GEOG 566

         Advanced spatial statistics and GIScience

June 11, 2017

Analyzing Hydrologic Patterns in a Single Basin in the Oregon Coast Range

Filed under: Final Project @ 5:13 pm


My master’s thesis involves quantifying geomorphic change in streams after the addition of a large wood jam for salmon habitat restoration purposes. Large wood has been shown to have many positive effects on salmon habitat, including creating pools and backwater areas that are essential for juvenile salmon survival. This past summer, I surveyed seven stream reaches within a single basin in the Oregon Coast range to capture the in-stream and floodplain topography before the addition of large wood restoration structures. Three other sites were previously established in 2015 for a total of ten study sites. I plan to survey these sites again next summer to capture changes in topography caused by the interaction of wood, water, and sediment.

A major part of my fieldwork is monitoring water stage height at different locations in the drainage basin to capture the hydrologic patterns that are driving the geomorphic change at each LW site, including the frequency and duration of inundation of these log jams. I formulated three questions to help guide my analysis of this data:


  1. How well correlated are stage gage data at different locations within a single basin, based on the amount of distance between them?
  2. How does upstream-downstream connectivity, drainage area, and position within the catchment affect the relationship between different site hydrographs?
  3. How can I model the physical water surface elevation at different cross sections at my log jam sites?


To conduct my analysis, I used water stage data gathered using Level Loggers at 6 different engineered log jam sites in the Mill Creek watershed. Three Level Loggers were installed in Fall 2015 and three more were installed in Fall 2016. Stage height was recorded at 15-minute intervals at each site. These stage gage instruments are pressure transducers. Therefore, I used a seventh gage that recorded atmospheric pressure to correct the stage height in MATLAB after downloading the data in the field.

I also have discharge rating curves for four of these stage gage sites, obtained through repeat discharge measurements taken at four of the stage gage sites. These can be used to calculate the discharge for any given stage height and obtain values that I can incorporate into my model of the physical water surface elevations.

The approximate location of each instrument was recorded using a handheld GPS unit. The streamwise distance between sites varied from 0.63 to 7.51 kilometers (Table 1) calculated using attribute table statistics in ArcGIS. The upstream drainage area of each stage gage site was from 4.02 to 21.76 square kilometers (Table 2), calculated using the NetMap Toolbox in ArcGIS.

Three water surface elevation (WSE) rebar were also installed at each log jam site. The top of each WSE rebar was surveyed in with a Total Station during my summer field season. During three subsequent visits to the field, I noted the distance between the top of the rebar and the water surface, marked off in 2-cm increments. This data could be later correlated with the stage gage data in order to answer my third question.


Question 1

I expected to see a negative relationship between distance between instruments and correlation coefficient.

Question 2

I expected sites with more similar drainage area sizes that are closer together geographically will have more similar hydrograph shapes. I expect that streams that are connected will also be more similar.

Question 3

I wanted to be able to use the step-backwater equation or HEC-RAS to model the physical water surface elevations of my sites, based on the physical observations of the WSE rebar and my stage height and discharge data.


I worked mostly in Excel and ArcGIS, using ArcGIS to derive basic spatial relationships between my stage gage sites and Excel to calculate statistics from these data.

Question 1

For Question 1, I used to Excel to import and organize all of my stage gage data. I identified the time period where I had data for all six gages, then created a table to set up a pairwise correlation coefficient analysis for all combinations of the six gages for that range. Finally, I calculated the correlation coefficient for each pair of stage gage arrays in the table. To measure the streamwise distance between each pair of stage gages, I used a previously created map in ArcGIS with GPS coordinates for the six stage gages and the stream polyline on which the gages were located, selecting all portions of the stream polyline between each pair of stage gage sites and using the Statistics tool in the stream polyline attribute table to calculate the sum of the selected lengths, which I input into another pairwise Excel table. Finally I used Excel to create an X-column with the distances between the paired gages and a Y-column with the correlation coefficients between the paired gages. Then I plotted X vs. Y as a scatter plot and examined the relationship.

Question 2

For Question 2, I used attribute tables in ArcGIS to determine the upstream drainage area of the four sites I was interested in. I chose to look at hydrograph relationships at four of my sites that are all have upstream-downstream connectivity but have different upstream drainage areas and are located on different tributaries. Then I used Excel to further manipulate the stage gage data that I had started to analyze for Question 1. I graphed the difference in stage heights over time between site with the largest drainage area, and each of the three other sites to examine patterns in the hydrographs, creating three “stage difference” graphs. I also graphed the raw hydrograph data for comparison.

