The UK Environmental Change Network datasets – integrated and co-located data for long-term environmental research (1993–2015)

.

Abstract. Long-term datasets of integrated environmental variables, co-located together, are relatively rare. The UK Environmental Change Network (ECN) was launched in 1992 and provides the UK with its only long-term integrated environmental monitoring and research network for the assessment of the causes and consequences of environmental change. Measurements, covering a wide range of physical, chemical, and biological "driver" and "response" variables are made in close proximity at ECN terrestrial sites using protocols incorporating standard quality control procedures. This paper describes the datasets (there are 19 published ECN datasets) for these co-located measurements, containing over 20 years of data . The data and supporting documentation are freely available from the NERC Environmental Information Data Centre under the terms of the Open Government Licence using the following DOIs.
The monitoring programme includes a wide range of physical, chemical and biological "driver" and "response" variables, identified by experts in the field as being important for the assessment of environmental change (see Table 2). A Statistical and Technical Advisory Group met regularly to review ECN monitoring activities. These measurements are made in close proximity at each site, using standard protocols incorporating standard quality control procedures . Data are managed by the ECN Data Centre, which has an integrated information system (Rennie, 2016) that stores all data and metadata collected by the networks which supply data to it. These data are held in standardised structures in order to support the cross-disciplinary analyses necessary for environmental change research. An associated summary database consists of monthly, quarterly and/or annual summaries of these data using summary statistics appropriate to each measurement, as advised by experts. These summary data can be explored through data visualisation interfaces available on the website (ECN Data Centre, 2019). The database uses the Oracle relational database management system with links to Arc GIS for spatial data handling. Data were regularly sent in from sites and were quality-assured be-fore being lodged in the database (information about quality control is in Sect. 4).
This paper describes the datasets for the high-frequency, co-located ECN measurements. There are 19 published datasets (Table 2), containing over 20 years of data , covering biological, meteorological and biogeochemical measurements (Rennie et al., 2016a(Rennie et al., , b, 2017a(Rennie et al., -p, 2018. They are hosted by the NERC Environmental Information Data Centre and are available to users under the Open Government Licence.

Methods
ECN measurements are co-ordinated and standardised across sites according to published protocol procedures . The protocol documents are included in the supporting documentation provided alongside every data down-  load. The protocols are designed to ensure consistency in methods and data handling over time and across the ECN's sites. Sites were visited on the same day each week, preferably on a Wednesday, to synchronise sampling, within the site and across the network. The protocol documents detail quality control procedures, e.g. correct handling of equipment and samples, maintenance schedules, and calibration specifications, as well as unambiguous instructions for measurement and data handling. Data requirements are an integral part of these protocols and include specifications of variables, units, reporting precisions, dimensions, resolutions, reference systems and quality assurance procedures. These specifications, together with as much information as possible about likely user requirements, were used in the design of the database and the construction of standard formats for data transfer and standard field forms for each dataset. Where available, existing data capture methodologies were used (e.g. the Rothamsted light trap network, part of the Rothamsted Insect Survey, 2019) to maintain compatibility with other sectoral networks.
At each site, an area of 1 ha (10 000 m 2 ) was selected and permanently marked. This is called the target sampling site (TSS), and destructive sampling within it was kept to a minimum. Many of the measurements are co-located within the TSS. Dispersed monitoring protocols (e.g. vegetation) also include plots within the TSS. The TSS was chosen to be rep-resentative of the predominant vegetation, soil and management of the site.
Some protocols (Sect. 2.15 to 2.19) have not been measured at all sites or have had varied uptake at sites over time, limiting their use for cross-site comparison. In addition, some protocols are designed as national-scale surveys, thus they have limited use for assessment of trends at individual sites. These limitations are discussed with each individual dataset. The methods for data collection for the 19 published ECN datasets  are summarised below.

