Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Kan akademin hjälpa industrin utveckla mer robusta grundvattenmodeller? En studie av moderna svenska industriframtagna grundvattenmodeller

Haraldsson, Emil LU (2024) In Examensarbeten i geologi vid Lunds universitet GEOL02 20231
Department of Geology
Abstract
Groundwater modeling is often employed by professional hydrogeologists to make predictions to be
used by regulators in decision-making. In Sweden, there are no laws governing the way in which a groundwater
model may be created or shaped, and therefore they may differ wildly from one another. Thus, the goal of this study
has been to research Swedish groundwater models created by the industry during the last 11 years to evaluate which
methods have been used in their creation, to then be able to recommend methods favored by academia that would
make the modeling more robust.
In the background section, the phrases parameterization, calibration, observation, sensitivity analysis, and un-
certainty analysis are explained. They all... (More)
Groundwater modeling is often employed by professional hydrogeologists to make predictions to be
used by regulators in decision-making. In Sweden, there are no laws governing the way in which a groundwater
model may be created or shaped, and therefore they may differ wildly from one another. Thus, the goal of this study
has been to research Swedish groundwater models created by the industry during the last 11 years to evaluate which
methods have been used in their creation, to then be able to recommend methods favored by academia that would
make the modeling more robust.
In the background section, the phrases parameterization, calibration, observation, sensitivity analysis, and un-
certainty analysis are explained. They all represent decisions modelers have to make during the modeling process.
It is in regard to these phrases that the groundwater models have been evaluated. Uncertainty analysis and probabil-
ity are discussed throughout this text, as the type of deterministic predictions regulators usually prefer stand in stark
contrast to the type of stochastic predictions that are favored by multiple leading names in groundwater modeling
academia.
The results show that the researched models are generally parameterized sparingly, not all the models have been
calibrated, the number of observations used is generally low, sensitivity analysis has been completed for but a few
of the models, and that uncertainty analysis has been completed for but a few of the models.
I recommend the usage of pilot points as a means of increasing parameter dimensionality, the usage of data
worth analysis to deduce if there are enough observations in a model, and the usage of PEST or PEST++ to com-
plete calibration, sensitivity analysis, and uncertainty analysis. Furthermore, I encourage modelers to increase the
transparency of the modeling process by clarifying and highlighting the choices they make during modeling. One
way to do this and at the same time increasing the reproducibility of modeling, is through script-based modeling. In
addition to this, legislation, or the lack thereof, is discussed as a potential obstacle for more probability-based mod-
eling.
My conclusion is that there are multiple weaknesses among the evaluated models, and that, by extension, there
may also be weaknesses in other groundwater models produced by the industry in Sweden during recent years. My
main recommendation is the usage of PEST or PEST++, as they are versatile tools with extensive documentation
and instructions. (Less)
Abstract (Swedish)
Grundvattenmodellering används ofta av yrkesverksamma hydrogeologer för att ta fram predikt-
ioner som ska fungera som underlag i beslutsfattandeprocesser. I Sverige finns det inga lagar som reglerar hur en
grundvattenmodell ska utformas, och de kan se väldigt olika ut. Därför har syftet med detta arbete varit att under-
söka svenska grundvattenmodeller skapade av industrin i beslutsändamål under de senaste 11 åren för att undersöka
vilken metodik som använts för att skapa dem, för att sedan kunna ge rekommendationer på metoder förespråkade
av universitetsvärlden som hade gjort modelleringen mer robust.
I bakgrunden förklaras begreppen parametrisering, kalibrering, observationer, känslighetsanalys och osäkerhets-
analys, som alla... (More)
Grundvattenmodellering används ofta av yrkesverksamma hydrogeologer för att ta fram predikt-
ioner som ska fungera som underlag i beslutsfattandeprocesser. I Sverige finns det inga lagar som reglerar hur en
grundvattenmodell ska utformas, och de kan se väldigt olika ut. Därför har syftet med detta arbete varit att under-
söka svenska grundvattenmodeller skapade av industrin i beslutsändamål under de senaste 11 åren för att undersöka
vilken metodik som använts för att skapa dem, för att sedan kunna ge rekommendationer på metoder förespråkade
av universitetsvärlden som hade gjort modelleringen mer robust.
I bakgrunden förklaras begreppen parametrisering, kalibrering, observationer, känslighetsanalys och osäkerhets-
analys, som alla representerar olika val modellerare måste göra under modelleringsprocessen. Det är med avseende
på dessa begrepp som grundvattenmodellerna utvärderats. Osäkerhetsanalys och sannolikhet diskuteras i texten, då
den sortens deterministiska prediktioner som beslutsfattare gärna tar ställning till står i kontrast mot den sortens
stokastiska prediktioner som förespråkas av flera framträdande namn inom universitetsvärlden.