Question 3

For Question 3, I scaled down my analysis and created a DEM of a single log jam site in ArcGIS. Then I interpolated an elevation profile perpendicular to each of the three water surface elevation rebar which I imported into Excel. Then I used an Excel spreadsheet that utilizes the step-backwater method to calculate and graph the water surface elevations at the three rebar locations.


Question 1

Correlation coefficients ranged from 0.86 (Site 9 vs Mainstem) to 0.97 (Site 26 vs South Fork). Distances between sites ranged from 625 m (Site 36 vs Mainstem) to 7512 m (Site 9 vs South Fork). Five out of the six site pairs with the lowest correlation coefficients included the Mainstem site as part of the pair.

The plot of distance vs. correlation coefficient yielded a slight negative relationship with an R2 of 0.0117 which indicates low correlation (Figure 2).


Question 2

All sites of interest have upstream-downstream connectivity with Site 26 (Figure 3). Upstream drainage areas ranged from 4.02 to 21.76 km2 (Table 4) while pairwise distances ranged from 0.63 to 3.76 km (Table 3).

Mainstem (MS) and South Fork (SF) are on the same tributary have different upstream drainage areas and are different distances from Site 36 (Table 3, Table 4, Figure 3). Both graphs showed little difference with Site 26 during baseflow conditions. The 26 v MS graph was relatively flat but it still had some high peaks during the two largest floods of the season which indicates other sources of water that are influencing the hydrograph (Figure 4). The 26 v SF graph had high positive peaks during floods (Figure 5).

Mainstem and Site 36 are a similar distance from Site 26 but have different drainage areas and are on different tributaries (Table 3, Table 4, Figure 3). They are on also on different tributaries (Figure 3). The 26 v. 36 graph looked very similar to the 26 vs. SF graph (Figure 5).

Site 36 and South Fork have similar drainage areas but are different distances from Site 26 and also are located on different tributaries (Table 3, Table 4, Figure 3). The two graphs were very similar in shape despite differences in geographic location, though the Site 36 hydrograph was more similar to Site 26 during baseflow conditions (Figure 5).

Question 3

Two of the WSE rebar (WSE 2 & WSE 3) are located very close together, at the upstream end of the site. The third rebar, WSE 1, is located at the downstream end of the site (Figure 6). Cross section numbers correspond with WSE rebar numbers.

The November and February WSE rebar observation dates resulted in similar, higher water surface elevations with lower water surface elevations in December (Table 5).

Stage height ranged from 0.364 to 0.692 m for the three WSE rebar elevation dates, with the lowest stage height in December and the highest in November. Discharge was correlated with stage height, with the lowest flow in December, 0.85 m3/s, and the highest flow in November, 2.16 m3/s (Table 6).

The plots generated from the step-backwater spreadsheet in ArcGIS looked relatively similar for Cross Sections 1 and 3, with a lower water surface elevation in December and two very similar, higher water surface elevations in November and February (Figure 7, Figure 8). This corresponds with the patterns I observed in the WSE rebar data and the stage and discharge data (Table 5, Table 6).

Graphing the WSEs at Cross Section 2 yielded different results, with a much larger separation between the November and February WSEs. This might be due to the different topography at Cross Section 2 which does not constrict the water in the same way as it does at Cross Sections 1 & 3. Figure 5 shows the simulated water surface elevation if the water flowed all the way across the cross section into a secondary channel, though in reality it probably only accumulates in the main channel (on the right) because of the obstruction in between the two channels.

Several of the calculated water surface elevation values have large discrepancies (~0.2-0.3 m) which indicate issues with the data (Table 3). I repeated the entire analysis and also tried changing the Manning’s roughness values in the spreadsheet which can have an effect on flow levels, but neither of these methods had any significant impact on the outcome.

Graphing the observed and calculated water surface elevations in comparison with the bed elevation also illustrates inconsistencies with the data. All three observed WSE lines follow the same general shape, increasing gradually and then dropping off. The calculated WSE line for December follows the same pattern, but the lines for November and February have the opposite shape, with an almost flat water surface for most of the reach and a sudden jump up at the last cross section (Figure 6).