Meteorology
Automatic weather stations (AWSs) were installed at all ECN terrestrial sites and situated in accordance with British Meteorological Office site requirements (Meteorological Office, 1982). The AWS was ideally located on, or within 500 m of, the TSS. The layout of the meteorological enclosure is provided in Fig. 2. Full details for the procedure for installing an AWS are provided in the protocol document (Burt and Johnson, 1996), but the instruments were fixed to two crossarms -one at 2 m above ground level and oriented east-west and the other a 1 m a.g.l. and oriented north-south. The wind vane and anemometer were located on the upper cross-arm and the air temperature and radiation sensors on the lower cross-arm. A number of the sites also had either a manual meteorological station (referred to as MM in Fig. 2) or a second AWS to quality check the data. In addition, the majority of sites have operated more than one AWS in the same location, e.g. when kit is replaced (see Sect. 3.1 for details on how this is recorded in the dataset). All ECN AWS instruments were subject to regular (normally annual or biannual) professional calibration checks by external contractors. The data are hourly summaries calculated from 5 s samplings and the variables recorded are listed in Table 3. Full operating procedures are provided in the protocol document (Burt and Johnson, 1996), which is included in the supporting documentation provided alongside the data download (called MA.pdf).

Atmospheric nitrogen
Passive diffusion tubes were used to measure the concentration of nitrogen dioxide (NO 2 ) at all ECN terrestrial sites. They were attached to a post at a height of 1.5 m a.g.l. in the meteorological enclosure (Fig. 2). As a control measure, blank tubes were also transported to the site but were not exposed on arrival. The blank tubes were returned to the laboratory the same day, stored in a refrigerator and analysed in the lab alongside the experimental tubes. In the early years of the ECN, the diffusion tubes were assembled and analysed locally, but these were replaced at some sites by commercially made tubes manufactured and analysed by Gradko Ltd. Comparability tests were conducted when this switch was made. The samples were collected fortnightly and the variables recorded are listed in Table 4. Full operating proce-   dures are provided in the protocol document (Bojanic, 1996), which is included in the supporting documentation provided alongside the data download (called AN.pdf).

Precipitation chemistry
Bulk (open funnel) precipitation collectors were used to measure the precipitation chemistry at all ECN terrestrial sites. These were situated in the meteorological enclosure ( Fig. 2), in an open location away from local sources of contamination (e.g. vehicle tracks or animal houses). Warren Spring Laboratory standard precipitation collectors were used, with the collecting bottle fixed 1.75 m a.g.l. The collectors were secured by guy ropes or bolted to a concrete base. The collector had a filter to prevent debris falling into the bottle and was kept dark and cool by a jacket. The collecting bottle was changed at the same time each week, and the funnel was replaced or cleaned with deionised water. The volume collected was recorded, and analysis of the samples were made by the analytical laboratories linked to each site. The cost of standardising methods of analysis across all ECN laboratories was prohibitive. Instead, the analytical guidelines (available in supporting documentation available with the data download) list approved techniques for each determinand with their corresponding limits of detection. The sponsoring organisations were responsible for maintaining their own continuity in methods for existing long-term runs of data. Each laboratory practised its own internal quality control, and most participated in national quality assurance schemes. As a quality check, a standard quality control solution was sent to the laboratories that analyse the ECN water samples. This solution was analysed alongside the samples collected in the field. The samples were collected weekly, and the variables recorded are listed in Table 5. Full operating procedures are provided in the protocol document , which is included in the supporting documentation provided alongside the data download (called PC.pdf).
Operating procedures for handling water samples (Adamson, 1996a) and analytical guidelines (Rowland, 1996) are also provided in the supporting information (called WH.pdf and WAG.pdf).

Soil solution chemistry
Water was collected from soils via suction lysimeters at the majority of ECN terrestrial sites. The lysimeters were installed at two depths within a 10 m by 10 m plot on the edge of (but outside) the TSS. Six samplers were installed in the A horizon and six others at the base of the B horizon (or at 10 and 50 cm if these soil horizons did not exist), ideally on a downslope to avoid debris from soil disturbance. Samplers were emptied and the water volumes collected on the same day each fortnight. A week after sample collection, the samplers were evacuated to 0.5 bar (or 0.7 bar for sites where insufficient soil solution could be collected), thus the water only accumulated over the second week of the fortnightly period. The chemistry of the water collected was analysed by the analytical labs associated with each site. At some sites, particularly in drier months, the volume of water collected may have been very small; in these cases, the samples were discarded or, if possible, combined (only samples from the same horizon were combined) for analysis (see Sect. 3.2 for details on how this is recorded in the dataset). The samples were collected fortnightly and the variables recorded are listed in Table 5. Full operating procedures are provided in the protocol document (Adamson, 1996b), which is included in the supporting documentation provided alongside the data download (called SS.pdf). Operating procedures for handling water samples (Adamson, 1996a) and analytical guidelines (Rowland, 1996) are also provided in the supporting information (called WH.pdf and WAG.pdf).

