Welcome to Climate Matcher!
The ability of correlative species distribution models (e.g.
MaxEnt
) to
accurately predict outside the fitted region is highly contingent on
the degree to which predictions are made under novel environmental conditions.
This is particularly important when modelling the potential distribution of invasive species.
Climate Matcher
is an easy to use tool that can readily estimate and map areas where
a model is likely to be extrapolating into novel environmental conditions, and thus, identify areas
where a correlative SDM is likely to be unreliable.
Climate data used in this tool:
This tool uses WorldClim2 (10 min; ~340 km 2 ) global climate raster data, with 19 climate variables available for selection.Selecting reference locations:
Reference areas are the region you plan to fit your model too. This should encompass the area in which an invasive species is thought to have spread or the region which is used to fit a correlative SDM. These regions can be selected by either choosing relavant countries from a dropdown menu, or by uploading a relevant input data source. Input data can be:- A .csv file: containing point coordinates of both species occurrences and background/pseudo absences over the extent of interest. Must contain 'latitude' and 'longitude' named columns; or
-
A .rds file:
containing a spatial object (e.g. SpatialPoints, sf object or other vector data).
If you have a .shp file, we recommend loading it into R using thesfpackage and then saving the output as an .rds for use with this tool.
Two methods for identifying novel climate are used in this tool:
-
Extrapolation detector(ExDet; Mesgaran et al. 2014 ):
An algorithm that can detect, distinguish
and quantify two types of extrapolation: novel univariate range and novel combinations of covariates
(i.e. novel correlations between covariates). Here negative value represents a novel environmental range.
Values between 0 and 1 are considered to contain analagous conditions, and those >1 are deemd to
contain novel covariate correlations.
-
Multivariate Environmental Similarity Surface(MESS; Elith et al. 2010 ):
An algorithm that distinguishes and quantifies
novel covariate ranges not present in the fitted/reference area.
This method can be used on any number of climate covariates and
will report the minimum extrapolation value per raster cell. Here a negative value represents
a novel environmental range. The more negative the greater the extrapolation. Between 0 and 100 is
considered analagous climatic conditions, with values close to 100 deemed near the median of the reference
area.
Note this method can only be used when two or more covariates are selected.
This method does not account for changes correlation structures among covariates (i.e. new combinations of covariates) but can be used on any number of covariates.
Climate Matching
NOTE: the climate matching algorithm is currently under development and requires additional testing before using to inform management. Use at own risk.The tool also contains a prototype implementation of CLIMEX's ( Kriticos et al. 2015 ) climate matching algorithm (CMI). This algorithm has been repurposed to work with WorldClim2 data and covariates. Specifically, it attempts to estimate a 'Composite Match Index (CMI)' as a product of the average temperature and precipitation indexes.
CMI is estimated as follows:
- Temperature Index: $$\text{Mean Temperature Index}_i = \frac{\sum\limits_{j=1}^{n}e^{(-\beta\times \Delta T_{i,j})}}{n}$$ Here i refers to a location of interest and j refers to a temperature variable from n selected temperature variables. \(\Delta T_{i,j}\), is the long-term absolute differences in the temperature variable between the reference area and the location i. \(\beta\) is a constant exponential rate parameter that defines the importance of a unit change in variable j. By default \(\beta\) is set to 0.1 for all temperature covariates (BIO1 to BIO11). This means that a difference of 1 °C results in a index score of 0.9 and a difference of 5 °C results in a 0.6 index score. As the importance of each variable is likely to differ depending on the species, we've allowed users to redefine \(\beta\) in the parameter settings menu (only shows up when CMI is selected)
- Precipitation Index: $$\text{Mean Precipitation Index}_i = \frac{\sum\limits_{j=1}^{n}e^{(-\beta\times \Delta P_{i,j}).}}{n}$$ Where, $$\Delta P_{i,j} = \frac{\text{abs}(P_R - P_j)}{1 + \alpha (P_R + P_j)}$$ Here i refers to a location of interest and j refers to a temperature variable from n selected precipitaion variables. \(\Delta P_i\), is the difference in annual rainfall between the reference location, \(P_R\), and the matching location, \(P_j\) adjusted so that a 100 mm difference in rainfall is more significant for locations with lower rainfall. By default this tool sets \(\alpha\) and \(\beta\) to 0.001 and 0.0004 respectively for all precipitation variables (BIO12 to BIO19). With these values, a difference in rainfall of 200 mm per year between two locations will result in an index of 0.64 if the average rainfall for the two locations is 400 mm, and a value of 0.85 if the average rainfall is 2000 mm. However, again, the importance of these variables will vary by species, and as such have provided the userwith the ability to change these variables under the parameter settings.
- Composite Matching Index (CMI): $$\text{CMI}_i = \left(\text{Mean Temperature Index}_i \times \text{Mean Precipitation Index}_i\right)^{\frac{1}{2}}$$