Towards Better Hydrologic Process Understanding at Shell Creek Watershed
Improving understanding of flooding and rainfall-runoff processes at Shell Creek Watershed, a rural Nebraskan Watershed. This study seeks to determine the relative influence of selected factors on flooding, as well as investigate a potential link between conservation practice implementation and flood reduction observed in the watershed.
icon search Searchable Transcript
Toggle between list and paragraph view.
[00:00:01.470]Hello. My name is Annushka Aliev.
[00:00:03.600]I'm an undergraduate researcher
from University of Maryland,
[00:00:06.780]and this summer I've been working
on improving hydrological process
[00:00:10.170]understanding at shell Creek watershed.
[00:00:14.700]Flooding is a significant and
recurring issue in Nebraska,
[00:00:18.270]and one recent example of
this is the March, 2019 flood.
[00:00:23.130]flood was the culmination of several
different factors and forecasting at the
[00:00:27.120]time looked into the interplay of these
factors and was able to save lives and
[00:00:31.530]property. In our work,
[00:00:33.930]we are focusing on a specific rural
watershed with a history of issues,
[00:00:37.800]including chronic flooding.
This history of issues
[00:00:40.980]led to the implementation
of lots of conservation
practices starting around 2005
[00:00:46.830]in the watershed, which makes it
particularly interesting to study.
[00:00:52.200]The main objectives of this
study are as follows: to identify
[00:00:56.550]the influence of various factors on
flooding and hydrology in Shell Creek; to
[00:01:01.410]develop a rainfall runoff model
using artificial neural networks
[00:01:05.280]to predict flooding;
[00:01:07.050]and to conduct a drought intensity
[00:01:09.660]and flood frequency analysis in order
to better understand hydrological
[00:01:14.130]processes in the watershed.
[00:01:16.500]And these will have forecasting
[00:01:19.800]as well as lay some groundwork for a
future study that is coming down the line.
[00:01:25.710]In order to accomplish these objectives,
three main pathways will be used.
[00:01:30.210]The main pathway - this center pathway -
selection and model development,
[00:01:35.340]which will aid in flood prediction and
determine the influence of variable
[00:01:39.930]predictive variables on our
target variable discharge.
[00:01:44.190]The other two pathways are
flood frequency analysis,
[00:01:47.010]which will tell us about
changes in probabilities
[00:01:49.980]-flood probabilities- and drought analysis,
[00:01:52.650]which tells us about wet periods and
dry periods. Flood frequency analysis
[00:01:57.120]and variable selection in this study,
will be run for three periods:
[00:02:01.020]the full period, the non conservation
period and the conservation period.
[00:02:06.870]An important thing to note
[00:02:08.159]is that only one gauge
provided up-to-date stream flow data,
[00:02:11.610]which was our target variable,
[00:02:13.350]so this gauge limited the time
span and scope of our study.
[00:02:17.370]This gauge is indicated with
a yellow star on this map.
[00:02:23.400]On the right,
[00:02:24.090]there is a diagram of an artificial
neural network, and artificial neural
[00:02:27.870]networks are the machine learning
algorithm that we used in our work.
[00:02:35.500]let's first look at the flood frequency
analysis and drought analysis.
[00:02:39.790]On the left, for the flood
[00:02:42.640]we compared full period
and non conservation period
in order to tell us about
[00:02:46.960]the conservation period.
[00:02:48.910]What we learn is that extreme flows
are of much greater magnitude in the
[00:02:54.460]but typical flows are actually
lower than in previous years.
[00:02:58.810]The drought analysis tells us a
bit about soil moisture
[00:03:02.350]- we can find that from the one month
SPI - and it also tells us about what dry
[00:03:07.060]periods - this is most clearly agitated
by the 12 month SPI. From these,
[00:03:11.920]we learned that soil moisture conditions
are typically wetter in spring and
[00:03:15.490]fall and sometimes summer,
[00:03:17.470]and that floods line up fairly well with
wet periods in Shell Creek watershed.
[00:03:24.010]From variable selection,
[00:03:25.720]what we find is that no single variable
is highly correlated with flooding for
[00:03:31.630]This is also supported by the model
performance changes from our leave-
[00:03:37.420]The artificial neural network was
run without each predictive variable,
[00:03:41.080]and most often this resulted
in a prediction performance
drop indicating that
[00:03:45.610]the model performs best
with all variables included
[00:03:50.680]So what can we learn
from these results?
[00:03:53.602]In terms of individual predictability
precipitation is most tied to flooding in
[00:03:58.390]Shell Creek watershed. However, this
correlation - like the rest - is low,
[00:04:03.430]so we suggest the inclusion of
all variables in order to best
[00:04:08.380]predict flooding in Shell Creek
[00:04:10.900]watershed. Of the models we
developed in this study,
[00:04:14.500]we suggest the use of the conservation
period model as it performed best,
[00:04:19.810]and most closely represents the
current conditions in the watershed.
[00:04:26.620]we identified that flooding decreased
in intensity during the conservation
[00:04:30.670]period, though not necessarily
due to conservation practices,
[00:04:34.570]as we noticed that this was a trend
that continued from slightly before the
[00:04:38.800]conservation period and can not
conclusively attribute decreased
[00:04:43.270]flooding to conservation
practices in the watershed.
[00:04:48.820]And that's my work from this summer.
[00:04:50.560]Here are some references that
were used in this poster,
[00:04:54.490]and I'd like to thank some people as well.
[00:04:56.710]Funding for this project was provided
by the national science foundation and
[00:05:01.810]go out to these people here for connecting
us on a more personal level with the
[00:05:06.040]watershed and allowing
us to tour it. Thank you.
Log in to post comments