Surface water chemistry
Dip samples from rivers and streams were collected. This was only done at sites where flowing water was present. Samples were taken at a representative location above a weir; some sites collect samples at multiple locations on the site (indicated by the location code in the dataset). The collecting bottle is rinsed in river water, and a 250 mL sample of river water is taken. The samples were collected weekly and the variables recorded are listed in Table 5. Full operating procedures are provided in the protocol document (Johnson and Burt, 1996a), which is included in the supporting documentation provided alongside the data download (called WC.pdf).
Operating procedures for handling water samples (Adamson, 1996a) and analytical guidelines (Rowland, 1996) are also provided in the supporting information (called WH.pdf and WAG.pdf)

Surface water discharge
Hydrological data from rivers and streams were collected by a logger at sites with a river or stream. Recording of river stage was done by a permanently installed weir, the design of which was determined by the conditions at the site. Data were recorded by a logger. The data are 15 min averages  Table 2. Full operating procedures are provided in the protocol document (Johnson and Burt, 1996b), which is included in the supporting documentation provided alongside the data download (called WD.pdf).

Moths
Light traps were used to sample moths (Macrolepidoptera) at the majority of the ECN terrestrial sites using the Rothamsted Insect Survey method (Rothamsted Insect Survey, 2019) at the majority of the ECN terrestrial sites. Where possible, the light trap was sheltered by vegetation and placed away from artificial light sources, in a location that was convenient for daily emptying. The traps require a continuous power supply so this often determined their location. Ideally, the traps were emptied daily throughout the year, but when this was not possible (e.g. for more remote sites or at the weekend) samples could accumulate. Samples from the sites were identified by a single expert contracted by the ECN. The data are stored within the Rothamsted Insect Survey database, as well as in the ECN database. A count of each species trapped was recorded. Full operating procedures are provided in the protocol document (Woiwod, 1996a), which is included in the supporting documentation provided alongside the data download (called IM.pdf).

Butterflies
Butterfly species were recorded on a fixed transect (which was divided into a maximum of 15 sections) at the majority of the ECN terrestrial sites. The transect was chosen to be broadly representative of the site and include areas under different management regimes. The length of the transect was dependant on the local conditions at the site. The national Butterfly Monitoring Scheme methodology was used (UK Butterfly Monitoring Scheme, 2019). The transect was walked at an even pace and the number of butterflies that were seen flying within or passing through an imaginary box (5 m wide, 5 m high and 5 m in front of the observer) were recorded. Sampling took place when the temperature was between 13 and 17 • C if sunshine was at least 60 %. However, if the temperature was above 17 • C (15 • C at more northerly sites), recording could be carried out in any conditions, providing it was not raining. Transects were walked weekly between 1 April and 29 September, providing the meteorological conditions were met. A count of each species observed was recorded. Full operating procedures are provided in the protocol document (Woiwod, 1996b), which is included in the supporting documentation provided alongside the data download (called IB.pdf).

Carabid beetles
Pitfall traps were used to collect carabid beetles (Carabidae) at the majority of the ECN terrestrial sites. A total of 30 traps were set, divided between three transects, in or adjacent to the TSS and in areas representing different habitats where possible. The traps were polypropylene, with a 7.5 cm diameter and 10 cm depth, and were filled with ethylene glycol preservative. They were buried with the top of the trap flush with the soil surface. The traps were set 10 m apart along the transect. A wire netting cage made from chicken wire was attached to the rim of the trap to reduce the number of small mammals inadvertently caught. Each trap also had a cover to help prevent rain flooding the traps and to reduce bird interference. Samples were analysed by a local taxonomic expert. The samples were collected fortnightly (between May and the end of October). A count of each species trapped was recorded. Full operating procedures are provided in the protocol document (Woiwod and Coulston, 1996), which is included in the supporting documentation provided alongside the data download (called IG.pdf).

