.  Pulling the data off of the web site was a little tricky. There are several available outputs for the data.   I tried copying and pasting the HTML data table into Excel, but that didn't work very well.  I then tried the tab separated data.  That output form copied nicely into Excel in tabbed columns that could be separated into the actual cells and columns by using a "Separate Text to Columns" command.   The problem with that was that it listed all the months in the same column.  That made it difficult to find an average value based on the month over a multiple year period.  So what I did to change the data so that I could perform individual calculations was to sort all the data by month, and then by year.  This put all the same months together, and listed them in ascending  years.  From here I could perform calculations for a specific month for a given time frame, in this case from 1961 to 1990.   I averaged all the months together over 29 years and then plotted the total monthly averages.  This graph, shown below, shows averaged monthly values in cfs over the 29 year period.

        One of the things to note about the averaged stream flow graph shown here is the bell shaped curve, peaking in the middle.  These high values of stream flow during the Spring and Summer months is because of the snow melt coming off of the mountains.  The curve on the graph shows that stream flow goes  down in the winter, when most precipitation comes in the form of snow, and then comes way up in the summer when all that snow starts to melt.  This annual phenomenon of snow melt caught me off guard. I had forgotten that the Upper Sevier Basin was high up in the mountains, where a lot of snow falls in the winter.  In fact most of the watershed was sitting more than 6,000 feet above sea level, and there was even a  popular ski resort  Brian Head is located within 20 minutes of several of the stream gages in this area.  This ski resort sits at close to 10,000 and gets 3-8 feet of snow every year.  I suspected that snow melt would be an issue, but I had no idea that the snow melt could influence the stream levels so much.  May and June especially seemed to be affected by the snow melt.

     Precipitation Data

        After I had gotten sufficient information for the stream flow data, I then went to work on getting precipitation data.  After looking around for a while on the internet, I found some pretty good sites that would give precipitation data for that area.  Again, I had problems with the data being inconsistent over the years.  That is, some rain gage stations had big holes in their records.  Records were unavailable for months or even years at time for various gages in the area.  I also found a little bit of trouble in finding sites with monthly measurements or averages.   Many sites had hourly or daily averages, but not monthly. At least not immediately visible on the web site.  I guess I could have hand calculated the months from the daily or hourly records, but I wasn't interested in doing that for thousands of entries over a 29 year period.  I wanted to find monthly data if possible.

                            Snotel Map of Gage Stations for Precipitation

    After looking through several sites, I finally found some data in the Snotel web site that I thought would work for me.  The Snotel site is actually a pretty amazing    web site, with a lot of computing power in the area of precipitation.  For example, by clicking on one of the precipitation gages shown above on the map, it takes you to a page for that gage station where you can create a graph and a table of daily precipitation data, snow water content, and max and min temperatures.   An example of a graph generated for the Box Creek gage in the Upper Sevier Watershed is shown below. The graph shows daily precipitation (blue) and snow water content (green) from 1961 to 1995.
 

 Graph of precipitation (green) and snow water content (blue) for a site in the Upper Sevier Basin

        Even though the graph shown above is pretty cool and has a lot of useful data, I actually obtained my precipitation data from a generic table of monthly averages from 1961 to 1990 for the entire state of Utah on the Snotel site.  A portion of this table is shown below.    The nice thing about this data was that is was already in monthly averages, so all I had to do was get the stream flow data to match and then I could compare the two.  I used this monthly data for several precipitation gages that were in the area of the watershed.  I was able to get the data into Excel and separate everything into columns and individual cells ok.  The only thing I had to do was move the columns around so that the year started with January and ended with December, rather than starting with October.  This way I could be sure I was staying consistent with my other data.

         I then used Excel to plot the resulting graph of average monthly precipitation from 1961 to 1990, which is shown below.   Notice that the trends in the curve show that there is a rise in precipitation in March, July and August, and a slightly in November.  This is typical for this region, where a lot of precipitation comes in the form of snow during the winter months, and then there is a rainy season in late July and August when most of the summer storms come.  Notice also that the amount of precipitation is really pretty low. The summer rainy season only averages 2 inches a month.   Again, this is a dry area of the state, and is considered desert or semi-arid desert by most accounts.
 


