BIODESERT is a flexible and scalable survey of multiple vegetation
attributes, soil properties, ecosystem processes and livestock
grazing-related variables that has been conducted in drylands worldwide
between 2016 and 2019. This survey is coordinated by the Dryland Ecology and
Global Change Laboratory at the University of Alicante (Spain,
The power and relevance of the data to be obtained by a survey like
Flowchart of the actions and measurements involved in the
BIODESERT survey. The numbers indicate the sections of the article where
each part of the survey is described. N
The flowchart of the BIODESERT survey is shown in Fig. 1. Fieldwork has been carried out with basic field equipment (tape, quadrats of different size, a compass, a clinometer, soil corer, hammer, plastic bags and a ruler), and no sophisticated field equipment or sensors were required to successfully execute it. We have done so deliberately, as this field survey needed to be conducted easily by researchers from across the world, who in many cases work in areas that are remote and difficult to access and where there are no nearby research facilities. It was also our intention to minimize the cost of doing the field survey as much as possible to ensure the widest participation possible, particularly for colleagues from the Global South and other dryland areas underrepresented in previous studies. However, these researchers needed to have local botanical knowledge and a good understanding of livestock characteristics at their sites as the identification of the plant species and of the grazing pressure present at each field site is an essential part of the survey.
The BIODESERT global survey was conducted at 98 sites located in 25 countries from 6 continents (Algeria, Argentina, Australia, Botswana, Brazil, Canada, Chile, China, Ecuador, Hungary, Iran, Israel, Kazakhstan, Kenya, Mexico, Mongolia, Namibia, Niger, Palestine, Peru, Portugal, South Africa, Spain, Tunisia and the United States of America; Fig. 2). This survey, and the processing of all the data and samples it has provided, has involved over 300 researchers, field assistants, students, laboratory technicians and site/laboratory managers working in over 100 institutions from 30 countries.
Location of the 98 sites surveyed with examples (insets a–g) of
local grazing gradients. Each white dot represents a 45 m
We surveyed multiple (from two to four, see below) 45 m
Hypothetical example of the survey scheme, showing the location of three survey sites (blue dots in the left panel) and the location of plots along watering points (WPs) within one of these sites. In the right panels, the image in the center shows the location of three plots, and the upper and lower images show the plots located at 250 and 2000 m from the watering point, respectively.
While the evaluation of grazing pressure at each site based on a combination of the distance to a water point and expert knowledge may seem subjective, each team doing fieldwork was very familiar with the main features of the grazing system present in the plots they surveyed. In eight of the sites surveyed, local grazing gradients were not established using water points but rather using fences separating paddocks with different grazing intensities. In any case, plots at each site were established in areas representative of the vegetation and soil types that can be found there, so the impacts of grazing pressure could be assessed without confounding factors associated with variation in soil, climate or vegetation type. Selected water points were separated from other watering points or elements that could interrupt the movement of livestock (e.g., fences) by at least 1 km to avoid confounding effects that could alter the effects of the distance from the water point on the ecosystem structural and functional attributes to be measured.
We conducted multiple validation tests of the heuristic value of grazing pressure obtained at each plot by (i) measuring recent grazing pressure in situ in all plots by counting, identifying and weighting the dung or pellets of all herbivores within quadrats (see Sect. 2.3.1 below), (ii) using livestock density data from local statistics, observations or interviews with local farmers (whenever available), and (iii) measuring the width and depth of all livestock tracks crossing each 45 m transect to derive a total cross-sectional area of livestock tracks for each site (see Sect. 2.3.2 below). All validation tests conducted, described in detail in Maestre et al. (2022), demonstrate that our approach quantified the full spectrum of grazing intensities found at each site.
The large variability in the characteristics of extensive grazing systems
present across the drylands surveyed did not allow us to find all grazing
intensities (ungrazed and low, medium and high grazing pressure) at all sites.
A total of 52 of the 98 sites surveyed had three plots corresponding to low,
medium and high grazing intensities. An additional fourth plot in an
ungrazed area was also surveyed at 35 of the sites. An ungrazed control plus
two additional grazing levels (medium and high or low and high) were
surveyed at eight sites. In three sites, only two plots could be surveyed
because they did not have either low or medium grazing areas. In total, 326
45 m
All measurements were taken within 45 m
Example of a transect line placed in the field. Photograph by Matthew Bowker.
