Replication Data for: Top-down vs. bottom-up control on vegetation composition in a tidal marsh depends on scale

Elschot, K. (Creator), Vermeulen, A. (Creator), Vandenbruwaene, W. (Creator), Bakker, J. (Creator), Bouma, T. (Creator), Stahl, J. (Creator), Castelijns, H. (Creator), Temmerman, S. (Creator), University of Groningen, 9-Feb-2017


  • Kelly Elschot (Creator)
  • Anke Vermeulen (Creator)
  • Wouter Vandenbruwaene (Creator)
  • Jan Bakker (Creator)
  • Tjeerd Bouma (Creator)
  • Julia StahlSOVON, Dutch Ctr Field Ornithol, (Creator)
  • Henk Castelijns (Creator)
  • Stijn Temmerman (Creator)
  • SOVON, Dutch Ctr Field Ornithol


==== Data collection and data files=====

==Study site==
The study area, Saeftinghe, is located in the Western Scheldt estuary in the Netherlands (51°21’N, 4°11’E). All datasets in this study were collected in this marsh, except for the total number of greylag geese in the Netherlands. It has a semidiurnal tidal regime with a mean tidal range of 4.9 m, and salinity varies between 5- 18 PSU. Permission to conduct this study was issued by Het Zeeuwse Landschap, which is the authority responsible for the protection of this national park. This study did not involve any endangered or protected species. Saeftinghe is considered as one of the largest brackish marshes in Western Europe, approximately 28 km2 in size, and is an important feeding habitat for large populations of wintering Greylag geese, Anser anser. While a small part of the marsh is grazed by cattle, the largest part is now abandoned but had been extensively grazed by sheep until 1993. Outside the cattle-grazed marsh, the most important vegetation types are dominated by Phragmites australis, Elytrigia atherica or Bolboschoenus maritimus. P. australis is mostly limited to the eastern part of Saeftinghe near the seawall, E. atherica mainly dominates the higher elevated levees bordering creeks and B. maritimus is mainly limited to the depressions between levees. This marsh has only become an important staging and wintering site for Greylag geese since the 1980s. Greylag geese prefer to feed on the below-ground storage organs of B. maritimus for which they grub into the marsh soil. A study by Castelijns et al. 1998 (Oriolus 64: 90-102) showed that during the winters between 1994 and 1997, the main food sources of Greylag geese in Saeftinghe consisted of the following: 49% tubers of B. maritimus, 33% above-ground plant parts of other marsh plant species, 10% agricultural plants (growing on arable fields adjacent to the marsh) and 8% seeds of E. atherica. They concluded this based on microscopic evidence found in fresh droppings that had been collected in the field together with field observations.

==Greylag goose numbers==
Dataset 1: Between 1987 and 2011, numbers of Greylag geese in Saeftinghe were estimated at monthly intervals from July until March of the following year by the local ‘natuurbeschermingsvereniging De Steltkluut’. The majority of the geese arrive in October and almost all leave again by the end of February, except for a small breeding population that remains year round. Goose numbers were estimated for the entire marsh area (28 km2) (data provided by Natuurberschermingsvereniging De Steltkluut). Goose counts were performed by several groups of people walking pre-determined routes, thereby covering the entire marsh area. By walking these routes simultaneously, they prevented that the same geese get counted multiple times. This data is found in: FILE: Greylag_Goose_counts.xlsx, SHEET: Greylag_Goose_counts_Saeftinghe. This data is collected and owned by The Steltkluutvereniging. Before using this data or for further details about this dataset please contact To include estimates of the population before 1987 in our analyses, we used population sizes estimated in the literature, which were estimated in Saeftinghe at monthly intervals as well: Castelijns H, Jacobusse C. Spectaculaire toename van grauwe ganzen in Saeftinghe. De Levende Natuur. 2010: 45–48.

