April 24, 2018

Building Blocks in GoldSim

Posted by 

One of the many strengths of GoldSim is its graphically enhanced user interface, which allows you to quickly build your model from reusable components. During a recent webinar, I built a reservoir operations model that receives water from a snowmelt driven watershed using 6 existing models found in our Model Library. Once I understood what my modeling requirements were, I was able to quickly compile a probabilistic system model that is powerful in demonstrating how to operate a dam that services changing water demands and climate scenarios, which protects against downstream flooding.

The model simulates the inflows to an existing reservoir in the Rocky Mountains just east of Salt Lake City, Utah called Little Dell Reservoir. Because of it's elevation and location, the runoff from this watershed (picture shown below) is mainly driven by snowmelt in the early spring time. The following models from our Library were used to build this model:

The model was quickly built by downloading the above models and simply copy-pasting them into a new model as localized containers. 
This process was repeated until the entire system model was constructed. 
Within each of these containers is the model downloaded from our Library. Each container includes references to input data and outputs. I had to link these inputs and outputs from one container to the next and these references are what causes GoldSim to draw the influence lines shown in the screen capture above. 

To calculate the snow accumulation and melting process, I used a model from our library that is based on the National Weather Service's SNOW-17 model. This model is mainly driven by temperature so I added another model that simulates the changing daily temperature probabilistically. This is done in our library model called the WGEN Weather Generator, which is based on the NRCS WGEN model.
The WGEN model is able to simulate rainfall using a Markov Chain with stochastically scaled precipitation amounts derived from historic rainfall time series. The data used to develop this model were obtained from the Parley's Summit SNOTEL site operated by NRCS. This site measures min/max daily temperature, water in the snowpack, and precipitation. Using this information, I calculated the statistical inputs required for WGEN to generate the stochatic rainfall, temperature, and radiation values on a daily basis. 

With daily temperature and precipitation values, I was able to use these to drive the Snow 17 model. 
The main output of the Snow17 model is the excess water, which is the sum of rain and melting snow. This data was fed into a runoff model called the Australian Water Balance model (AWBM) which has been used for calculating the rainfall-runoff process for watersheds in arid regions. The resulting runoff was linked to the inflow of the reservoir model. The reservoir model simulates deliveries made through an outlet works to fulfill downstream water demands while also providing controlled and uncontrolled flood controls through the same outlet works and also a spillway. The objectives of the dam operations are to provide water to supply these downstream demands while also protecting downstream properties from flooding by making strategic flood releases through the controlled outlet works and avoid uncontrolled releases over the spillway. This is more efficiently done by watching the snowpack and temperature levels as a way to predict how much additional water will be flowing into the reservoir as the snow melts. By running the model in slow motion (which is what was done during the webinar), you can proactively drop the water level in the reservoir in anticipation of large inflows from heavy melting rates from a snowpack high in water content (snow water equivelent or "SWE"). Conversely, if there is a small amount of SWE then the decision to hold the water level without releasing flood flows can be safely made.

I also used the stochastic temperature values to estimate reference evapotranspiration (ETo) in preparation for determining crop demands so that I could simulate dam releases required to fulfill these demands. The ETo was used to calculate water demand for a given crop mix assumed for downstream agricultural areas. The end result is a powerful model that provides insight into many real-world possibilities based on uncertainty of the weather. 

In the future, this model could be improved by adding in capability to estimate the future inflow statistics in order to preemptively open the flood gates at the safest and most efficient time. If the gate is opened too early, you risk losing too much water that could have been reserved for serving irrigation needs but if opened too late, you risk spilling water over the spillway at a much higher and uncontrolled rate, which puts downstream properties at risk.

No comments:

Post a Comment