Six different vegetation surveys took place at each plot as described in the next subsections: (i) “patch”–“inter-patch” survey with the line-intercept method (mandatory), (ii) cover and diversity survey with quadrats (mandatory), (iii) cover and diversity survey using the line-point intercept method (mandatory), (iv) plant functional trait survey (mandatory), and (v) facilitation survey (optional). All the vegetation surveys took place at each site during the peak of the growing season or immediately just after.
A patch is defined as a long-lived feature such as perennial plants and shrub branches making contact with the soil that is able to collect runoff water, as well as entrained sediment and nutrients (Tongway and Hindley, 1995). Patches are structures that accumulate resources by restricting the downslope flow of water, topsoil and organic matter and are separated from each other by areas of bare soil or annual vegetation (inter-patch zone). Inter-patches are characterized as zones where resources such as water, soil materials and litter are freely transported either downslope, when water is the active motive agent, or down wind, when eolian processes are active (Tongway and Hindley, 1995, 2004). Patches can comprise physical features such as furrows or bays created by active land-forming processes or biological features such as plants or fallen logs. The deposition of alluvium or litter is a common identifying factor in helping to recognize patches in the field (see Tongway and Hindley, 2004, for additional guidelines to assist the identification of patch and inter-patch zones in situ). While multiple types of inter-patch zones can be established, for the sake of simplicity, and to obtain data as standardized as possible among sites, we only differentiated two types of inter-patches: bare soil and rock outcrops. We considered inter-patches dominated by annual plants or biocrusts as bare soil (but with a cover of annuals or biocrusts) because our study focused on perennial plants.
The use of the line-intercept method allowed us to collect a continuous record of patch (perennial plants, accumulations of permanent litter, stems and logs placed on the soil surface) and inter-patch zones along each transect (Fig. 6). Once we identified a patch, we established its boundaries and measured its length along the transect and its width at right angles to the transect line (Fig. 7). Clumps of grasses, or of grasses and small shrubs growing closely together or connected with litter bridges, were considered to constitute a single patch. See Fig. 8 for an illustrated example of the patch–inter-patch survey.
Illustration of how to measure individual grasses when they form the patches on a monitoring transect. Redrawn from Tongway and Hindley (2004).
A diagrammatic illustration showing how to measure the length (along the transect) and the width of patches (right angles to the transect). Redrawn from Tongway and Hindley (2004).
Illustrated example of a transect and its information once recorded in the field data sheet. Redrawn from Tongway and Hindley (2004).
Along each of the four transects located in the field, we also placed 25
consecutive quadrats (1.5 m
The line-point intercept method (Levy and Madden, 1933) is a quick, easy and accurate method for quantifying vegetation cover. This survey complemented the information provided by the line-intercept and quadrat surveys described in the preceding sections and allowed us to compare our results with those obtained using other widely used dryland monitoring methods (Herrick et al., 2005; Oliva et al., 2020). In this survey we gathered information on three vegetation categories (Fig. 9a) according to the guidelines provided by Muir et al. (2011).
Along each of the four 45 m transects established in each plot, a measuring tape was fixed as tense and as close to the ground surface as possible (Fig. 5, thread under shrubs using a steel pin as a needle). A measurement was always recorded for the ground cover category. Measurements were made for midstory/overstory categories only in those cases when these elements were observed. Every 20 cm along the tape, a metal needle (2 m height) was dropped vertically, and the relevant cover type was recorded in the appropriate category (Fig. 9b):
“Green leaf” is a leaf with green pigmentation (one that is actively
photosynthesizing) that is attached to the plant. “Dry leaf” is a leaf with non-green pigmentation (one that is not actively
photosynthesizing). This can include senescent (alive) vegetation as well as
dead vegetation. It must be attached to the ground or plant.
Examples of bare ground areas with lichen-dominated
cyanobacterial biocrusts
When midstory plants were present on the transect, we recorded them using the same needle directly above the position used with ground story cover. The species name and its status (green leaf, dry leaf or branch) were recorded. When Overstory vegetation was present, we recorded the species name and its status (dead or alive; if alive we recorded if the pin projected over a green leaf, dry leaf, branch or canopy gap; Fig. 9a).