Dataset 2: In addition, the numbers of Greylag geese were counted between 1980 and 2011, throughout the Netherlands, including nature reserves and farmland (that were known for being bird hotspots). These bird counts were performed between September and April the next year, in one pre-specified weekend and replicated each month. The calculated maximum of goose numbers for each year was provided by Sovon, Dutch Centre for Field Ornithology. This data is found in: FILE: Greylag_Goose_counts.xlsx, SHEET: Greylag_Goose_counts_NL. Before using this dataset or for further details please contact: To gain more insight in the Greylag goose numbers for different areas in the Netherlands you can also visit the website:

==Top-down control by geese grubbing at the local scale==
To determine the strength of top-down control by Greylag geese on the local vegetation, we used bare patches within the vegetation created through grubbing by Greylag geese and subsequent revegetation of these bare patches. Using ArcGIS, we analysed false-colour aerial photographs from 1979, 1990, 1998, 2004 and 2008. We identified bare patches in the eastern region of Saeftinghe, an area approximately 7 km2 in size, for which the most central point was 51°21’45N, 4°11’46E. These aerial photographs are georeferenced and can be found in the Saeftinghe.mxd GIS file in map: A shape file in this folder kuilenpunten.shp are the location of the bare patches identified. We used aerial photographs provided by Rijkswaterstaat (2004 and 2010 were already digitalized by Rijkswaterstaat, all other years were provided by Rijkswaterstaat but scanned and georefenced for this manuscript). The photo’s that we scanned for this manuscript were limited to the eastern marsh where my research area was focussed.

For each bare patch, we determined years that they were present as well as absent using the GIS file. Once the bare patch was no longer visible on the aerial photograph, we assumed vegetation had re-established at these bare patches. In this way, we could determine the minimum number of years that it had taken for vegetation to re-establish in each bare patch, i.e., the minimum regeneration time. For example, when a bare patch present in the photo taken in 1990 was no longer present in the photo taken in 2004, we assumed revegetation of this bare patch had taken at least six years when we measured the vegetation composition in 2010. As we did not have aerial photographs for every year, we established four classes of minimum regeneration time: 0, 2, 6 and 12 years.
After determining the coordinates from the photographs, we visited these regenerated bare patches in the field in July and August of 2010. We measured the vegetation composition in 2 m x 2 m plots using the decimal scale of Londo. Some identified patches were not bare patches but rather locations where two creeks met which simulated bare patches in the photographs. If so, then we did not measure the vegetation. In order to assess the impact of Greylag geese on the vegetation composition, we compared the vegetation composition in regenerated patches with the vegetation composition in depressions unaffected by geese. Thus, we measured vegetation composition in 13 control plots located in vegetation dominated by B. maritimus, where no visible signs of goose grubbing were present. As geese prefer to feed on the below-ground tubers of B. maritimus, we considered the B. maritimus-dominated vegetation type as the original vegetation type before geese activity resulted in the bare patches. As only three bare patches were recorded in the aerial photograph of 1979 that were no longer present in the aerial photograph of 1990 and therefore we excluded the minimum regeneration time of 20 years from further analyses. Additionally, we removed in total seven bare patches that got overgrown by expanding reed beds.
The vegetation data of the bare patches and the control plots can be found in the file: regeneration_patches_Saeftinghe.accdb that is located in the folder: GIS_files_aerial_photographs_and_bare_patches.

==Bottom-up control by sediment accretion at the landscape scale==
To determine the strength of bottom-up control on the marsh vegetation exerted by vertical sediment accretion, and the subsequent increase of marsh elevation relative to sea level as well as its effects on the succession of B. maritimus vegetation towards other vegetation types, we measured both changes in vegetation type as well as long-term changes in surface elevation. To determine the cover of different vegetation types in Saeftinghe, we used vegetation maps generated by Rijkswaterstaat that had been produced using a widely practiced and validated method [53]. First, aerial photographs from 1979, 1998, 2004 and 2010 were analysed. Based on the false colour ranges, different putative vegetation types were identified. This was followed by multiple vegetation composition measurements performed in the field for each of the identified putative vegetation types. Ultimately, specific vegetation types were linked to colour ranges in the aerial photographs, and used to generate vegetation maps that overlapped the entire marsh.
Generally, the pioneer marsh harbours a combination of Salicornia europaea and Spartina anglica, which is replaced by species such as Puccinellia maritima, Aster tripolium and Glaux maritima. Ultimately, the lower marsh becomes dominated by a cover of B. maritimus and the higher marsh by E. atherica and P. australis. In this study, we focused on the total cover of the three climax vegetation types, dominated by either B. maritimus, E. atherica or P. australis. When one of these three species was either dominant (at least 50% cover) or indicated as co-dominant in a vegetation type, we included that specific vegetation type in the analysis. All other vegetation types were excluded from the analysis.
Data provided by rijkswaterstaat were used for this analysis. These vegetation maps are located in the folder: Vegetation_maps_rijkswaterstaat. Here, you find both the maps that can be added to GIS as a layer. In the attached attribute table the cover of the different vegetation types are given. They use a typology called SALT97 in earlier photographs and SALT08 for the latest time steps (after 2008). The attribute tables were exported to an excel file which is added as: Calculated_cover_climax_vegetation types.xlsx in the folder: Vegetation_maps_rijkswaterstaat. Here, we calculated the covers which were used in our analyses. In the same folder pdf files are given with each map that explains what each vegetation type is. These are given in Dutch. For an English explanation you can use the following link:, here the RWS (=Rijkswaterstaat) SALT CODE NL explains the different codes as well. For more information you can also contact: or visit for an online view of the vegetationmaps provided by Rijkwaterstaat.

