[GRASS-user] calibration, sensitiviyy analysis tools in GRASS?

GRASS GRASS user
1

Dear list,

I’m looking for tools to perform calibration (parameter estimation), sensitivity analysis and uncertainty analysis on a groundwater model that i built with r.gwflow.

Does anyone have any pointers, or expertize in this? I see some reference to r.ucode in this article:

Natural Hazards

January 2014, Volume 70, Issue 2, pp 1153–1179 "r.massmov: an open-source landslide model for dynamic early warning systems"

What i’m looking to do is gain insights on a very complex urban groundwater problem with data scarcity: for example, the relative importance of different sources of uncertainty (hydrogeologic parameters versus human influence like pumping and return flows).

Thanks,

Vishal

–

Vishal K. Mehta, PhD
Sr. Scientist
Stockholm Environment Institute - US
133 D St Suite F
Davis CA 95616
www.sei-us.org

2

Dear Vishal,

I am not aware of any GRASS internal modules/add-ons to perform sensitivity/uncertainty analysis. However, you could try to use R in combination with GRASS (https://grasswiki.osgeo.org/wiki/R_statistics).

R offers useful packages to perform sampling-based uncertainty analysis of model parameters e.g. Latin-Hypercube-Sampling (LHS, package ‘lhs’) combined with partial rank correlation analysis (PRCC, package ‘sensitivity’).

I am not exactly sure if this is what you’re looking for: First, you could generate LHS-based parameter combinations for your GRASS groundwater model and save these parameter combinations into file. Second, this file can be accessed in a loop within GRASS to repeatedly calculate the groundwater model for each parameter combination. Third, the results can be collected and analysed using e.g. PRCC to identify those model parameters which are most influential on your model result. This is more an R-approach where you use GRASS just for running your groundwater model. Hope that helps.

Best regards,
Johannes

···

On Wed, Jul 27, 2016 at 8:11 PM, Vishal Mehta <vishalm1975@gmail.com> wrote:

Dear list,

I’m looking for tools to perform calibration (parameter estimation), sensitivity analysis and uncertainty analysis on a groundwater model that i built with r.gwflow.

Does anyone have any pointers, or expertize in this? I see some reference to r.ucode in this article:

Natural Hazards

January 2014, Volume 70, Issue 2, pp 1153–1179 "r.massmov: an open-source landslide model for dynamic early warning systems"

What i’m looking to do is gain insights on a very complex urban groundwater problem with data scarcity: for example, the relative importance of different sources of uncertainty (hydrogeologic parameters versus human influence like pumping and return flows).

Thanks,

Vishal

–

Vishal K. Mehta, PhD
Sr. Scientist
Stockholm Environment Institute - US
133 D St Suite F
Davis CA 95616
www.sei-us.org

grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user