Vegetation traits of pre-Alpine grasslands in southern Germany

Study area

The study area is located in the TERENO Pre-Alpine Observatory^28,29 in southern Bavaria, Germany (Fig. 2). The ten test plots are situated on three sites at different elevations: Fendt (FE), Rottenbuch (RB), and Eschenlohe (EL). Table 1 gives an overview about the main characteristics of these sites and the plots.

Geologically, the plots in FE are located in the major structural unit of the molasse basin of the Bavarian Alpine foreland, while the other ones lie in the major structural unit of the Alps with the major tectonic units folded molasse (RB) and northern calcareous Alps (EL)³³. Glacial erosion and Quaternary deposition processes influenced most of the molasse area. Therefore, alluvial structures and moraines largely effect soil parent material here²⁸. The dominant soil types in the northern part of the study area are Cambisols, Luvisols, and Regosols, and in the southern part Rendzic Leptosols and Calcaric Cambisols. Gleysols and Histosols characterize areas along the course of rivers and areas of recent and paleo lakes²⁸. According to the Köppen-Geiger climate classification the study area has a warm temperate climate without a dry season and warm summers (Cfb)³⁴. Mean annual precipitation at the study sites varied between 1008 mm and 1419 mm³⁵, and mean annual air temperature between 8.0 °C and 8.6 °C³⁶ (see Table 1). The land cover around the study sites is characterized by a mix of pastures, natural grasslands, forests (needle-leaf, broad-leaf, mixed), discontinuous urban fabric, and in EL additionally peat bogs³⁷. All ten plots are situated on managed grasslands, the dominant land use around the study sites. The management intensity of the plots range from very extensive management with only one cut and no fertilizer application per year to intensive management with five cuts and five slurry applications per year (Table 1).

Sampling design

The field campaign with UAS flights and vegetation sampling took place on 24–25 April 2018 at ten different grassland plots (FE1, FE2, FE3, FE4, RB1, RB2, RB3, EL1, EL2, EL3). Plots were selected by i) visual characterisation of standing biomass to select plots that differ in management as well as soil nutrient and water status (based on talks with local farmers and corresponding field visits; no specific method was applied), and that fulfil other criteria such as ii) homogeneous, flat area, iii) accessibility (including permission by farmers), and iv) proximity of the plots to ideally cover several plots with one UAS flight.

When designing the sampling strategy, a perspective linkage of the sample data to Sentinel-2 images with a spatial resolution of 10 m × 10 m was taken into account. Therefore, we adapted the sampling strategy proposed by Baret et al.³⁸ for the validation of medium spatial resolution land satellite products. The authors suggested relatively flat and homogeneous validation sites of 3 km × 3 km for validating data of sensors with a spatial resolution of up to 1 km × 1 km. Their validation sites were sampled at several so called elementary sampling units (ESUs, 20 m × 20 m). These ESUs were spread across the validation site using a division of the site in nine 1 km × 1 km squares (three to five ESUs per square) with a higher sampling density in the central square (five to seven ESUs)³⁸.

In our study, we used 30 m × 30 m plots that were ideally sampled at 12 subplots (corresponding to the ESUs of Baret et al.³⁸) of 0.25 m × 0.25 m (Fig. 1). We divided each plot in nine equally sized squares of 10 m × 10 m, in which we randomly placed one subplot. Following the suggestions of Baret et al.³⁸, we sampled the central square with a higher density (i.e. with four subplots). Compared to Baret et al.³⁸ we targeted a smaller number of subplots per plot (12 instead of 30 to 50) as our plot size is notably smaller than their plot size and hence it is easier to select a homogenous area. During the sampling campaign, we needed to reduce the number of subplots in the central square for EL1 (11 subplots were sampled) and EL2 (9 subplots) due to time constraints. In all other plots, we sampled 12 subplots.

Preparations in the field

Some arrangements needed to be done in the field prior to sampling to prepare accompanying UAS flights. The resulting images of these UAS flights were used among others for retrieving the exact location of the subplots. After the localisation of the plots in the field (aiming for a north-orientation of one plot site), bright 0.5 m × 0.5 m flakeboards were distributed in the plots at the approximate locations of the subplots (Fig. 3). These flakeboards were used to identify the location of the subplots in the orthophotos that were generated from images of the UAS.

