Mapping data provide information about the distribution of a habitat or species. They are one of the most basic information used in biodiversity conservation. Besides determination of the range of a habitat or species, mapping data are frequently used to assess status and trends of habitats and species in applied conservation. Mapping data are usually based on observations made by numerous volunteers at different years and locations. While some locations may be surveyed repeatedly by the same person(s) across years, in general mapping data rarely are based on repeated surveys and thus, strictly speaking, are not monitoring data. Nevertheless, they are often used to assess status and trends of biodiversity. Here we will briefly outline potentials and limits of such assessments and provide relevant links to statistical methods that can be used for estimating trends from opportunistically collected data.
Distribution mapping versus monitoring. Distribution maps usually result from surveys carried out by different observers at different locations and times across a geographically wide area. Observers and locations usually change through time; only exceptionally, the surveys are systematically repeated in time. Biodiversity monitoring, in contrast, is the systematically repeated observation or measurement of biodiversity at the same location(s) through time with standardized methods.
Analysing status and trends from distribution mapping data - Potentials and limits. Distribution data provide information about the presence of a habitat or species at particular locations at a particular time. Survey efforts are usually insufficient to be certain about absences and absences are rarely documented. As a consequence, it is rarely known whether an unoccupied locations means true absence or presence but no or insufficient surveys for the habitat or species.
The number of locations, or the number of grid cells occupied in gird based distribution maps, is often used as an index of (relative) abundance, which then is used as a basis of inferences about status and trends. Ideally, an index of abundance correlates linearly with absolute density or abundance but non-linear, monotonously increasing indices are sometimes sufficient, especially if orders of magnitude suffice as precision. Populations generally differ in size and the density of populations within an area depends on habitat quality and other factors. Likewise, the extent of a habitat will differ within its distribution area in relation to abiotic factors. Thus, a linear relationship of grid cell occupancy of habitats or species with abundance is unlikely. Nevertheless, a non-linear relationship may still hold in most cases.
In addition to the relationship with abundance, detection probability either must be known or remain constant for an index to be suitable for assessments of status and trends. Unfortunately, this is rarely the case with mapping data. Rather, detection probability usually will change with time, location, and the interaction of these two factors. While this prevents the use of mapping data for inferences about absolute status or trends, it may still allow assessment of status and trends relative to other habitats or species included in the same distribution surveys. Strictly speaking, this requires that there is no shift in effort to record different species or habitats, e.g., all species and habitats of a fixed list are surveyed and recorded. If additionally mapping surveys cover many sites, inferences about the relative abundance of species or habitats should be fairly safe. Notwithstanding, one should be aware that this may not hold for very secretive or unpopular species or habitats that are compared to highly conspicuous or popular species or habitats or when particular species or habitats were targeted at particular years of the survey!
Inferences about temporal trends are more difficult because survey efforts are rarely constant through times in mapping data, rather they often may have increased due to an increase in the number of surveyors or the targeting of previously unsurveyed regions. In addition, it usually is unknown whether an unoccupied location that previously was occupied signifies a loss or is due to a lack of survey. One may still analyse the change of occupancy for a particular species relative to another one, but inferences are valid only if survey efforts remained the same for the compared species or habitats. Thus, it requires among others, that there was no shift in the likeliness of reporting rare and common species or habitats through time. This assumption will be violated, for example, if rare species, specific habitats, or particular geographic regions were targeted at particular times.
An alternative option that, to our knowledge, has never been considered so far, is to extract from the mapping database time series of observations that were contributed by the same observer - provided that survey efforts of the observer are also documented, i.e., one extracts a small subset of data that form monitoring data. These time series could then be analysed with methods for presence/absence or count data suitable for monitoring data. Missing data may be common in the extracted time series but methods to account for missing data and to integrate different monitoring schemes - in the case that the different observers monitored different sets of species or habitats - are available.
A second alternative may be to assess whether the distribution data could be treated like haphazardly collected monitoring data. This requires that some locations or grid cells have been surveyed in several years and some of them repeatedly within the same year. It also requires that either absences were also recorded (which is rarely the case in mapping surveys) or that all species/habitats from a list of targeted species/habitats are recorded so that absences can be reconstructed. See Estimating population trends from opportunistic observations for an explanation of the approach.
- Kery M., Royle A., Schmid H, Schaub M., Volet B, Häfliger G, Zbinden N (2010): Site-occupancy distribution modeling to correct population-trend estimates derived from opportunistic observations. Conserv. Biol. 24: 1388-1397.
EuMon core team; May 2013