We aimed to estimate the trait distributions that characterize the functional diversity exhibited by the dryland plant communities surveyed while considering the intraspecific trait variability and the spatial distribution of species traits within each plot. These measurements can provide important insights on the drivers and consequences of both aridity and grazing pressure on ecosystem structure and functioning (see Gross et al., 2013, 2017; Valencia et al., 2015; García-Palacios et al., 2018; Le Bagousse-Pinguet et al., 2017, 2019, 2021 for examples). All the protocols described followed the standardized protocols commonly used in trait-based research (Pérez-Harguindeguy et al., 2013). The plant trait survey involved the following steps:
Five 1.5 m In each quadrat, the cover of all perennial species was noted (these
measurements were obtained from Sect. 2.2.2 described above). In each quadrat, the tallest and most well-developed individual of each
species was selected. Functional traits were measured on the selected individuals.
Traits related to plant size (maximum vegetative height, plant lateral spread and maximum leaf length) were measured as follows:
Measuring plant height
Once plant and leaf size measurements were completed, leaves of the selected individuals were harvested and taken to the lab for subsequent leaf trait measurements. We sampled and stored leaves for trait measurements following these guidelines:
We only measured the last mature (i.e., well-developed) leaves at the top of
each individual to avoid shaded leaves. For perennial plants with evergreen
leaves, we avoided measuring old leaves from past growing seasons. We also
avoided damaged leaves, such as those that showed signs of herbivory,
senescent leaves, or leaves with the presence of fungi or other
pathogens/parasites. For plants with large leaves, we harvested at least two leaves per
individual. For small leaves, we sampled enough leaves to obtain Once harvested, leaves were placed in a wet paper towel and stored in a tagged
plastic Ziploc bag with a little water to prevent dehydration. These plastic
bags were stored in a cooling box for transportation to the lab. We used
only one zip bag per sampled individual and tagged the bag with the name of
the site, transect and quadrat number, as well as with the genus and species name.
Once in the lab, the plastic bags and leaves were stored in a refrigerator
in the dark before analyses and, ideally, processed within 48 h of sampling.
Leaf traits measured in the laboratory included leaf area
(cm
We first measured leaf fresh mass (mg). To do this we carefully removed excess of water from the leaves with a dry paper towel and weighed all the leaves of a given sampled individual using a precision balance. This gave us a measure of fresh leaf mass. It is critical that any excess water is removed before measuring fresh weight.
After measuring fresh mass, all the leaves of the sampled individuals were
placed onto a sheet of white paper, avoiding any leaf overlap, and both a 5 cm
Examples of pictures taken for leaf area measurements. Photographs by Nicolas Gross.
This survey was set up to obtain a proxy of the magnitude and direction of biotic interactions among vascular plants using an observational and comparable approach (for examples on the use of this type of data, see Soliveres et al., 2014; Xu et al., 2015; Berdugo et al., 2017, 2019). We first randomly selected 25 discrete perennial plant patches (“nurse plants” sensu Soliveres et al., 2014) for each of the dominant species within each plot. This resulted in 75, 50 or 25 plants per plot if there were three, two or one dominant species, respectively (Fig. 13a). If a given plot did not contain 25 patches of each dominant species, all available patches from the dominant species were surveyed until at least 25 patches in total were selected. We evaluated the cover (and number of individuals, whenever possible) of all perennial species growing beneath each nurse plant. To obtain comparable samples of species richness and abundance in surrounding open areas, the shape and area of each sampled nurse plant were delineated using wire rings. The rings were randomly placed in open areas away from each sampled nurse, and the cover of every perennial species identified within each ring was assessed (and the number of individuals counted when possible). The rings in open areas were located at least one patch radius away from the edge of the nurse plant (Fig. 13a).
Example of layout of sampling areas in a plant patch and in an
adjacent bare ground area (“Open”)
Sometimes we found a situation like that presented in Fig. 13b, in which one of the nurse
species formed large patches, and it was not possible to find an adjacent and
continuous open area of the same size. In this case, we surveyed
quadrats/rings located in adjacent open areas until an area of the same size
of the nurse plant was surveyed. For instance, if the nurse plant of Fig. 13b labeled as “Patch 1” had an area of 10 m
While we used the distance from water points as a surrogate of grazing pressure, we also conducted in situ assessments of recent and historical grazing pressure to verify the presence of a grazing gradient at each site and to obtain alternative estimates of grazing pressure, which were used to test the suitability of the BIODESERT sampling design (see Maestre et al., 2022, for a validation of grazing pressure estimates).
