Negative Values of Moisture
-------- ------ -- --------
The relative humidity can have negative values and values greater
than 100%. Some causes are:
Gibb's Phenomenon
The Reanalysis forecast model is spectral with a quadratic grid.
This means that quadratic quantities can be calculated exactly
for the resolved spherical harmonics. This is considered important
as this helps preserve kinetic energy in a adiabatic model. One
drawback of the quadratic grid is that it is prone to "Gibb-ing".
Sharp gradients gradients tend to produce "over-shoots" or "ringing"
after a transformation from grid to spectral and back to the grid
representations. It is not uncommon to have negative humidity in
regions of large moisture gradients.
Humidity Advection
Traditionally GCMs use a time step which is slightly shorter than
what would cause an instability. It is possible that a too large time
step could cause an overestimate of |dq/dt| from the moisture advection
term in regions of a large gradient.
Finite Precision of Grib
GRIB data has a finite precision and the precision for specific
humidity was set at a too low value. During winter over the polar
regions, saturated q may have, for example, a value of 0.28 g/kg.
In converting to GRIB, the moisture is converted to the nearest 0.1 g/kg.
Thus it possible for the GRIB file to have a value of 0.3 g/kg which
would be super-saturated. This quantization also plays havoc with
the variance estimates.
Effects of Negative Humidity
(1) Radiation code uses RH = 0 whenever RH < 0
(2) pressure level data: 0 <= RH <= 100
(3) precipitable water uses negative RH to preserve water budget