iMOD User Manual version 5.2 (html)

7.16iPEST Analyser

WHY?
Model optimization is a proces that generates many output variables, such as objective function values, parameter sensitivities, parameter adjustment factors, reliabilities, spatial and temporal residuals. Many of these need to be studied at the same time to understand the values and qualify the calibration results. The iPEST Analyser is a tool that provides insight in the above mentioned variables in a dashboard view in which all of these are visible.

WHAT?
It takes a folder with the output of the iPEST or iPESTP. This tool reads all files that are saved via a optimization with iPEST and/or IPESTP.

HOW?
Select the option Toolbox from the main menu and then choose the option iPESt Analyser to display the iPEST Analyser window.

iPEST Analyser window.

Select an Optimization Cycle	Select one of the available cycles to display the corresponding results.
Read iPEST Logfiles	Click this option to select the folder that contains the iPEST output files. If iPEST has been used, this is the PEST folder that contains files with the extension “_MF2005”. For simulations that used iPEST Parallel, select the folder IPEST, files here does not have the extension “_MF2005”.
Parameters	Select the parameters that need to be displayed in the graphs on the sections B,D and E.
Periods	Select the periods (transient simulations only) that need to be displayed in the graphs on the sections C and F.
Layers	Select the filters within the selected model layers that need to be displayed in the graphs on the sections C and F
Filters	Select the filters that need to be displayed in the graphs on the sections C and F.
A
Quadratic, Absolute Difference, Difference:	This graph shows the selected cycle of the optimization. The blue line shows the total objective function value (quadratic differences between measurements and observations), the green line the objective function value from the quadratic difference of the changes in parameters. This option shows the objective function value in different forms. Example of the different objective function value representations.
B
Iteration, Parameters, Labels	This option shows the sensitivities (%) per parameter and per iteration (Iteration). Each line represents a single parameter as selected in the list to the right. If Parameters are selected, the sensitivities per parameters for a single optimization are displayed in a bar graph, if Labels are checked, each parameter is labelled. Example of the different sensitivity representations.
C
Iteration, Parameters, Labels	This option shows the residuals in a scatter plot per iteration (Scatter; the bandwidth (red/green) can be modified in Bandwidth), or in a histogram (Histogram). The lower- and upper limits for the histogram is defined automatically at the 10% and 90% percentiles. The figures are presented for the selected Periods, Layers and/or Filters. Any (de)selection in any of these lists, change the scatter plot and the histogram. Example of the different residual representations.
	Properties Click this button to ... not yet filled in
	Legend Click this button to ... not yet filled in
D
Iteration, Parameters, Labels, Include Reliability	This option shows the multiplication factors per parameter and per iteration (Iteration). Each line represents a single parameter as selected in the list to the right. If Parameters are selected, the multiplication factor per parameters for a single optimization cycle are displayed in a bar graph, if Labels are checked, each parameter is labelled. It is also possible to visualize reliability intervals for each parameter (Include Reliability). Example of the different sensitivity representations.
E
Correlation, Eigenvalues, Variance, Labels	This option shows the correlation matrix per iteration whereas the correlation is expressed in red and green colours, the more vibrant the colour the higher the correlation. The exact values can be plotted if the Labels is checked. If Eigenvalues are selected, the eigenvalues are presented in a bar graph for a single optimization cycle, it represents the uniqueness of the sensitivities per iteration cycle, if Labels are checked, each eigenvalue is labelled. It is also possible to visualize the variance values per eigenvalue (which is in fact the square-root of the eigenvalue) (Variance). Example of the different correlation and eigenvalue representations.
F
Difference, Progressive Diff., Time Series, Labels	This option shows the residual spatially (for transient simulation this will be the average difference value) and in time series (transient simulations) per iteration. If Progressive Diff. is selected, the differences between two adjacent optimization cycles are presented. It Time Series is selected, the time series of the measured and observed time series are plotted. This is done for the selected Periods, Layers and/or Filters. If Labels are checked, the value of the residual will be plotted (obsolute yet). Example of the difference between observations and measurements in spatial plots and time series.
	Properties Click this button to ... not yet filled in
	Legend Click this button to start the Adjust Legend window. Here the colouring can be modified used to display the residuals.
	Zoom In Click this button to zoom IN on the centre of the current graphical dimensions.
	Zoom Out Click this button to zoom OUT on the centre of the current graphical dimensions.
	Zoom Rectangle Click this button to zoom in for a rectangle to be drawn. Use you the left-mouse button to determine the lower-left corner of the rectangle, click again for the upper-right corner (or vice-versa).
	Zoom Full Click this button to zoom in on the entire extent.
	Move Click this button to move the current display. Click the left-mouse button on that location where you want to move from, repeat this after the display has been refreshed (automatically). Use the right mouse button to stop the moving process.
Help	Click this button to start the Help functionality.
Close	Click this button to close the iPEST Analyse window.