CIVL3431[6431*] – Land Surface Processes and Management
Assignment 2 – Hydrologic Modelling and Sensitivity Analysis 30 % [24%*] Due: 25th September Friday 5 pm
This assignment uses a simple lumped rainfall-runoff model, consisting of two storage ‘buckets’ (see Fig 1).
Figure 1: Bucket Model
Figure 2: Change of Runco
The upper storage represents a small interception canopy (depth = INT), where rainfall is intercepted. When this store is full, any excess moisture (throughfall, TF) enters the soil moisture bucket.
The amount of excess rainfall (TF) infiltrates into the soil moisture bucket is determined based on the storage capacity of the bucket (i.e., min (TF, Smax-Smin)). Then, water leaves the soil moisture bucket through two routes – baseflow (Qb) and surface runoff (Qs).
1. Baseflow is calculated as a fixed fraction of the moisture stored (S) in the bucket: Qb = K1.S
2. Surface runoff is generated as a linear function (Runoff coefficient, RunCo) of the amount of storage (S) above a minimum threshold (Smin) together with TF: Qs=TF.RunCo. Note when the bucket is full, the amount of TF becomes surface runoff (i.e., RunCo = 1). When S is at or below Smin, no runoff is generated (i.e., RunCo = 0). See Fig 2.
ET is calculated from two storages:
1. The interception canopy at the potential rate given that there is enough moisture in interception canopy (ET should be PET or whatever in the bucket which is smaller).
2. The soil store at [(S-Smin)/(SMAX-Smin)]*PET, provided that there is sufficient moisture in the bucket (i.e ET=0, if S =Smin).
Part 1: Developing hydrologic model with 2 storage buckets (20% - 16%*)
Using this model (Fig 1), you will be expected to manually calibrate the four model parameters so that the modelled streamflow series best mimics the observed series for the given calibration data (first 150 time steps), and then use this calibrated model to predict the response of the model to the validation data (last 150 time steps). The observed data used to calibrate the model is in RRdata.csv.
Explain this bucket model and discuss how you calibrated it and how your predictions are affected by your choice of parameters.
Your results should include a hydrograph for the full 300 time steps and a summary of the peak flow and total (cumulative) flow for both the calibration and validation periods. Include a printout of your source code in your report.
Part 2: Sensitivity analysis based on Monte Carlo Simulation (10% - 8%*)
With the model from Part 1, demonstrate the following;
Use Monte Carlo sampling to sample from specified uniform ranges of each parameter to produce randomly sampled parameter sets. Sample 10000 randomised parameter sets.
a. Create dotty plots for each parameter with efficiency 0
b. Select the best 100 parameter sets according to the Nash-Sutcliffe efficiency. Plot the cumulative distributions for each parameter for the 10000 prior samples and the best 100. Calculate the max, min, and range for the peak discharge as well as total discharge from top 100 samples.
Explains sensitivity approaches and an interpretation of your results. Which parameters display the most sensitivity?
Identified cases of plagiarism will be dealt with formally through University procedures.
Assignment Checklist and Guideline
Similar to the previous assignment for this assignment you can use MS excel or any other programming language such as Matlab to develop the model and determine the best parameter set for the model.
However the MS excel file or the computer code you developed needs to be submitted electronically via blackboard on or before the due date. In addition to your MS excel sheet and/or program please provide a report describing your work to meet the assignment requirements.
Structure of the report
Structure your report according to the standard report format including, 1.Introduction, 2. Materials and Methods, 3 Results and discussion, 4. Conclusions, and 5. References and Appendices if applicable. Include the following items in appropriate sections of your report.
Checklist for submission: The items are not in order of how they need to appear in the report. You need to include them in the appropriate sections of the report.
- Provide a simple overview of the conceptual models
- Describe the model given in the assignment. Describe different model components, modelled processes, and how they are simplified in the model. Identify different conditions, parameters, forcing variables, and dependent variables.
- Describe what data is available to you.
- Provide the “Pseudo code” you developed for this model in the appendix.
- Describe the calibration process you used to find the best parameters (or Parameter range).
- Give the range of parameters you are going to use to calibrate the model.
- Provide reasons for selecting the initial conditions and the parameter range.
- Provide the results of your calibration process. (Dotty plots for the best parameter sets “Best criteria” is up to you, best 100, best 1% of parameter sets etc.).
- Comment on whether some parameters show equifinality.
- Provide the relevant hydrographs for the best parameter set (as described in the assignment).
When you discuss the process of calibration and the process of validation, you can include the 2 hydrographs for the first 150 time steps (calibrated) and the second dataset (time steps 150-300) as validation of the model.
Finally to show that the model is fairly capable of predicting the observed values for all the data range with the calibrated parameters (best parameters), plot the observed and predicted hydrograph for the entire dataset (300 steps). You can either place it in the report where suitable or include that as an appendix.
- Summary of peak flow and cumulative flow for both the calibration and validation periods. (compare observed and simulated quantities)
Peak flow charts are simply bar chart comparing observed peak flow value (highest Qobs), initial simulated peak flow value(with the first parameters you used in the model without any calibration (Qsim)) and the
final simulated peak flow value (generated form the best parameter set found by the calibration process). Cumulative flow chart is another bar chart (similar to the Peak flow chart) comparing the total amount of flow for the 300 time steps for observed (SQobs), initially simulation (with initial parameters) (SQsim), and SQsim generated using the best calibrated parameter set.
- Include the source code of your program (if you used a programing language) as an Appendix. Part 2:
a. Dotty plots are to be plotted similar to those presented in the lecture (wk 5), using cumulative likelihood distribution and the feasible range of parameter value. Please use the same procedure in developing these plots.
b. Show the maximum, minimum, and range of total and peak flow from the best 100 runs.
Note that the matlab code and excel sheet provided last lecture (wk 5) show the step by step instruction to prepare the cumulative distributions and dotty plots.