Spittle bugs
Populations of Philaenus spumarius and Neophilaenus lineatus were monitored annually at the majority of the ECN terrestrial sites. In mid-June, counts of the spittle produced by nymphs were made in 20 quadrats (0.25 m 2 ) randomly placed near the TSS. Also, in late August, the proportions of each colour morph of the adult P. spumarius were estimated using sweep netting on the TSS when the weather conditions were dry. Colour polymorphism is likely to be environmentally determined (Whittaker, 1965) and therefore an indicator of environmental change. The samples were collected annually (nymphs in June and adults in August). A count of each species and colour morph was recorded. Full operating procedures are provided in the protocol document (Whittaker, 1996), which is included in the supporting documentation provided alongside the data download (called IS.pdf).

Baseline vegetation
This was a one-off survey at the start of ECN monitoring to establish a vegetation map at all sites. It allowed a vegetation map to be generated and the plots for continuous monitoring (see Sect. 2.12, 2.13, 2.14) to be selected. An approximately regular grid, coincident with the UK National Grid, was superimposed on the site map, scaled so as to provide approximately 400 sample grid positions. This ensured the plot locations were unbiased and relocatable. Additionally, no more than 100 points (infill points) were chosen to ensure all vegetation types were represented. A 2 m × 2 m plot was centred on each grid and infill point, oriented using magnetic bearings. These plots were permanently marked (the plot corners are marked with buried metal stakes). Species presence was recorded in the plots. Where the plots fell in woodland, the trees and shrubs were recorded in a 10 m × 10 m plot centred on the 2 m × 2 m plot to provide a more representative sample of the canopy and understory. Full operating procedures are provided in the protocol document (Rodwell et al., 1996), which is included in the supporting documentation provided alongside the data download (called V.pdf).

Coarse-grain vegetation
A random selection was made of 40 of the 2 m × 2 m plots from the regular grid set-up for baseline survey vegetation recording (Sect. 2.11) at the majority of the ECN terrestrial sites at the onset of ECN monitoring. Where infill plots were included in the baseline survey, up to 10 of these plots were also randomly selected, providing a total of up to 50 of these plots for coarse-grain monitoring. The plots were permanently marked. Where plots fell in woodland or scrub, the associated woodland protocol was also undertaken (see Sect. 2.13). The protocol was undertaken every 9 years. Species presence was recorded in each of the twenty-five 40 cm × 40 cm cells within the plots. Full operating procedures are provided in the protocol document (Rodwell et al., 1996), which is included in the supporting documentation provided alongside the data download (called V.pdf).

Woodland vegetation
Where grid and infill plots selected for coarse-grain sampling (Sect. 2.12) fall in scrub or woodland, 10 m×10 m plots (which were centred on the 2 m×2 m plot used in the coarsegrain survey) were used to record trees and shrubs. Species dominance was assessed within the plots. A total of 10 cells, each 40 cm × 40 cm, were selected at random within the plot and marked. Seedlings were counted by species in each cell. Additionally, an individual tree was chosen nearest the centre point of the cell and monitored for height and diameter at breast height (dbh). The protocol was undertaken every 9 years, but dbh was measured every 3 years for sites where there was woodland. The variables recorded are listed in Table 6. Full operating procedures are provided in the protocol document (Rodwell et al., 1996), which is included in the supporting documentation provided alongside the data download (called V.pdf).

Fine-grain vegetation
At least two 10 m × 10 m plots from each vegetation type present at the site were randomly selected (from the plots selected in the baseline survey (see Sect. 2.11). The plots were chosen to coincide with the original grid and infill plots where possible but otherwise were selected using randomly selected pairs of co-ordinates. The plots did not coincide with the coarse-grain sampling plots (see Sect. 2.12) to avoid repeated disturbance to the plots. Ten 40 cm × 40 cm cells were selected randomly within these plots. This survey was under- taken every 3 years, but some sites chose to do this survey annually to provide a better temporal range. The same plots were visited on each occasion, but often a smaller number of plots were chosen to do the annual survey. Species presence was recorded within the cells. Full operating procedures are provided in the protocol document (Rodwell et al., 1996), which is included in the supporting documentation provided alongside the data download (called V.pdf).

Frogs
It is difficult to monitor populations of adult frogs; therefore, phenological observations were made of selected pools and ditches, and the number of egg masses were assessed as an indicator of the "health" of frog populations at sites with standing water present. Additionally, a 250 mL water sample was taken from the spawning area and analysed. The time at which frog breeding starts in the UK varies greatly; therefore, observations of frog behaviour were made at the appropriate time for each site. The variables recorded are listed in Table 7. Full operating procedures are provided in the protocol document (Beattie et al., 1996), which is included in the supporting documentation provided alongside the data download (called BF.pdf).