                Plot of Monthly Ave. Precipitation (1961-1990)

 
Data Analysis
            Once I had these two sets of data plotted so that I could see the trends, I studied them to try and look for any correlation.  In order to aid me in observing the 2 plots together, I created a log scale graph of both stream flow and precipitation.  This plot is shown below.  I discovered that the stream flow graph did not match at all with the precipitation graph as I had expected.  The precipitation varies up and down, but the stream flow only goes up in the summer time.  This is when I first realized that the discrepancy that large could only come from the snow melt.  Like I mentioned earlier, I expected the snow melt to be a factor, but it's effect on the stream flow was much bigger than I originally thought.  At this point, not knowing how to account for the snow melt, I proceeded on to the runoff ratios.


                    Plot of comparison on a log scale
 

Runoff Ratios
   Once I had both the precipitation and stream flow long term monthly averages plotted in Excel, now I wanted to develop a relationship that related the two graphs.  A brief observation of the situation indicated that this would be a simple linear function.  What I did was to divide the precipitation by the stream flow to get a ratio.  I placed the stream flow value in the denominator because I wanted to be able to isolate the precipitation.
 
 

               In other words,

                    R=i/Q

        where R= the runoff ratio,
                                i = precipitation in inches,
                       and Q = stream flow in cfs.

     Once this relationship had been established,  I realized that using these ratios would allow one to find Q from a given precipitation, as well as allow one to find the precipitation i from from a given flow.  In other words, this relationship allows for the prediction of stream flow levels from a given precipitation. Or, one can even back calculate to find out the precipitation levels in the basin upstream by taking the measured stream flow and solving for the precipitation. The resulting plot of the averaged monthly ratios are plotted below.

Adjusting for snow melt
   It's pretty obvious that the ratios have a large fluctuation from the high to the low, so the next step was now to try and adjust for snow melt to obtain more accurate ratios.  Since I didn't really know how to compensate for the snow melt, I decided to just neglect the snow melt entirely.  The stream flow graph show that the months of May and June are the biggest times for snow melt, but you can still find snow up in the shadows in the mountains until July.   So what I decided to do was to neglect the summer months entirely from April through July.  I then recalculated the runoff ratios neglecting the summer months and re-plotted the data.  This revised plot is shown here.
 

            Now I had revised ratios that looked a little better, but now I noticed my ratios were greater than one.  Originally I had thought that the units would not really matter at all when I was figuring the ratios, because it was simply a ratio.  As long as the units are consistent, i.e. stream flow in cfs and precipitation in inches, then the ratio should still work.  These were the default units of all the data that I obtained from the Internet and it would also make it easy to use the ratio when making predictions of stream flow or precipitation, because these are standard units and you wouldn't have to do any unit conversion.  You could just put them straight into the formula.  This was the idea that I presented in class for my oral presentation and Power Point slides in November, 2002.
 

Adjusting for Consistency of Units in the ratios

        After my presentation, Dr. Tarboton brought to my attention that my runoff ratios were greater than one, which should not be the case.  Also, it is generally better to express precipitation and runoff both in the same units, e.g. inches.  This can be done by dividing the rainfall runoff by the basin area and doing unit conversions.  Doing this will make my numbers to be less than one for the ratios.  The runoff should then be a fraction of the rainfall that is comparable to other watersheds.  As I have it now, basin area is buried in the runoff coefficient which makes general interpretation of it difficult.
 

        Side note:  As I was in the process of developing these new regional runoff ratios, I realized that I had multiplied the original ratios by 100 to make the numbers bigger and easier to deal with.  I  had figured that since it was just a ratio, including a constant multiplier in the formula won't make any difference as to the function of the ratio in the calculations.  Therefore all my ratios were actually smaller than one before I multiplied them by 100, and the need for dividing the runoff by the area is not as crucial.  However, just to make the units consistent and to see how it will change my graphs, I  continued to divide the runoff by the area.  That process went as follows:


 In looking at my drainage area, the main stream gages that I used for my data had a drainage area of 340 square miles. Dividing this area into the runoff with the appropriate unit conversions will give the following monthly averages for precipitation runoff:


 
 

Using this new data to calculate runoff ratios, I plotted the following data in Excel:
 
 
 


 

Now, just to compare the shape of the revised regional runoff ratios to the original runoff ratios, I plotted the following  on a log scale in Excel.


