Channels
The purpose of this project is to assess the trends in the Winooski River's channel and discharge. As well as to document geomorphic changes from aerial photography. Note all data is publicly available from USGS and NOAA.
The River Though Time
The image carousel to the left begins with a 1927 aerial photo, from UVM Special Collections, of the Winooksi River at Williston, days after the historic flood. The farmland is completely washed out, with eroding cut banks forming on the hillside hundreds of yards west of the river bank. The bridge was lost. The only safe areas were high ground farms. Pools of water and extra streams were cut into the landscape.
The second image is rephotography by Elizabeth Stanley Mann in 2004. The river has returned completely back to the channel and out from the farmland. the eroded cutbanks from the flood have been repaired and sown with trees.
The third image and onward are wider angle images captured from GoogleEarth in 2004, 2006, 2008, 2009, 2011, 2015, and 2018. They show little change in the riparian zone, but there is evidence of substantial river slow-down with deposition of sediment in the central channel sandbar. As time elapses, you can see the western channel slowly shrink, until the west bank becomes a point bar of the river.
A Very Bad® Flood
Gathering discharge data of the Winooksi River near Essex Junction, Vt (1927-2018) I constructed relative probability to peak annual flow, and mean annual flow as seen in Figures 1, 2 and 3 (independent variable plotted on the vertical axis to emphasize river height in flood stage). Using Normal Probability and logarithmic scales, the data plots quite linearly (meaning a standard exponential relationship) r^2 = 0.889, except for the extremely low discharge of 1965, and the extremely high one of 1927. Visually, one can see the disparity (figure 2), the 1927 flood had a peak discharge more than twice as high as that expected in a 100 year (1% chance) flood. Making it more to be a 100,000 year flood, or rarer.
Data and calculations for this project can be found here:
The peak annual discharge data (1927-2018) has a mean of 22082 standard deviation of 11505, resulting in a relative standard deviation of 0.5210
The peak annual discharge data - without 1927 - has a mean of 21083, a standard deviation of 6403.35, and a resulting relative standard deviation of 0.3037
The annual mean discharge data (1928-2018) has a mean of 1817 and standard deviation 441.87, resulting in a relative standard deviation of 0.2431
This means that the 1927 flood accounts for 171.55-214.32% increases in deviation from the expected trend.
Looking further, the r^2 given in figure 1 shows a decently well fit line. however when removing the '27 flood, the data shows an r^2 of 0.998. Furthermore, if we plot the monthly mean discharge data (figure 4) from 1928 onward, we arrive at an r^2 of 0.992
The monthly mean discharge data has a mean of 1818, standard deviation of 1562.66 and a relative standard deviation of 0.85945 - this accounts for seasonal variability. Which is why it was not used to calculate relative deviations with peak flow statistics.
Further, to examine monthly variation in discharge, we can plot each year's data by month number (figure 5), which shows regular trends of high water - with very few outliers) that point to large quantities of snowmelt between March-May. The high amount of outliers during the Summer and Autumn months are likely caused by storms. These would the most likely causes of landslides seen in Project 1.
Figures 6 shows the monthly variation in Winooski River mean monthly discharge by year, with outliers plotted. Figures 7 and 8 explore NOAA Burlington Hourly Precipitation from 1948-2013. As the data do not show the same high points, it is likely that rainfall in Burlington has little effect on the behavior of the Winooski River, rather precipitation in the Green Mountains will.