In compilation of WetCID for the 27 wetlandscapes, we employed three sources of primary data. These were (1) local site survey data, depicting general characteristics of each wetlandscape (catchment) and its geographical characteristics (including shapefiles for its spatial extent) and associated hydrological, climate and land-use conditions and their observed or perceived changes; (2) gridded historical data time series of monthly precipitation and temperature from Climate Research Unit Time-Series version 4.02 (CRU_TS4.02; Harris et al., 2014); and (3) historical data of annual land cover and its changes from the NOAA-HYDE dataset provided by the NOAA's National Climate Data Center (Jain et al., 2013; Meiyappan and Jain, 2012).

The survey for local site data was given to researchers within the Global Wetland Ecohydrological Network (GWEN) (http://www.gwennetwork.se/, last access: 9 May 2020). The GWEN researchers responding to the survey specified the relevant wetlandscape extent (total hydrological catchment with wetlands) and provided boundaries in GIS format for the 27 wetlandscapes, located as shown in Fig. 1. Information and data of all three types (local survey-based, hydroclimate, land use) were collected and synthesized for each of these wetlandscapes from all three sources (1–3). In addition to the local survey information, data on hydroclimate and land-use variables were thus also compiled from the global datasets in both gridded and aggregated form for each wetlandscape, as described further in the following.

A questionnaire for collecting local site knowledge and information on the availability and accessibility of local data was developed during a GWEN workshop held in Santa Marta, Colombia, on 24–28 April 2018. The questionnaire was sent out by email after the workshop to all participating GWEN researchers. The researchers responding to it related their answers to a specific wetlandscape in which they had active research.

The questionnaire comprised two main parts. Part 1 contained general questions about the geography, climate, hydrology, and wetland-relevant human activities and changes in the wetlandscapes. Part 2 focused on the availability and accessibility of local site data, structured into “Hydroclimate”, “Land use” and “Other” data (see templates in the database files for a full outline of the questionnaire). The collective knowledge obtained on conditions and changes in the 27 wetlandscapes and on data availability–accessibility is summarized in Sect. 3.1.

To complement this local knowledge and information basis, we further extracted and synthesized data for the 27 wetlandscapes from relevant global hydroclimate and land-use datasets as described below.

The temperature and precipitation data taken from the CRU_TS4.02 global datasets (Harris et al., 2014) covered a 30-year period (1981–2010), to be consistent with the time span of existing global land-use change data. CRU_TS4.02 provides hydroclimate data with spatial resolution of 0.5∘×0.5∘ and at a monthly temporal scale. In preparing temperature and precipitation datasets for each wetlandscape, the gridded data within the area of the wetlandscape were extracted from the global datasets and also spatially aggregated over that area, based on area-weighted averaging over the grid cells covered by the shapefile of each wetlandscape (catchment). This provided wetlandscape-specific data time series for each variable at each grid cell and aggregated over the whole wetlandscape. To facilitate analyses at different spatial resolutions, both the gridded and the aggregated time series were included in WetCID for each of the 27 wetlandscapes.

In addition to the gridded and aggregated data time series, period-specific temperature and precipitation changes were also calculated for each wetlandscape by dividing the total 30-year time span of the collected data into the two 15-year periods, 1981–1995 (Per1) and 1996–2010 (Per2). Such period-specific change quantification can facilitate relatively simple and straightforward analysis of how these hydroclimatic changes correlate with and may have driven other wetlandscape changes (e.g., in runoff, evapotranspiration, wetland area) between the same time periods (Destouni et al., 2013; Jaramillo and Destouni, 2014, 2015). Absolute and relative (%) changes between these periods (AbsChng and RelChng, respectively) were calculated from the mean annual values of temperature and precipitation during Per1 and Per2 as 1AbsChng=VarPer2‾-VarPer1‾,RelChng=VarPer2‾-VarPer1‾VarPer1‾×100, where VarPer1‾ and VarPer2‾ are average temperature (in C∘) or precipitation (in mm yr-1) over Per1 (1981–1995) and Per2 (1996–2010), respectively. Equation (1) was applied to both temperature and precipitation data to calculate their absolute changes in each wetlandscape, while Eq. (2) was only applied to precipitation data to calculate the corresponding percentage change in precipitation.

The NOAA-HYDE dataset was used to estimate land uses and their changes in each wetlandscape. NOAA-HYDE estimates annual changes in land-cover area over the global land mass, starting from a base map for the year 1765. The estimations follow a predefined pathway, determined by relevant land-use and management datasets (cropland, pastureland, urbanization, timber harvesting), to obtain forest area distributions close to satellite-based estimates of forests in recent years (Meiyappan and Jain, 2012). NOAA-HYDE data cover the period 1770–2010 with yearly temporal resolution and spatial resolution of 0.5∘×0.5∘, from which data for the period 1981–2010 were used for the development of this database, consistent with the hydroclimate data period described above.