Birds -breeding bird survey
Bird species were recorded on two transect lines (within a 1 km square) at the majority of the ECN sites. Counts were made in the morning, ideally no later than 09:00 UTC. Transects were walked, at a slow and methodical pace, when the visibility was good and there was no strong wind or heavy rain. All birds that were seen or heard, as well as their distance (there are four distance categories) from the transect were recorded. The methodology used was that of the Breeding Birds Survey (BBS, 2019) organised by the British Trust for Ornithology (BTO). The transect was walked twice each year (once between April and mid-May and the second between mid-May to late June). Full operating procedures are provided in the protocol document (Sykes, 1996a), which is included in the supporting documentation provided alongside the data download (called BB.pdf). This protocol replaced the Common Bird Census (see Sect. 2.17) in 1999. The methodologies of the two surveys are different, thus it is unfortunately not possible to create a single time series from both datasets. Please also note that the Breeding Birds Survey is designed to be a national-scale survey, therefore the site-based ECN data are limited in the amount of information that they can provide on the precise relationships between population levels and environmental change. It is recommended that the ECN data are used in conjunction with data from more widespread monitoring programmes (i.e. those of the BTO) so these limitations can be mitigated.

Birds -common bird census
Bird species were recorded in a plot that was, ideally, a minimum of 40 ha in farmland and 10 ha in woodland. The methodology used was that of the Common Birds Census (CBC, 2019) organised by the BTO. A total of 10 visits were made between mid-March and late June, spaced evenly through the season. Cold, windy and wet days were avoided. The CBC uses a mapping method in which a series of visits were made to all parts of a defined plot during the breeding season and contacts with birds by sight or sound were recorded on large-scale maps. Information from the series of visits was combined to estimate the number of territories found. Within the CBC protocol, some species were also monitored by nest counts on the plot or by a combination of nest counts and territory estimation. Full operating procedures are provided in the protocol document (Sykes, 1996b), which is included in the supporting documentation provided alongside the data download (called BC.pdf). The CBC was the standard protocol at lowland ECN sites until 1999 when it was replaced by the BBS (see Sect. 2.16). The methodologies of the two surveys are different so it is unfortunately not possible to create a single time series from both datasets. A few sites continued the CBC alongside the BBS for a few years to allow for a comparison. Additionally, historical data (pre-ECN) was obtained for the Wytham site. Therefore, the date ranges for individual sites in this dataset are not consistent. As with the BBS, the CBC was designed to be a national-scale survey, thus similar limitations apply to the site-based ECN data provided in this dataset.

Bats
Bat species were mapped (using a bat detector) and their behaviour recorded at the majority of the ECN sites. One or more kilometre-sized squares were selected at the site. This selection did not need to be random as long as the square was reasonably typical of the site and that fieldwork could be conducted safely at night. The square was divided into two and a transect selected through each of these half squares. The methodology was based on that used in the Bats and Habitats survey organised for the Joint Nature Conservation Committee (Walsh et al., 1995). The transect was walked four times in each year (once in each 3-week period between June and September). Bat detectors were used during the sur-vey and the frequency of the detector was tuned to could be altered during the survey if that helped ensure all species were recorded (in particular to distinguish between Pipistrellus species). Surveys were not carried out when rain was heavy or there were strong winds. A count of each species observed and their behaviour was recorded. Full operating procedures are provided in the protocol document (Walsh et al., 1996), which is included in the supporting documentation provided alongside the data download (called BA.pdf).
The methodology is somewhat limited in the amount of information that it can provide about the precise relationships between population levels and environmental change. Nevertheless, by linking the ECN results to those from more widespread monitoring programmes, these limitations can be mitigated.

Rabbits and deer
There were no practicable methods of making direct measures of the population size of the rabbit and deer populations; therefore, an index method based on dropping counts was used to estimate relative abundance at the majority of the ECN sites. The butterfly monitoring transect was used. A second transect that covered habitat types not present on the butterfly transect was also selected. Dropping counts were recorded on a transect twice a year (once in late March and again in late September). Droppings on the transect were cleared 2 weeks before sampling took place. At Moor House, the same methodology was also used to estimate the relative abundance of grouse. Full operating procedures are provided in the protocol document , which is included in the supporting documentation provided alongside the data download (called BU.pdf).