The overall purpose of my research is to explore the response of site-specific features that vary with scale to engineered large wood jams within an individual basin. This kind of research can help refine stream restoration efforts and optimize resource allocation. For this class, I focused on understanding the hydrologic patterns on my watershed. The patterns of discharge, floodplain and large wood inundation throughout the basin can help clarify the picture of what is truly driving the geomorphic changes that we measure in the watershed.

Question 1

When addressing Question 1, I learned that regardless of location in the basin, the water levels are responding relatively uniformly, with only a slight delay between sites that are spaced further apart. Instruments are located on different tributaries so the relationships are slightly more complicated than I expected but I think I will still be able to apply what I have learned from this analysis to extrapolate to other parts of the basin where I do not have stage data.

Question 2

For Question 2, I learned that sites with similar drainage area sizes and geographic locations behave more similarly, and that drainage area has a greater influence on the similarity of hydrographs than distance does. I was also able to identify a portion of the hydrograph that is still missing and attribute it to the water flowing through North Fork Mill Creek, which is not instrumented. It would be interesting to install a LevelLogger on this tributary in the future to see if it completes the picture

Question 3

The results of Question 3 were not very conclusive. My analysis showed that I do not have enough data points to understand what is really going on. Two of the cross sections (2 & 3) are not spaced very far apart which could also have contributed to the issues with the data. Despite these issues, this analysis gave me insight into what I need to do in the future to ensure the robustness of my dataset, namely more, spaced-out water surface elevation rebar at each site to create a more complete water surface elevation profile.

All of my findings reinforce relatively well known hydrologic concepts. I will be able to provide more interpretation of these patterns after I have gathered my second round of geomorphic data. Then I hope to be able to correlate these hydrologic patterns with the changes we see in local channel morphology around each log jam site. This kind of information will help restoration planners determine where in the basin is the best place to put a log jam.

Learning Outcomes

At the beginning of the term, I assessed my proficiency with ArcGIS, Python, and R as follows:

ArcGIS: Intermediate level (I spent last term working with it. I can perform most basic manipulations and figure out what I don’t know how to do from the help menus and the internet.)

Python: Beginner level (I have used it in two classes to perform basic calculations and plot data/results.)

R: Very limited experience (I used it a few times in my undergraduate Statistics class but that was 7 years ago so I will probably be very rusty.)

I was excited to expand my skills in Python and R but the class did not end up providing me with a lot of opportunities to do so. I was encouraged to use software that I knew and therefore ended up conducting my analyses entirely in ArcGIS and Excel. I did not have the type of spatial data that was suited to large-scale spatial statistics anyways, so my only exposure to methods such as hotspot analysis, spatial autocorrelation, regression, and multivariate methods was through the tutorial presentations, where I learned about how other people had used these approaches. It would have been more educational to apply these concepts to my own data but unfortunately that was not really an option.


Tutorial 1

In Tutorial 1, one of my reviewers suggested that my method for calculating distance between points was not very efficient. It is possible that I could have used Python to automate the process, but the process of teaching myself how to use Python in order to perform this relatively simple if repetitive task would probably have taken more time than just doing it the way I did it. If I had to perform this manipulation thousands of times instead of less than twenty times, it may have been worth it to explore other options. The same is true for my methods for calculating correlation coefficients.

Tutorial 2

In Tutorial 2, I got a lot of questions about other factors besides drainage area and geographic location, such as elevation, that could also be influencing the hydrograph patterns. One reviewer also suggested that I compare sites with similar drainage areas more closely to see if drainage size really is the biggest determining factor in flow variation under different flood and drought scenarios. These are all really great suggestions that I look forward to exploring in the future.

Exercise 1

I did not get any suggestions for how to improve Exercise 1.

Exercise 2

Based on the comments I got on Exercise 2, I converted my units of drainage area from m2 to km2 which are more reasonable and understandable. I also clarified the naming conventions for my sites. I combined what had initially been three separate graphs into one graph to more efficiently illustrate hydrograph patterns. Finally, I added a discussion of the limitations of my approach.

Exercise 3

Apparently my WSE rebar labeling was not very clear so I improved that for my final write-up.


Print Friendly, PDF & Email

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

© 2019 GEOG 566   Powered by WordPress MU    Hosted by