The NOAA-HYDE land-cover maps show the percentage of grid cell area containing 28 different land-cover types (see Table A1 in Appendix A). In this study, we reclassified these 28 land-cover types into 10 distinct land-cover types, namely urban, shrubland, grassland, pastureland, cropland, forest, water, desert, tundra and savannah, by combining similar land-cover classes (see Table A1). As done for the hydroclimate data, the gridded land-use data were also spatially aggregated over each wetlandscape based on the area-weighted averaging method (with weights of specific land-cover area in each grid cell relative to total wetlandscape area). This provided a wetlandscape-specific data time series of annual land use and land cover for each of the reclassified 10 land-cover types. The final WetCID files comprised gridded time series data on absolute grid cell area (in km2) covered by each land-cover type, time series data in percentage of grid cell area covered by each land-cover type, and aggregated absolute and percentage time series data for each wetlandscape area.

Analogous to the hydroclimatic changes, period-specific change quantification can facilitate relatively simple and straightforward analysis of how different types of land-use changes between time periods correlate with and may have driven associated wetlandscape changes (Destouni et al., 2013; Jaramillo and Destouni, 2015). Equation (1) was therefore also used to calculate absolute change in the area of each land-cover type (km2) within each wetlandscape between Per1 (1981–1995) and Per2 (1996–2010). In the land-use case, VarPer1‾ and VarPer2‾ represent annual average area covered by a land-cover type within each wetlandscape during Per1 and Per2, respectively. Furthermore, the corresponding change in relative land-cover area (ChngRel in % of total wetlandscape area) was calculated as 2ChngRel=VarPer2‾-VarPer1‾AreaC×100, where AreaC is the total wetlandscape (catchment) area (in km2) and VarPer1‾ and VarPer2‾ are the annual average areas covered by each land-cover type in the wetlandscape during Per1 and Per2, respectively.

Table 1 summarizes some general geographical, climate and wetland type information provided by GWEN researchers in the survey information forms. Each site represents either an individual wetland or a wetlandscape (e.g., a catchment) including multiple wetlands. The country, main climate zone and wetland area relative to total wetlandscape (catchment) area are also given for each site in Table 1. Moreover, a summary of the availability–accessibility of local data on the wetlands, hydrology, climate and land uses as well as the wetlandscape (catchment) area in each of the 27 wetlandscapes is also shown in Fig. 2. The variables of evapotranspiration and soil moisture were revealed as having large data gaps (red color in Fig. 2), indicating an overall need to use other data sources (e.g., gridded global data products) for quantifying these variables and associated processes. Figure 2 also highlights the variability in data availability and open accessibility among the sites. For instance, no open-access data sources have been reported for the considered variables in the arid subtropical sites 13, 16 and 17, whereas open-access data sources have been reported for most variables in the cold Swedish sites 4 and 6 and the American subtropical sites 15 and 19.

The synthesized survey dataset also contains information about different types of wetland, hydroclimatic and/or land-use changes observed or perceived to have occurred in the 27 investigated wetlandscapes (Fig. 3). Substantial changes are reported for most of these wetlandscapes, but a few sites have no known changes (e.g., in the arid Moroccan site 17) or have important knowledge gaps regarding changes (e.g., in the cold Swedish sites 2 and 5, even though availability to at least some data is relatively good there). The information on local data availability–accessibility (Fig. 2) and observed or perceived change occurrence (Fig. 3) summarized and structured in WetCID can guide further study directions and support identification of key needs for complementary new local data and/or use of additional large-scale (regional-global) gridded data. Furthermore, the wetlandscapes of WetCID are located in different regions of the world, with seven sites in northern Europe (sites 1–7), seven in the Amazon and Caribbean region (sites 20 and 23–27), four in North America (sites 10, 15, 18 and 19), three in the Middle East (sites 12, 13 and 16), two in the Mediterranean region (sites 11 and 14), two in Siberia (sites 8 and 9), and two more in other parts of the world (northern Africa and eastern Asia). As such, regional patterns and characteristics can be identified and regional strategies developed, e.g., to enhance availability of data and information and determine further research needed to bridge region-specific knowledge gaps and decide on relevant management plans for each region's wetland ecosystems. Such regional characterizations and assessments can be initialized with the current version of WetCID and further updated as more data for already-included and possible additional regional wetlandscapes become available in future database versions.