To assess recent grazing pressure, we identified and counted the dung or
pellets of all herbivores, an approach widely used to estimate the abundance
of wild herbivores and livestock, including cattle, sheep, deer, kangaroos
and other large mammalian herbivores such as giraffes, across terrestrial
ecosystems (Plumptre and Harris, 1995; Piana and Marsden, 2014; Bisigato and
Bertiller, 1997; Marques et al., 2001; Miyashita et al., 2008; Johnson and
Jarman, 1987). Within each 45 m transect surveyed, we placed two 25 m
At each plot, we calculated dung/pellet mass using either direct
measurements or estimates using dung/pellet counts. In some plots we
collected, dried and weighed all dung (or pellets) found in the quadrats
and expressed it as a mass per square meter for each plot and herbivore type.
This was typically done where dung mass was low or where the main herbivores
did not produce clearly defined pellets (such as horses, cattle, donkeys,
giraffe, elephants, buffalo, camels, hartebeest, wildebeest and zebra).
Alternatively, we counted dung and pellets of each herbivore in all quadrats
but collected it from four large (25 m
We quantified the size and density of livestock tracks (semipermanent landscape features that are formed when livestock traverse the same path to and from water; Lange, 1969) as a measure of historic grazing by livestock. These tracks are clearly visible over many decades and become wider and deeper as the intensity of livestock grazing increases. The intensity and size of livestock tracks are therefore a useful indicator of the history of livestock grazing (Pringle and Landsberg, 2004; Val et al., 2018). To assess the level of historic grazing pressure, we measured in all plots the width and depth of all livestock tracks crossing each of the 45 m transects to derive a total cross-sectional area of tracks for each site. As a note of caution, this technique is not very useful in sandy soils where tracks tend not to form or are not sustained over time. Thus, at those plots with sandy soils we validated grazing intensity with the previous methods.
Soil sampling was conducted in all the plots surveyed for vegetation during the dry season (summer). The soils were dry at the time of sampling, and no rain occurred in the surveyed plots at least 15 d prior to the soil survey. This survey was conducted following a stratified random sampling according to the following protocol and guidelines.
When only a dominant perennial plant species was present in a plot, five
sampling points were randomly placed in open areas devoid of perennial
vascular plants (cover
Examples of placement of sampling points in open (blue) and vegetated (green) microsites in different dryland vegetation types (in the examples shown there are no biocrusts). Please note that the location of the sampling points in the pictures is made for illustration purposes only and are not located at a proper scale. A minimum separation distance between samples and between open and vegetated microsites should be established (see the main text for details). Photograph credits: Juan Gaitán, Eduardo Pucheta, David Eldridge and Mohamad Mjanku.
Examples of semi-arid shrublands
When more than one dominant species was present (e.g., as in Fig. 13),
sampling at plant microsites was conducted in five randomly selected
individuals of the dominant species (one sampling point per individual),
recording the species under which each sample was collected. When biocrusts were present (Figs. 9 and 15), five additional sampling
points were placed in each biocrust-dominated area (biocrust microsite
hereafter). Before collecting the soils from biocrust microsites, a 20 cm Each sampling point was separated by at least 3 m to avoid spatial
dependency. Sampling points in open and biocrust microsites were also
separated from the nearest plant patch by at least 1 m. When this was not
possible due to the small-scale pattern of vegetation, samples were taken
beyond the influence of the nearest plant canopy (at least one patch radius
away from the edge of the nearest perennial plant patch). A composite sample obtained from four soil cores (0–7.5 cm depth; Fig. 16)
was collected at each sampling point, bulked and homogenized in the field
(
All soil samples were stored upon field collection in properly labeled Ziploc-type plastic bags (one sample per bag). In the case of samples from plant and biocrust microsites, the name of the plant species or biocrust type (moss, lichen, cyanobacteria, mixed; Fig. 10) was included on the bag label. During the transportation to the laboratory, the soil samples were maintained, when possible, in a cooler with ice packs.
