Preparation and Evaluation of GroundWater Simulation

A serie of demonstrations for how to setup the model and how to test the model for simulations of GroundWater Level can been viewed. All the examples will be based on a course given to Swedish Geological Survey during January 2017. You are invited to view the example below for education purposes.

The examples are mostly made for new users of the model but they may also be useful for experienced users since they demonstrate a number of options that have been developed during the latest years.

Deeper understanding about specific processes may still be best studied for the Tutorials.

To follow the process of how to setup the model and apply the model I recommend you to follow the links below.

(1) Organise the windows of the interface to follow the style given here.

(2) Preparation for creating input file of climate variable and ground water will provide example of new ways for how to convert  from text files and for how to view climate data and other time series. A mixing of two climate file with different coverage in time and by variables is demonstrated.

(3) Creating of simulation of will be made in a number of steps  to represent a complete hydrological simulation of water balance and ground water dynamics. The first will show show how to make the link to climate data from a previous empty document.  After this the document will add : (a) Soil profile description using data from data base,  (b) a simple vegetation cover will be defined and finally  the drainage component will be defined.

(4) The link of the simulation to measured data that will be used to evaluate the performance. The performance variables and the options for impoving a simulation results are made with single runs and multiple runs. A first and a Second try will be made to improve the simulation by single runs. That is to change some specific parameters step by step.

(5) A multiple run simulation will be defined based on the results obtained from the single run simulations. The multiple runs will first be pre-evaluated prior after a quick test. This will be followed by a realistic evaluation of 1000 runs to understand the performance of the model. The first focus is to find and describe how an accepted ensemble of runs can be identified from the available performance indicators.  The parameter sensitivity and the possible use of the data to contrain the uncertainty of the model will be shown. This will be followed by a detailed evaluation of the residuals and the search for patterns that are not explained by the model. Finally a feature to define the ensemble by considering transformed version of the simulated and measured data are done.  Two ensemble of accepted runs will be possible to compare.

(6) A new multirun apprach is also made based on Bayesian calibration instead of the GLUE style of calibration applied in the examples above.  The demonstration shows both how to setup the model for a Bayesian calibration and how to view the results obtained. The posterior distributions can be compared also with the corresponding subjective performance indicators as was used in the GLUE style.

The Example described in the process of setting up the model is made with Lappträsket in the north of Sweden. Already prepared simulation with ready documents to simulate the hydrology are available for LiaTorp, Motala – till Soil and Motala -Sandy soil.  In the case of  the Motala- sandy soil  also a complete Multi Runs is made. The simulations for the 3 additional sites are recommended as short cuts if you do now want to make the setup of the model by yourself.

After following the steps above the student/user is expected to be able to further develope the modelling to cover new sites with new data.  It will also open for new comparisions and especially by adding more data on the real world conditons.