Additionally, several ground control points (GCPs) were distributed in the overflight area of the UAS. The exact location of the GCPs’ centre was measured with a Global Navigation Satellite System (GNSS) receiver (Viva GNSS GS 10, Leica Geosystems AG, Switzerland) in static mode for 10 minutes. The data from the GNSS was reprocessed with Leica Geo Office 8.3 software (Leica Geosystems AG, Heerbrugg, Switzerland) utilising reference data from the satellite positioning service of the surveying administration of the federal states of Germany (SAPOS) for the real reference stations 0285-Garmisch, 0270-Bad Tölz, and 1271-Weilheim. The reference data were obtained via the SAPOS website of Bavaria (https://sapos.bayern.de/). The accuracy of the used GNSS in post-processing mode is 0.003 m in horizontal direction and 0.005 m in vertical direction³⁹. The transformation of the corrected coordinates from ellipsoidal heights to physical (geoid-based) heights (height system: DHHN2016, EPSG 7837) was done with the online processing service “CRS-Transformation Bayern” from SAPOS (https://sapos.bayern.de/coord_tm.php). The transformation accuracy for this height transformation is 0.005 m⁴⁰.

The UAS flights were conducted after the preparation of the respective field site, followed by the field measurements and vegetation sampling. A RGB camera (Sony Cyber-shot WX 220, Sony Corp., Minato, Japan) mounted on a fixed-wing UAS (eBee, senseFly, Cheseaux-sur-Lausanne, Switzerland) was used to acquire high-resolution images of the study sites. In total, four UAS-flights were necessary to cover all ten plots – one in FE, one in RB and two in EL (EL-N, EL-S).

Acquisition of field measurements and field samples

The methods for acquiring in-situ data of canopy height, destructive vegetation sampling, and subsequent sample processing were adapted from the Integrated Carbon Observation System (ICOS) instructions for vegetation measurements in grasslands^41,42.

Canopy height measurements and sampling for biomass and element content measurements

After the UAS flight, first the subplot area was identified (0.3 m south of the corresponding flakeboard, one site centred and parallel to the flakeboard, see Fig. 3b). Second, the bulk canopy height of the grassland canopy within the subplot was measured with a platemeter, which had the same area as the subplot for destructive sampling (0.25 m × 0.25 m) and was build according to the ICOS instructions⁴¹. The plate of the platemeter was constructed from acrylic glass and weighed 1680 g. Third, a metallic sampling frame (size: 0.25 m × 0.25 m × 0.03 m) was put on the subplot. After verifying that the sampling frame was not sliding on the vegetation, the vegetation within the sampling frame was clipped down to stubble height (0.03 m) with a manual grass cutter for later determination of biomass and element contents. Finally, the clipped vegetation was put in a labelled paper bag, then in an airtight plastic bag and afterwards in a cooling box until further processing in the laboratory.

Sampling for leaf mass per area determination

Additional samples were taken outside the subplots (within a radius of 2 m; one sample per subplot) to determine LMA. First, the area percentage of the PFTs legumes, other forbs and graminoids of the 30 m × 30 m plot was visually estimated (rough estimation based on field observations, no specific method applied). Then, the corresponding number of samples for each PFT was determined in relation to the number of subplots. That is, if the percentage of PFTs is e.g. 50% graminoids, 25% legumes, and 25% other forbs, and we have 12 subplots, there were six samples for graminoids, three for legumes and three for other forbs in this plot. At each subplot we took one LMA sample for just one specific PFT. One LMA sample consisted of fully expanded, undamaged leaves of the selected PFT originating from different individuals. The number of leaves per sample varied between one and seven depending on the leave size and the species composition of the PFT. However, the species composition could just be considered for dominant and easily distinguishable species, as the team was not specifically trained in grassland botany. Therefore, multiple species samples occurred mainly for forb samples. The clipped leaves were wrapped in humid paper. Then, the sample was put in a labelled plastic bag which was hermetically closed before it was placed in the cooling box.

Sample processing in the laboratory

After transportation, the samples were stored in the fridge and/or in a cooling room at 4 °C until further processing. Sample processing in the lab started within one to nine days after the sampling date.

Samples for the determination of biomass, green area index, and element content

The samples were removed from the plastic bags and weighed. The weight of the fresh material was multiplied by 16 to determine the fresh weight per 1 m² (Biom wm). Afterwards, the sample material was sorted into the PFT non-green vegetation (NG; representing photosynthetically inactive structures), legumes (L), non-leguminous forbs (F), and graminoids (G). Then, the fresh weight of each PFT was determined. The sorting was done either on the full sample or on a representative smaller subsample in case there was a lot of sample material. For taking a subsample, the vegetation material was thoroughly mixed and then a handful of material was taken out. The remaining material was dried and further processed/analysed for C and N content like the PFT specific samples.