Once in the laboratory, soil samples were sieved (2 mm mesh) and air-dried
at room temperature for 1 month. In samples obtained from biocrust
microsites, visible biocrust constituents (e.g., lichens and mosses) were
carefully removed with a knife (avoiding losing soil) before air-drying the
samples. After sieving but before air-drying the samples, 40 g of sieved (2 mm mesh) soil of each sample was introduced in a plastic bag, labeled
properly and frozen at
Additional sampling was conducted to measure soil bulk density at every
site. Bulk density is the weight of soil for a given volume, and it is used
as a measure of soil compaction and to estimate nutrient stocks. Bulk
density was measured at each plot following the cylindrical core method
(Arshad et al., 1996). Volumetric soil cores were collected to 7.5 cm depth
with a cylindrical corer to calculate soil bulk density (the mass of soil
for a given volume, Mg m
Thus, for each plot we obtained three sets of samples:
All dried and frozen samples were shipped (following current national and international regulations and with the proper import/export permits) to the Dryland Ecology and Global Change Laboratory in Spain, where they were either analyzed or distributed to other laboratories.
Measuring soil nitrogen (N) supplies with ion exchange membranes (IEMs; Subler et al., 1995; Fig. 17) has become a popular approach for measuring in situ soil nutrient availability for plants and microbes in dryland ecosystems (e.g., Drohan et al., 2005; Castillo-Monroy et al., 2010; Durán et al., 2013, Delgado-Baquerizo et al., 2014). IEMs exhibit surface characteristics and nutrient uptake phenomena that simulate a plant root surface (Drohan et al., 2005). Unlike traditional measurements, results obtained by resins integrate nutrient diffusion capacity in soil, and studies have shown that ion concentrations in resin extracts correlate to the plant uptake of such ions (Qian and Schoenau, 2007). Overall, IEMs provide a reliable method to estimate inorganic N and P production in soils (Subler et al., 1995) and are a good alternative to more destructive and time-consuming traditional methods for estimating soil inorganic N and phosphorus (P) availability (Durán et al., 2013).
View of ion exchange membranes placed in the field on the left and after field collection on the right. Photographs by Andrea P. Castillo-Monroy.
We conducted an IEM survey following a stratified random sampling according to the same guidelines described in Sect. 2.4 for soil sampling. At each sampling point, this survey was conducted as follows:
One anion and cation IEM (types AMI-7001S and CMI-7000S, Membranes
International Inc, Ringwoods, NJ) was carefully inserted 5 cm into the
soil using a small spatula or knife to limit soil disturbance. Within each
sampling point, IEMs were located 10 cm apart (Fig. 17). Once installed in
the field, the location of the IEMs was recorded with a GPS and marked
with a metal stick/tag to avoid loss of membranes during the retrieval
phase following incubation. IEMs were incubated in the field for 4 weeks. This incubation was done
during the main rainy season. After the incubation period, IEMs were carefully removed from the soil using
a knife or a small shovel, taken to the laboratory and dried at ambient
temperature for 2 weeks. Once dried, the IEMs were carefully brushed to remove soil particles, placed
into separate plastic bags, and shipped to the Dryland Ecology and Global
Change Laboratory for analyses.
We quantified biological activity in situ using the tea bag index technique (Keuskamp et al., 2013). This measurement gives us insights into potential leaf litter decomposition at each site and complements the information obtained from the soil sampling and IEMs described in Sect. 2.4 and 2.5. This survey was conducted using commercial tea bags from Lipton (Lipton Green tea, EAN 87 22700 05552 5, and Rooibos tea, EAN 87 22700 18843 8) and were provided to all the participating groups by the Dryland Ecology and Global Change Laboratory, so the same material was used across all the plots where these measurements were conducted.
This part of the survey was conducted following the same stratified random sampling detailed for the soil and IEM surveys (five sampling points in each of bare, plant and biocrust microsites; the same sampling points were used for deploying IEMs and tea bags). At each sampling point, this survey was conducted during the main rainy season following this protocol:
The initial weight of the tea bag was measured (0.000 g). The bag plus string weight without content ( Two Lipton tetrahedral bags (one of green tea, the other of rooibos) were
properly marked and buried in an 8 cm deep hole, which was filled with soil
while keeping the plastic label visible above the soil surface. Once installed in the field, the location of the tea bags was recorded with
a GPS and/or marked with a metal stick/tag, to avoid loss of membranes
during the retrieval phase following incubation. Tea bags were retrieved after incubating them for 90 d and placed in
paper bags properly labeled until lab processing. In the lab, soil particles adhering to the tea bags were carefully removed. The whole tea bags were dried in an oven at 70 Dried tea bags were placed in a Ziploc bag, labeled and shipped to the
Dryland Ecology and Global Change Laboratory in Spain.