Datasets
The ECN datasets are listed in Table 2, together with their citation information, the frequency of measurement and the variables collected. The NERC Environmental Information Data Centre (the repository that hosts the datasets) provides data and supporting information as separate packages -this allows improvements to be made to the supporting documentation over time if necessary while maintaining a persistent, citable dataset. The DOI for each dataset links to a landing page that contains separate links to download the data and the supporting information.
Each dataset follows the same basic structure: -SITECODE -site code (see Table 1); -SDATE -date of sampling; -FIELDNAME -the variable being measured (these are described below and in the supporting information); -VALUE -the value of the measured variable.
All the datasets have this structure in common but some of the datasets may also contain some additional information where necessary for the measurement. This is fully documented in the supporting information. For the majority of datasets, the entire time period is included in the data download; however, two large datasets are split into yearly time slices to make downloading easier for the user (see Sect. 3.1 and 3. 3) The supporting information, i.e. the protocol document, supplementary data and quality information, is provided with each dataset. It is important to refer to this information prior to analysing the data. The supporting information is provided in a zip file using the "supporting information" link on the relevant page for each dataset (Rennie et al., 2016a(Rennie et al., , b, 2017a(Rennie et al., -p, 2018. All the zip files contain a document called ***_DATA_STRUCTURE.doc (where *** is the ECN measurement code; see Table 2). This document contains detailed information about the structure of the dataset, location information for the sites, information about the variables measured, and documents for any additional information needed to understand the dataset and provides any coding lists used.
Some usage notes are included below.

Meteorology
Given the size of this dataset, the data have been split into yearly csv files. Users are advised to open these files in a text editor or to use a statistical package to analyse these data as the file sizes remain too large for a software package like Excel to open. Over the period of data collection, the majority of the ECN sites have operated more than one AWS in the same location -e.g. when kit is replaced. In many cases, these have been run concurrently to enable cross-checking of data. Replacement AWSs are indicated by the "AWSNO" field in the dataset -these are ID numbers assigned sequentially. Users should be aware of the AWSNO when analysing the dataparticularly when two AWSs have been run concurrentlyto avoid misleading results by inadvertently combining data from two AWSs.

Soil solution chemistry
Where samples were combined, this is indicated in the data with the replicate IDs XXS (combined shallow samplers) and XXD (combined deep samplers) in the datasets. Occasionally, the suction samplers were replaced, this is indicated in the data with a new replicate ID.

Surface water discharge
Given the size of this dataset, the data have been split into yearly csv files. Users are advised to open these files in a text editor or use a statistical package to analyse these data as the file sizes remain too large for a software package like Excel to open.
One site (Moor House -Upper Teesdale) uses an Environment Agency logger to record water discharge. The Environment Agency uses the WISKI format to record these data (the Hydrolog format was used prior to 2004). Both of these formats include quality information that is available in this dataset (for Moor House only). An explanation for these quality codes is provided in the supporting information.

Carabid beetles
There is an additional data column in this dataset that applies to only one species (Pterostichus madidus), where additional information was collected on gender (M or F) and leg colour (red, R, and black, B). The ratio of leg colour is thought to depend on ecological factors (Terrell-Nield, 1992).

Standards and coding lists
The ECN forms part of a global system of long-term, integrated environmental research networks; see Sect. 5 for more details. Therefore, it primarily uses the LTER-Europe controlled vocabulary, EnvThes (EnvThes, 2019), as the basis for the semantic harmonisation of data with its European and International partners. The ECN uses a number of coding lists within its datasets. Where possible, existing coding systems were used to maintain compatibility with other related data resources. The coding lists used by the ECN are listed in Table 8. These coding lists are fully documented in the supporting information.

Dataset completeness
The majority of the ECN sites have been collecting the full suite of ECN measurements since 1993 but two sites joined the network later -Yr Wyddfa (Snowdon) in 1995 and Cairngorms in 1999. However, it should be noted that many of the sites are in remote locations, which means that site managers are occasionally unable to attend the sites for health and safety reasons, causing gaps in the dataset. In particular, there was a foot-and-mouth disease outbreak in the UK in 2001, which meant a number of the sites could not be visited for biosecurity reasons and that the data for that year are patchy. In addition, Rothamsted ceased biological monitoring in 2011 and Drayton left the network in 2014.

