Skip to main content
SLU:s publikationsdatabas (SLUpub)

Sammanfattning

Monitoring programs designed for evaluation of trends rely on relatively frequent measurements in time but usually do not have spatial representation that is dense enough to connect them to explanatory variables that vary geographically. Other programs, like the so-called Omdrev-programs, monitor many objects, but with sparse temporal resolution and have therefore not been considered much in trend evaluations. In this report we evaluate the potential of geographically weighted regression models for trend analysis, including explanatory variables on three monitoring programs in Sweden: the Swedish Soil Inventory, the Swedish lake survey and the Swedish national environmental groundwater monitoring program, with measurements once each six or ten years for each of the objects.

Geographically weighted regression models (GWR, Brunsdon et al., 1998) have recently been adopted to allow the evaluation of regional trends (von Brömssen et al., 2023). They have also been tested to identify groups of observations which have similar levels or similar trends for a number of potential explanatory variables and exhibit similar trends in a response variable (von Brömssen et al., 2025). Both approaches rely on the principles of smoothing over windows of data which allows the use of not only the site of interest but also include sites that are either geographically or thematically similar. Since this approach allows an analysis even if the number of observations at a specific site is low, they are relevant for the analysis of trends in monitoring programs with low temporal but high spatial resolution. This report will focus especially on multiscale GWR (Fotheringham et al., 2017), since it is reasonable to assume that different explanatory variables work on different scale, e.g. that some might affect the response variable locally and others more regionally.

Trends in acidification were quite easy to evaluate during the 1980s and 1990s, where strong declines in acidifying variables were observed. Now, both the magnitude of trends in pH and SO4 is smaller and both the magnitude and direction can vary substantially over Sweden (von Brömssen et al., 2021) due to several underlying natural processes and anthropogenic activities (Grennfelt et al., 2020). Temporal trends in pH could be influenced by reduced levels of sulphur deposition (Garmo et al., 2014; Vuorenmaa et al., 2018) as well as by changes in climate and increased eutrophication (Minella et al., 2013). Levels of pH in Sweden are generally low due to the high levels of natural organic acids and can vary due to stream flow conditions (Erlandsson et al., 2011).

For the current report we will investigate if a single explanatory variable can be added to GWRs additional to time and study its influence on trends in pH. We will:

- investigate the usefulness of both basic and multiscale GWR to identify effects of a driver variable on trends in pH

- determine if data collected in typical “omdrev”-surveys are sufficient to obtain reliable results from multiscale GWR

- evaluating pH trends in different media - soil, surface and ground water

- provide open-source R code that can be used to study additional variables in these monitoring programs.

Publicerad i

Rapport (Institutionen för energi och teknik, SLU)
2026, nummer: 135
Utgivare: Department of Energy and Technology, Swedish University of Agricultural Sciences

SLU författare

Associerade SLU-program

Försurning

UKÄ forskningsämne

Markvetenskap
Oceanografi, hydrologi, vattenresurser
Sannolikhetsteori och statistik

Permanent länk till denna sida (URI)

https://res.slu.se/id/publ/146278