The distribution of marsh plant species is determined for a large part by the local marsh surface elevation and the marsh surface increases in elevation as marshes become older. Therefore, an area of approximately 2 km2 was used to estimate long-term changes in marsh surface elevation (51°21’48N, 4°11’15E). Data were available for the years 1931, 1951, 1963, 1992, 2004 and 2010. Data were provided as Digital Terrain Models (DTMs) with a resolution of 20 m x 20 m for the years 1931, 1951, 1963 and 1992. These were based on topographic and bathymetric surveys performed by the Dutch and Belgian waterway management authorities. Topographic surveys resulted in elevation data with a resolution of 1 measurement point per 7500 m-2. The elevations were mapped to the precision of 0.1 m relative to the Dutch Ordnance Level (NAP), which is close to mean sea level at the Dutch coast. This resulted in a maximum vertical error of + 0.05 m. For the more recent time steps 2004 and 2010, DTMs with a resolution of 2 m x 2 m were available based on LIDAR data. These LIDAR surveys were carried out during low tide with a resolution ranging from 1 point per 16 m-2 to several points per m-2 and a vertical accuracy of ±0.2 m. The channel networks for 1931 and 2010 were merged and used as a mask to exclude grid cells located within the tidal channel network. The changes in the creek edges between the tidal channel networks of 1931 and 2010 were fairly limited because of the slow migration rate of the channels. Thus, we used this mask for all the time steps, as the creek edges in the intervening time steps between 1931 and 2010 would also be located within this mask. Besides the mean platform elevation of the selected site, the standard deviation was also calculated to represent the spatial variation in the marsh platform elevation. Historical data on mean high water level (MHWL) and mean high water level at spring tide (MHWLS) were derived from the nearby tidal gauge station at Bath, the Netherlands. The marsh surface elevation and the levels of MHWL and MHWLS are expressed in metres above NAP. This data is given in: sediment_accretion_rate.xlsx. This dataset is also explained in: Wang C, Temmerman S. Does biogeomorphic feedback lead to abrupt shifts between alternative landscape states?: An empirical study on intertidal flats and marshes. J Geophys Res Earth Surf. 2013;118: 229–240. doi:10.1029/2012JF002474.
For more information you can contact:

Several datasets used in this chapter belonged to other institutes and cannot be used without their consent. Rijkswaterstaat provided the vegetation maps and Aerial photographs. Sovon provided the data on the geese maxima for the Netherlands and the Steltkluut vereniging provided data on the geese counts in Saeftinghe. The marsh accretion rate is estimated by Wouter Vandenbruwaene of the Flanders Marine Institute.
For additional information you can contact:
Date made available9-Feb-2017
PublisherUniversity of Groningen
Temporal coverage1980 - 2011
Date of data production1980 - 2011
Geographical coverageSaeftinghe, Western Scheldt estuary, the Netherlands (51°21’N, 4°11’E)
Access to the dataset Open

    Keywords on Datasets

  • vegetation composition , tidal marsh , Greylag geese
Related Publications
  1. Top-down vs. bottom-up control on vegetation composition in a tidal marsh depends on scale

    Elschot, K., Vermeulen, A., Vandenbruwaene, W., Bakker, J. P., Bouma, T., Stahl, J., Castelijns, H. & Temmerman, S., 3-Feb-2017, In : PLoS ONE. 12, 2, 17 p., e0169960.

    Research output: Contribution to journalArticleAcademicpeer-review

View all (1) »

View graph of relations

ID: 50055971