Data quality
Quality control is central to all stages of ECN data collection and management and is handled through a number of steps.

Standard operating procedures
As described in the Sect. 2, data collection procedures were co-ordinated and standardised across the sites through published protocols.

Data transfer templates
Data were checked and formatted by data providers prior to being submitted by email (in standardised, comma-separated files). Detailed data transfer documentation for each protocol guided the preparation of these files to ensure comparability of data across sites and over time. This documentation includes rules for handling missing values and data quality information. To aid site managers, a bespoke set of data entry templates were developed for each protocol, using MS Access, to improve data handling efficiency (Rennie, 2016). These templates incorporate quality-checking procedures and help to ensure that quality-checked, standardised and formatted data were submitted by site managers. The design of the templates takes into account ease of use, with the main emphasis being on minimising error. This type of data entry software is particularly useful where numeric coding systems for species are in use; numbers are less memorable and mistakes in one digit of a code can produce serious errors. For example, the software uses drop-down lists of codes (which are dynamically linked with a list of the species names) so that the codes can be cross-checked against the species name to ensure that the correct code is chosen.

Data verification
In addition to the checks made in the templates, standard verification procedures were applied to all data before import into the database. The procedures performed numeric range checks (i.e. checking if a value falls within a specified range), categorical checks (e.g. checking that a species code appears on the standard code list), formatting (i.e. that the dataset conforms to the specified data format) and logical integrity checks (i.e. checking the data make sense, e.g. that the dates in one dataset match those in a related dataset). Appropriate range settings for ECN variables were selected following discussion with specialists in each field. These ranges are held in a table in the database and the data are checked against this before being committed to the database. Where values fell outside these ranges, a cautious approach was adopted towards discarding data on the principle that apparent errors could be valid outliers. Data values identified by validation software as "out of range" were treated in one of three ways.
-Where values were clearly meaningless due to a known cause (e.g. an instrumentation fault that could not be back-corrected), the data were discarded and database fields set to null (no data), and quality flags were added to the database. -Where values were clearly in error, or out of range due to known calibration errors and could be backcorrected, the data were corrected (these changes were flagged in the database).
-Where there was no straightforward explanation for outliers, the data were stored in the database, accompanied by quality flags (see Sect. 4.4).

Quality flagging
The ECN site managers assigned quality codes to indicate factors that may affect the quality of the data being collected, including deviations from the protocol, faulty instrumentation and common problems. They picked these from a standard list of ECN quality codes; these quality codes are included in the data download, and an explanation for the codes is provided in the supporting documentation. Site managers could pick as many quality codes as were applicable. Occasionally, an unusual event took place that was not covered by these codes. In that case, the site manager attached text explaining the circumstances. This is indicated by a quality code "999" in the data download. This quality text is available in a file called ECN_***_qtext.csv (where *** is the measurement code; see Table 2), which is provided in the supporting documentation.

Quality assessment exercises
Samples were kept where possible (e.g. archived invertebrate samples), meaning the accuracy of identification can be assessed at a later date if necessary. Occasionally, quality assessment exercises have been run by appropriate experts to check, for example, consistency in species identification across sites (Scott and Hallam, 2003). The quality of more ephemeral measurements such as meteorology or water quality can only be similarly assessed by running duplicate or parallel systems. Duplicate systems are expensive, and in practice assessment normally involved regular checks for instrument drift and recorder error. Where possible, when new instrumentation or methods needed to be introduced, new and old systems were run in parallel to assess their relationship. This is assessed by the individual site manager, who must satisfy themselves that the new systems compare well before proceeding with the switchover.