Data for long-term average temperature and precipitation conditions, and changes in these between Per1 (1981–1995) and Per2 (1996–2010) at the 27 wetlandscape sites, are presented in Fig. 4. The horizontal axis in the diagrams shows the wetlandscape site numbers in order of their latitude from north to south, covering the latitude range from 70∘ N to 25∘ S. The increase in average temperature and precipitation with decreasing latitude (Fig. 4a–b) illustrates that the wetlandscapes also cover a wide range of hydroclimate conditions, from low to high temperature and precipitation values (see also Fig. 1). Temperature has increased over almost all wetlandscapes and considerably more so in the more northern and colder areas than in the warmer areas around and south of the Equator (Fig. 4a–b). In contrast, precipitation changes are relatively small, varying around zero, in the more northern, colder and drier areas, while precipitation has mostly increased in the warmer and also wetter areas around and south of the Equator (Fig. 4c–e). Overall, the changes in mean annual temperature range from zero to +1 ∘C, while the changes in precipitation range from -70 to +170 mm yr-1, with the Iranian site (12) (Lake Urmia catchment) exhibiting the greatest increase in temperature (+1 ∘C) and the greatest relative decrease in precipitation (-17 %).

Figures 5 and 6 exemplify gridded variability and change data for temperature and precipitation over the Volga (no. 9) and the León–Atrato (no. 23) wetlandscapes. The data time series of wetlandscape-aggregated annual average temperature and precipitation in these wetlandscapes (Fig. 5) exemplify such data prepared and included in WetCID for all 27 wetlandscapes. These two wetlandscapes were chosen for data exemplification because they represent different hydroclimatic conditions, with Volga being cold and dry while León–Atrato is warm and wet (Fig. 5), as well as having different sizes, with Volga being the largest (1 360 000 km2) and León–Atrato (2344 km2) one of the smallest studied wetlandscapes. The data for these examples (Fig. 5) are consistent with corresponding data implications across the different wetlandscapes over the world (Fig. 4) in indicating an overall positive (warmer–wetter) spatial correlation between long-term average temperature and precipitation. Temporally, however, the recent changes in these variables imply a negative correlation (towards warmer and mostly drier conditions) for the Volga wetlandscape (Fig. 6a and c) as for several other northern wetlandscapes in WetCID (Fig. 4). In contrast, a positive correlation (towards mostly warmer and wetter conditions) is implied by the recent temporal changes in the León–Atrato wetlandscape (Fig. 6b, d), as one of the most southern wetlandscapes in WetCID (Fig. 4). Such spatiotemporal sign shifts and dipole emergence in temperature–precipitation correlations have been noted in other recent studies of long-term variations and short-term changes of hydroclimate over Europe (Charpentier Ljungqvist et al., 2019). WetCID can facilitate further studies of these correlation conditions for and across the different wetlandscapes around the world.

The data for the Volga and León–Atrato examples also emphasize that wetlandscapes can have very different area extents (spatial scales), with potentially important implications for the spatial resolution (Fig. 6) and related usefulness of data provided in WetCID. For example, the Volga wetlandscape includes 982 grid cells with complete or partial coverage in the hydroclimate datasets, while the León–Atrato wetlandscape only includes four of such grid cells. Most of the available global datasets from climate and earth system models have coarser spatial resolution than the size of most individual wetlands. Thus, model data for individual wetlands are subject to high uncertainty, whereas data aggregated over whole wetlandscapes have greater potential for accuracy (Bring et al., 2015), highlighting the need for considering the whole-wetlandscape scales in assessments of how wetland systems interact with hydroclimate and land-use changes.

The aggregated and gridded land-use data in WetCID can also be used for different types of whole-wetlandscape analyses. Figure 7 summarizes the data for the long-term average relative area of each land-cover type (% of total wetlandscape area), and associated absolute area changes (km2) and changes in relative area coverage (% of total wetlandscape area), for different land-cover types across the 27 wetlandscapes. The data reveal, for example, the high percentage of forest area in wetlandscapes at high latitudes and in the tropics, while relative cropland area increases towards the temperate regions (Fig. 7a). Figure 7 also summarizes the different types of land-cover transformations, for example from “forest” into “cropland and pastureland” in the tropical Mekong wetlandscape (21), “pastureland” into “grassland” in the temperate Irish wetlandscape (7) and into “cropland” in the borderline cold–dry Iranian wetlandscape of the dramatically shrinking Lake Urmia (12) (Khazaei et al., 2019), “shrubland” into “cropland” in the borderline temperate Iranian wetlandscape (13), and “cropland” into “shrubland” in the warm temperate Greek wetlandscape (14).

The data time series of different land-cover types and their changes between Per1 (1981–1995) and Per2 (1996–2010) show, for example, forest and (decreasing) cropland, followed by pastureland and grassland, to be dominant in the large Volga wetlandscape, while forest, pastureland and (decreasing) cropland areas dominate the small León–Atrato wetlandscape (Fig. 8). Gridded maps of land-cover area changes in these wetlandscape examples (Figs. 9–10) again demonstrate large spatial-resolution differences, with potentially important implications for the usefulness of land-use datasets for wetlandscapes of smaller scales. For example, in the most northern Swedish–Arctic wetlandscape (1), grassland is obtained as the second most dominant land-cover type after tundra (Fig. 7a), which is not normally seen in this northern Arctic region.