I just ran the 3 ECHO probes used during SCP through a beaker of sand with various amounts of moisture (manually mixed in). The results are:
Variable |
Qsoil.g |
Qsoil.c |
Qsoil.c2 |
|---|---|---|---|
S/N |
ECHO 001 |
ECHO 002 |
ECHO 012 |
ID |
28 |
29 |
2A |
0ml/2700ml = 0% |
-3.14 |
-2.95 |
-2.33 |
200/2800 = 7.1% |
3.71 |
3.91 |
4.17 |
400/2900 = 13.8% |
15.87 |
16.00 |
16.30 |
600/2900 = 20.7% |
21.05 |
22.21 |
24.02 |
800/2800 = 28.6% |
25.33 |
25.78 |
29.12 |
This suggests that there are small gain/bias errors that have probe 001 read lower that probe 002 which is lower than 012 -- in other words, an order of 001/002/012
However, the one set of gravimetric samples showed that probe 001 was within 1% of its gravimetric sample, whereas probe 002 was 2% high and probe 012 was 3% low. This would imply that the order was 012/001/002.
Furthermore, the data consistently show an order 012/002/001, though 001 being larger undoubtedly is real.
I am forced to conclude that 002 reading higher than 012 in the field is not a calibration difference. Thus, this difference must have been due to spatial inhomogeneity in the soil.
For my records, here is code to plot this:
ref = c(0,7.1,13.8,20.7,28.6)
e01 = c(-3.14,3.71,15.87,21.05,25.33)
e02 = c(-2.95,3.91,16.00,22.21,25.78)
e12 = c(-2.33,4.17,16.3,24.02,29.12)
matplot(ref,cbind(e01,e02,e12),xlim=c(-5,30),ylim=c(-5,30))
abline(h=0,v=0)
abline(0,1)