Space for time – How do peatland vegetation patterns develop through time?
Posted On 21 January 2019
In this project (my GIS thesis) I used aerial photographs of peatlands (RBG and IR) to classify various landcovers and to understand how peatland vegetation patterns develop through time. I first mentioned it here.
Peatlands form a significant global terrestrial pool of carbon. Northern peatlands cover just 2 to 3% of the earth’s surface. However, nearly a third of the world’s terrestrial soil carbon is stored in these peatlands. These peatlands are also located in areas that are expected to have the highest increase in precipitation and temperature in the next decades.
Hummock-hollow patterns formation
Peatland ecosystems often show spatial vegetation patterns, such as regular string patterns of densely vegetated ridges (hummocks), perpendicular to the slope, which are interspersed with sparse vegetated pools (hollows). Here is what a maze-like pattern looks like from the top. In green the hummocks, in yellow and red the hollows. Trees are also visible, on ‘islands’ as well as on the hummocks themselves.
The patterning itself can be explained by a positive feedback between plant productivity and groundwater depth on elevated, drier sites, mainly due to increased production of vascular plants. The positive feedback is maintained because the vascular plants on hummocks attract nutrient flows in the water, due to differences in the transpiration rate. Although models suggest that small random hummocks will grow together as a peatland grows older, due to the positive feedbacks, there is insufficient empirical data on how these patterns form.
From carbon sink to source
As northern peatlands are located in areas that are expected to have the greatest increase in precipitation and temperature in the next decades, internal peatland dynamics may be affected in such a way that peatlands may switch from sinks to sources of atmospheric carbon.
However, peatlands are quite resilient to changes in climatic conditions. Nevertheless, this resilience is lost when environmental thresholds are passed due to changes in climate, which in turn may cause a shift to another stable state, with different surface patterns. This transition also results in different rates of carbon decomposition and sequestration; however, these systems are not fully understood yet. This transition is called a catastrophic shift, when an ecosystem shifts to an alternate stable state, which is not quickly reversed or cannot be reversed at all.
This catastrophic shift in relation to spatial vegetation patterns has been analysed already. Desert vegetation pattern development has been modelled and showed that they become more regular when a catastrophic shift to a barren state is approaching. If these patterns change predictably, they may be used as an early-warning signal for a nearing catastrophic shift. This raises the question whether peatland vegetation patterns could also be used as an early-warning signal for a catastrophic shift.
Different aged peatlands along a spatial chronosequence (a sequence of peatlands that differ in their profile due to differences in their age, also referred to as a Space-for-Time approach) can be analysed, and as such a larger spatial extent can be examined in a nondestructive way, at the same time. The chronosequence allows comparison of the links between vegetation and the environment. Aerial imagery of these differently aged peatlands can then be used to understand temporal processes. In an area where new land emerges due to isostatic rebound, such as Sweden and Finland, this method can be used to understand the dynamics of vegetation patterning. This is conveniant, as one of the processes of peatland formation is because of land becoming available due to glacial retreat.
Materials and methods
To establish a chronosequence, an area of isostatic rebound is required, which is the case for the Gulf of Bothnia, the sea between Sweden and Finland. the chosen study area is in the north-eastern province of Västerbotten in Sweden, located north/north-east of the city Umeå (63 49’ 42.31″ N, 20 15’ 34.99″ E). This area receives an annual average precipitation of 600 mm and has an average temperature of 2.6 C.
The specific study area is a transect from the coastland inwards and covers an area of 30.609 ha. The land cover in this area mainly consists of forest (23.795 ha) interspersed with many small to large lakes (983 ha) and peatlands (5.214 ha), with an occasional village (10 ha), agricultural lands (278 ha) and open fields (326 ha, Table 3.1). The transect rises from sea level to +260 m.a.s.l. nearly 40 km land inwards and varies from 5-20 km in width.
Four main types of data are used in this study, namely:
1) aerial images, which include red, green & blue (RGB) bands and one infrared (IR) band;
2) a digital elevation model (DEM);
3) vector maps such as a terrain map ; and
4) field data.
Without spending too much time on the details, in short the steps taken included:
The image below summarises the land cover classification throughout the transect. As peatlands develop, the amount of sedge decreases, while hummock increases. Hollows also increase, however, their percentage of land cover varies a lot with each age class. The amount of mudbottom remains stable, except for the 1000 years.
Here is an example of a some classified hummocks and hollows:
So now imagine many of these patterns throughout the transect. The easiest way to quantify them is with certain metrics. Here is an example:
The major range is an example of one of these metrics. It shows that young peatlands (2000-3000 years) have lower major ranges than older peatlands (up to 8000 years), although low values also exist in old peatlands. This is probably due to the age calculation, which only takes the elevation into account, not any new formation of peat through new dams for example.
This research aimed to understand the spatial vegetation pattern development through time, using highresolution images for a chronosequence of peatlands. These patterns were classified using supervised maximum likelihood classification and analysed using various pattern metrics based on hummock patches. Based on the results, the aerial images can be used to classify peatlands into five classes, which can then be quantified using pattern metrics. The hummock patches start off as small clumps of patches when the peatland is young, which grow together to form elongated patches as the peatland grows older.
This can be seen in the aerial images and its’ classification when going land inwards from the coast, as well as in the GAM figures. The minor and major range of the patches had the highest correlation with age, followed by the radius of gyration, which all increases with age. The fourth best correlation with age was the perimeter-area ratio, which decreased with age. Of the used terrain characteristics, the total catchment area above a eatland influences the pattern development the most, followed by the slope length above the peatland. There is a large variation in patterns and where in the landscape they develop, which also results in weak relationships.
Although the result showed that patterns change throughout the transect, the peatlands younger than 2000 years were underrepresented in the applied sampling strategy. This was caused by an insufficient presence of hummocks in the young peatlands, which were used in the pattern analysis.
If you would like to know more about this subject, don’t hesitate to contact me!