This is a late reply to the comments by Mike O'Kane concerning calibration
of soil moisture sensors.

These comments are about commercial use of soil moisture sensors, for
scheduling irrigations, identifying through-drainage and rooting depth,
irrigation efficiencies, sprinkler uniformity, etc.  Typically a grower will
be monitoring soil profiles at 30-100 sites, with the possibility for annual
crops of the sensors needing to be re-installed each year.

Because of the variabilty of soil moisture, to get a 'reasonably' accurate
soil moisture value from gravimetric tests about 15-20 samples are required.
At each sensor, therefore, 15-20 samples will be required at the dry end,
and another 15-20 at the 'wet' end.  With say, 6 sensors, this is about 300
samples.  Now a major problem is that unless these samples *are* taken, and
taken carefully, the resulting calibration equation can be wrong.  To give a
specific example, a State Department of Agriculture here in Australia
decided that their soils needed special calibrations.  The final printout
was from Minitab (on computer paper...), with excellent r^2 values.
Unfortunately, the soil samples were taken in a narrow range and as the soil
dried the equation predicted a negative water content!

It seems to me that unless users are extremely carefull they can very easily
be worse off re-calibrating at each site.  Surely their best check is to see
whether their soil monitoring system is giving changes in soil water content
commensurate with applied volumes.  If 10 mm is applied, did the sensor
record 10 mm - its a good check on any system.  Commercial consultants are
looking at *differences* in water content - has it got wetter, and if so by
how much.

Installation is another problem, as outlined by Bruce on ?/?, that must be
considered.  Once again, a commercial grower may need to install 30-100
sites at the start of each season.  (It has been reported that consultants
are charging $1-2,000 for each installation of a particular soil moisture
system)  Is the sensor calibration sensitive to 'practical' installation
techniques.  Does the calibration change if the soil cracks, or the sensor
moves slightly ?

Although automatic loggers are very attractive compared with a neutron
probe, their full cost needs to be considered.  For example, if the logger
is at the sensor it must still be visited (by a reasonably skilled operator)
to download data.  If the logger is connected to a central station by cable,
then the cost and more importantly the maintenance of the cabling can be
very high.  Remote access to the soil moisture sensor is attractive - but
nevertheless most agronomists I have met wouldn't dream of making irrigation
decisions without actually looking at the crop.

Neutron probes still have a lot of pluses.  Access tubes are easily
available, can be installed cheaply and in quite appalling soils (how does
RSA grow such fantastic crops in what is mainly gravel and rock ?); they can
be moved easily; lateral profiles (around sprinklers) can be easily measured
and then abandoned; they can measure to almost any depth (we have a client
just about to install 12 m tubes at different radial distances from tree
wind breaks); the instrument itself is the calibration standard for other
instruments;

Nevertheless, a reasonably accurate, easily installed, and loggable soil
moisture system is required.  It seems to me that the sensor will have to be
buried in the soil (eg TDR) to measure a an 'adequate' volume of soil.
Incidentally a trial here in Australia of DeltaT ThetaProbes installed
horizontally into the walls of a bore hole and using the manufacturers
calibration are agreeing with neutron probes within about 2% VSW.

Measurement frequency is an interesting problem. When I first saw hourly
data I was fascinated to watch the fluctuations in crop water use as clouds
passsed over the sun; but watching consultants make irrigation decisions has
shown me that there can very easily be too much data.  From a purely
arithmetic point of view, how do you calculate daily water use from, say,
hourly data ?  I tried smoothing and digital filters and similar solutions,
but in practice the best method is the difference between the, say, 6 am
reading on one day and the 6 am reading the next day - that is ignore all
the other data.  In fact, with a crop on say a 7-14 day irrigation cycle,
experienced consultants will say that they only need 2-3 readings in that
period, and any more just confuses their analysis.  I have seen them totally
ignore and not plot readings taken too frequently.

With a logger I think we need a recording system whereby a user could
temporarily switch to say hourly logging to pick up the drainage after an
irrigation, but then for routine use it should be 24 hours, at say 6 am.
But try getting a logger manufacturer to do that...

We have a client about to install a network of buried TDR's (probably the
Delta-T ThetaProbe), connected to a small radio transmitter at the surface.
The radio link will work over about 1 km, runs on dry cells (ie no
re-charging), will take 4 sensors, and will cost about $AUS400 (plus
sensors).  The system could also handle the Campbell sensor, or in fact
other sensors such as temperature.  The basic logger itself is about
$AUS1,200, but can have almost unlimited extra channels added.


----
Trevor Finch
Research Services New England
8/16 Nicholson St, Balmain NSW 2041 Australia
email: rsne@mpx.com.au
tel: +61 (2) 810 3563
fax: +61 (2) 810 3323