           This final graph above shows that my initial assumptions were right.  Since this is only a ratio, whether or not the units are consistent really doesn't matter that much.  The fact that the two plots shown above match almost exactly with each other, even though they're on a different scale, proves that this is still a simple linear function, and it doesn't matter what type of constant multiplier is placed in the relationship, such as dividing by the area.  It's still a simple linear relationship and can still be used to predict stream flow in the manner mentioned above in the Runoff section.   Of course just for good habits sake, it's probably better to use consistent units in the ratio.  And it is important to remember that these ratios are only valid for the area and gage stations used for these calculations.  In order to develop ratios for any other region or watershed, there must be calculations performed for data from that area, and that area only.
 
 

 
Conclusions

         The back calculation of the precipitation was not something that I had not considered or predicted until I actually saw the simple relationship written down on my notebook.  In many cases the precipitation gages do not represent a completely accurate picture of where the rain is falling.  Anyone who has spent much time in the mountains of Southern Utah will attest the fact that you can be in the middle of a solid downpour, while you're friend can be on the next hill over and be completely dry.  Rain storms can be extremely localized, which can cause errors in rain data.  For example, consider the possibility that there is a rain gage, located at the foot of hill. There is no gage on the other side of the hill, which area is still in the same watershed as the rain gage.  It's possible for one side of the hill to have several big storms while the other side of the hill gets only a few drops by the time the storms crosses over the hill.  The streams downstream in the watershed could be swollen with rain runoff, but the precipitation gages did not measure any moisture that would cause such an event.  In this case, one could use the runoff ratios to determine the precipitation above the stream gage where there are a lack of rain gages.  Now of course, this would not allow you to know exactly where the rain was coming from, but you could at least come up with an average rainfall over the catchment area of the watershed.

          This relationship is admittedly only a simple reduced model of what actually happens in real life, but if it were developed further it could be a simple yet powerful tool for helping to predict floods and stream flow levels for a given area.  Our technology is such that the weather people can usually predict fairly accurately how much moisture a given storm is going to produce.  Using this runoff ratio relationship would allow for advanced warning systems to be in place hours, if not days in advance, depending on how far upstream the storm is located.  Some of the ways that this model could be developed further would be to include other factors such as snow melt, infiltration, evaporation, vegetation, and all other factors that will affect whether or not runoff will actually reach the streams.    I think the 2 biggest factors in the runoff behavior is the snow melt and the infiltration of the soil.  Future work in this area should include a way to account for those 2 factors.

        If I had more time to explore for this project, I would explore the Snotel web site further.   The fact that there is data available that will provide the moisture content from a snow storm could provide a means to compensate for the melting snow runoff that had such a big influence on the stream levels.   Given enough time, it should be possible to develop a relationship that will accurately account for snow precipitation and corresponding moisture content in a manner such that you could predict the level of runoff from the melting snow that will reach the streams.  This concept really intrigues me, but it is beyond the scope of this paper at this time.

          The final suggestion for improving the accuracy of this study would be to take smaller areas of data.  Studying smaller areas where the rainfall is more uniform may be better to establish general runoff ratios.  Unfortunately there is a limit to how small the areas can get because of lack of equipment and funding.  Maybe in the future this can be resolved with the further development of technology.

 References
 

NOHRSC                                                   http://www.nohrsc.nws.gov/

USGS                                                        http://www.usgs.gov/
                                                                    http://edcnts14.cr.usgs.gov/Website/nhdserver/viewer.htm

EPA                                                           http://www.epa.gov/

EPA Basins                                               http://www.epa.gov/OST/BASINS/b3webdwn.htm

NRCS SNOTEL                                       ftp://ftp.wcc.nrcs.usda.gov/data/snow/ads/ut/ut6190pr.html
                                                                  http://www.wrcc.dri.edu/snotel/snoareas.html
DR. Tarboton, USU Professor                http://www.engineering.usu.edu/dtarb/

USUWRL                                               http://www.engineering.usu.edu/uwrl/atlas/ch4/ch4hybasinsg.html