Resultaten visar att de undersökta modellerna generellt är lågparametriserade, att alla modeller inte är kalibre-
rade mot observationer, att antalet observationer som använts generellt är lågt, att känslighetsanalys endast genom-
förts för ett fåtal av modellerna, och att osäkerhetsanalys endast genomförts för ett fåtal av modellerna.
Jag föreslår användandet av pilotpunkter som ett sätt att öka parametriseringsgraden bland modellerna, använ-
dandet av datavärdesanalys för att avgöra om mängden observationer är tillräcklig bland modellerna, och användan-
det av PEST eller PEST++ som verktyg för att utföra kalibrering, känslighetsanalys och osäkerhetsanalys. Dessu-
tom uppmanar jag modellerare att, för ökad transparens, tydliggöra sina resonemang kring de val de gör under mo-
delleringsprocessen. Ett sätt att göra detta, och samtidigt öka reproducibiliteten av modelleringen, är genom använ-
dandet av skriptbaserad modellering. Utöver detta diskuteras lagstiftningen, eller avsaknaden därav, som ett poten-
tiellt hinder för mer sannolikhetsbaserad modellering.
Min slutsats är att det finns flera svagheter bland de undersökta modellerna, och i förlängningen potentiellt även
i övrig modellering som genomförts av industrin i Sverige under de senaste åren. Min huvudsakliga rekommendat-
ion är användandet av PEST eller PEST++ eftersom de är mångsidiga hjälpmedel med omfattande dokumentation
och manualer. (Less)
Please use this url to cite or link to this publication:
author
Haraldsson, Emil LU
supervisor
organization
course
GEOL02 20231
year
type
M2 - Bachelor Degree
subject
keywords
Hydrogeologi, grundvatten, grundvattenmodellering, beslutsstöd, litteraturstudie
publication/series
Examensarbeten i geologi vid Lunds universitet
report number
672
language
Swedish
id
9144643
date added to LUP
2024-01-12 13:07:52
date last changed
2024-01-12 14:06:52
@misc{9144643,
  abstract     = {{Groundwater modeling is often employed by professional hydrogeologists to make predictions to be
used by regulators in decision-making. In Sweden, there are no laws governing the way in which a groundwater
model may be created or shaped, and therefore they may differ wildly from one another. Thus, the goal of this study
has been to research Swedish groundwater models created by the industry during the last 11 years to evaluate which
methods have been used in their creation, to then be able to recommend methods favored by academia that would
make the modeling more robust.
In the background section, the phrases parameterization, calibration, observation, sensitivity analysis, and un-
certainty analysis are explained. They all represent decisions modelers have to make during the modeling process.
It is in regard to these phrases that the groundwater models have been evaluated. Uncertainty analysis and probabil-
ity are discussed throughout this text, as the type of deterministic predictions regulators usually prefer stand in stark
contrast to the type of stochastic predictions that are favored by multiple leading names in groundwater modeling
academia.
The results show that the researched models are generally parameterized sparingly, not all the models have been
calibrated, the number of observations used is generally low, sensitivity analysis has been completed for but a few
of the models, and that uncertainty analysis has been completed for but a few of the models.
I recommend the usage of pilot points as a means of increasing parameter dimensionality, the usage of data
worth analysis to deduce if there are enough observations in a model, and the usage of PEST or PEST++ to com-
plete calibration, sensitivity analysis, and uncertainty analysis. Furthermore, I encourage modelers to increase the
transparency of the modeling process by clarifying and highlighting the choices they make during modeling. One
way to do this and at the same time increasing the reproducibility of modeling, is through script-based modeling. In
addition to this, legislation, or the lack thereof, is discussed as a potential obstacle for more probability-based mod-
eling.
My conclusion is that there are multiple weaknesses among the evaluated models, and that, by extension, there
may also be weaknesses in other groundwater models produced by the industry in Sweden during recent years. My
main recommendation is the usage of PEST or PEST++, as they are versatile tools with extensive documentation
and instructions.}},
  author       = {{Haraldsson, Emil}},
  language     = {{swe}},
  note         = {{Student Paper}},
  series       = {{Examensarbeten i geologi vid Lunds universitet}},
  title        = {{Kan akademin hjälpa industrin utveckla mer robusta grundvattenmodeller? En studie av moderna svenska industriframtagna grundvattenmodeller}},
  year         = {{2024}},
}