WorldClim bioclimatic variables

[Chris]

Hello

Can anybody confirm what the units are for the bioclim variables (Bio1
to Bio11) provided by the WorldClim data set?

The WorldClim website doesn't actually say what they are but does say
that the variables are derived from temperature variables that are
degrees celcius x 10 ,which makes me think I need to divide the values
by 10 if I want to know the absolute values. I don't think this will
work for Bio3 and Bio4 though (isothermality and seasonality) ....

I am working with data extracted at the 5 arc mins resolution for the
region of Nepal and the annual mean temperatures are ranging from -100
to 247 which doesn't make sense. Divided by 10 they are much more
realistic.

Another question is: from the point of view of the operation of the
Maxent model does it matter that the values are scaled up, so long as
they are correct relative to each other?

Any thoughts or ideas on the above would be greatly appreciated,
thanks.

[Peter D. Wilson]

Chris,

WorldClim temperature variables are degrees Celcius x 10 and chopped to
an integer. This means that you can only recover true values to one
decimal place when you divide by 10 to convert back to prpoer
temperatures. Precipitation is in whole millimetres.

Isolthermality is a dimensionless ratio between Meant monthly
temperature and Annual temperature range. the stored value in WorldClim
files is to 2 decimal places, so it is the true value times 100 and
chopped to an integer.

Temperature seasonality is calculated for WordClim data as 100*(standard
deviation of monthly mean temperature) and stored as an integer. It has
the same units as temperature.

The answer to your second question is that MaxEnt does not require data
to be in particular units, or to be real numbers or integers. It does
require that data used to train (fit) a model is exactly the same scale
and units as data used to project the model, say onto future climate
conditions. So, you do not need to rescale data from WorldClim to use it
in a MaxEnt model.

Cheers,

Peter Wilson

[Marnin Wolfe]

Sort of an add on question...

Does anyone have an opinion on standardizing the variables by subtracting their mean and dividing by the standard deviation? This puts all variables in the same units. Since MaxEnt uses variable covariances, I am guessing its response curves are affected by the scale differences!

Would it be a more fair interpretation of the relative importance of variables if we examined the standardized ones?

Cheers,

Marnin

[Chris]

Thanks very much, Peter. Your explanation helps clarify how WorldClim
data is constructed. In MaxEnt my projection data will be the same
units and scale as the fitting data.
Chris