The total hemi-surface area of the green material (GA) needs to be obtained to determine the GAI of a sample. As we have flat vegetation structures, we used a planimeter (LI-3100A Area Meters, LI-COR, USA) to obtain the hemi-surface area. The hemi-surface area was measured separately for each PFT of a sample, but not for the non-green vegetation. The GAI of a certain PFT (GAI_PFT) was then calculated by dividing the hemi-surface area of the PFT (GA_PFT) by the area of the subplot (A; 0.25 m × 0.25 m) and in case a subsample was taken by multiplying with the ratio of the fresh weight of the sample (Biomass wm_sample) to the subsample (Biomass wm_subsample):

The total GAI (GAI_tot) of the sample was calculated by summing the GAI of each PFT:

After the planimeter measurements the samples were dried in an oven at 65 °C until constant weight was achieved. Then, the dry weight of each PFT of each sample was measured (dm_PFT). The total weight of dry biomass of a sample (Biomass dm) was obtained by summing up the dry weights of all PFTs and if applicable the dry weight of the remaining material (dm_rest) of a sample and then scaled to 1 m²:

The percentage of each PFT (P_PFT) with respect to the total dry biomass weight of the sorted material (either subsample or full sample) (Σdm_PFT) was calculated as follows:

The plant water content (PWC) was calculated from the weight of the fresh (Biom wm) and dry biomass (Biom dm) as follows:

Finally, the dried vegetation material was ground in a ball mill for elemental analysis of C and N.

Samples for the determination of leaf mass per area

After returning from the field work, the leaves for the LMA determination were rehydrated until full turgescence. When the leaves were fully expanded, the last mature leaf from each tiller (in case tillers were sampled) were separated and the petiole was recut at the base of the leave blade. Then, the hemi-surface area of each sample was determined with the planimeter. Afterwards, the samples were dried in an oven at 65 °C until constant weight was achieved, before the dry weight of the LMA samples was measured. The LMA of a sample i was calculated by the ratio of the leaf dry weight (W_i) to its fresh area (A_i):

Analysis of C and N content

The C and N content of the milled vegetation samples was determined using an elemental analyser (varioMax CUBE, Elementar Analysesysteme GmbH, Germany) operated in CNS (carbon, nitrogen, sulphur) mode with the plant method and a weighted sample of 17 mg at the laboratory of the Technical University Munich, Chair of Soil Science, in Freising (Germany). The detection limits of the instrument were 0.020 wt.% for C and 0.015 wt.% for N.

Due to the sorting of the sample in different PFTs, the C and N contents were obtained specific for each PFT. In cases where a subsample was taken for the sorting into PFT, C and N contents were additionally measured also for the mixed remaining sample material.

Mean concentrations of C and N for each subplot (plant community C and N) were calculated as follows:

where (bar{E}) is the mean element content (C or N), P_PFT is the percentage of the PFT (NG = non-green, L = legumes, F = non-leguminous forbs, G = graminoids), and E_PFT is the element content (C or N) in the PFT.

Retrieving the coordinates of the subplot centres

The single images from each UAS flight were processed with the photogrammetric software PIX4D (Pix4Dmapper Pro, Pix4D S.A., Prilly, Switzerland) to obtain orthophotos. The GCPs in the orthophotos were used to georeference the orthophotos. The final spatial resolution of the orthophotos was 0.036 m (FE), 0.034 m (RB), 0.030 m (EL-N), and 0.043 m (EL-S).

Afterwards, the coordinates of the subplots centres were manually extracted from the high-resolution georeferenced orthophotos utilizing QGIS (Version 3.0.0-Girona)⁴³. The bright flakeboards were visually localised in the images. Then the subplots centres were identified (0.425 m perpendicular away from the middle of the southern flakeboards site) and their coordinates extracted.

Characterisation of grassland type and plant community composition on the plot-level

Vegetation relevés were carried out on 18–19 June 2020 for each of the ten 30 m × 30 m plots. Within each plot all vascular plant species were systematically determined and their cover was visually estimated to the nearest percentage as a proxy for abundance. Data is provided as percentage cover per plot. Plant species names were updated according to The Plant List, a working list of all known plant species, aiming to be comprehensive for all species of vascular plants, including flowering plants, conifers, ferns and their allies, and of bryophytes, including mosses and liverworts (theplantlist.org). The grassland type was classified sensu Oberdorfer (1977 and updated since)⁴⁴. Please note that vegetation relevés from 2020 may slightly differ in relative coverage to actual species specific data from sampling in 2018. Nevertheless, most grassland species reach life spans of several decades and persist through time.

Source: Ecology - nature.com