ECN datasets in context
The ECN is nationally unique with its focus on highfrequency and co-located measurements. It provides a rare opportunity to link pressures and responses to investigate relationships between environmental variables and explore environmental change over significant timescales. The data included within these datasets have been the focus of a number of peer-reviewed scientific publications over the past 20 years. For example, linking meteorological data with invertebrate species data for exploring the impact of drought (Morecroft et al., 2002), exploring trends in the physical and biological environment (Morecroft et al., 2009), determining that hydrochloric acid deposition was a driver of UK soil acidification (Evans et al., 2011), and investigating declines in carabid beetle biodiversity (Brooks et al., 2012). Many of the datasets were incorporated in papers forming a journal special issue marking the first 20 years of the ECN (Sier and Monteith, 2016b). This special issue demonstrates how effective the datasets are in assessing and interpreting environmental change, covering a breadth of topics, such as trends in weather and atmospheric deposition ; trends in dissolved organic carbon (Sawicka et al., 2016;Moody et al., 2016); various aspects of change in UK plant communities Morecroft et al., 2016;Pallett et al., 2016;Milligan et al., 2016), ecosystem services , and carabid beetle communities (Eyre et al., 2016;Pozsgai et al., 2016); the use of digital imaging to assess vegetation cover (Baxendale et al., 2016); and the response of Lepidoptera communities to warming (Martay et al., 2016). A full catalogue of the peer-reviewed papers that have used ECN data are available on the website (ECN Publications Catalogue, 2019).
ECN sites cover a wide range of UK habitats but, given their focus on high-frequency data, are costly to run and are relatively few in number. The representativeness of ECN sites was compared to data obtained by the UK Countryside Survey (CS -Countryside Survey, 2019). The survey is based on a stratified random sample of 1 km squares from the intersections of a regular 15 km grid superimposed on the rural areas of Great Britain. Analysis revealed that the British ECN sites effectively span the range of values for both temperature and rainfall and cover a similar range of vegetation types to the CS, with the exception of arable, a land use category not assessed at ECN sites but present on several sites (Dick et al., 2011).
ECN sites contribute to a number of national monitoring programmes, e.g. Rothamsted Insect Survey (Rothamsted Insect Survey, 2019), Countryside Survey (Countryside Survey, 2019), the UK Butterfly Monitoring Scheme (UKBMS, 2019), the Breeding Bird Survey (BBS, 2019), the United Kingdom Eutrophying and Acidifying Network (UKEAP, 2019), and the Cosmic-ray Soil Moisture Monitoring Network (COSMOS-UK, 2019). The ECN's focus on multidisciplinary, co-located measurements can help integrate these networks and provides temporal-scale context for observations made by these networks, for example by providing information on year to year variation in vegetation communities to help inform how CS data can be influenced by weather variability (Scott et al., 2010).
The ECN is formally recognised as the UK's contribution to a global system of long-term, integrated environmental research networks and is a member of LTER-Europe (the European Long-Term Ecosystem Research Network - Mirtl, 2010) and ILTER (International Long-Term Ecological Research - Kim, 2006

Data availability
Provision of easy access to data has always been central to the ECN's strategy to provide a resource for environmental research, policy purposes and public information. The ECN datasets are hosted by the NERC Environmental Information Data Centre (EIDC, 2019) managed by the UK Centre for Ecology and Hydrology (UKCEH). The EIDC manages nationally important terrestrial and freshwater science datasets and is a CoreTrustSeal accredited data repository. EIDC has a registration system -users need a free account to download data. The ECN datasets can be discovered and downloaded through the EIDC's data catalogue (the Environmental Information Platform, EIP). The datasets are listed in Table 2, together with their citation information. They should be cited for every use using the information provided (Rennie et al., 2016a(Rennie et al., , b, 2017a(Rennie et al., -p, 2018. The ECN datasets are available under the Open Government Licence (Open Government Licence, 2019), and they are available as comma-separated files. Temporal extensions, provided as additional time slices, to the datasets will be created as further data become available.

Conclusions
The datasets collected by the UK Environmental Change Network are an invaluable and nationally unique resource, which, over the years, has proved useful to a range of users, including the scientific community and national policy makers. The co-location of high-frequency meteorological, biological and biogeochemical measurements means the ECN datasets are ideally placed for the development of clearer process understanding and assessing the impact of shorter-term events, such as droughts, on ecosystems. This 2-decade ECN data record provides a long-term baseline of environmental variability across a wide range of UK habitats against which environmental changes can be assessed.
Author contributions. SR was responsible for the management of the ECN Data Centre, publication of the datasets and led the writing of this paper. CA, SA, DB, SB, VB, JD, BD, CM, DP, RR, SMS, TS, CT and HW are the current site managers and are responsible for site management, data collection and quality checking. All co-authors contributed to the writing, discussion and review of this paper.
Competing interests. The authors declare that they have no conflict of interest.