Soil Organic Matter Management: The Roles of Residue Quality in C Sequestration and N Supply - Sustainable Management of Soil Organic Matter - page 176

Table 3.10.1. NIRS process and results.

1. Calibration process: scan a set of 'known' samples - classified on the basis of colour, chemistry
and presence/abundance of artefacts - and analyse with the calibration programme to develop
a function with the fewest possible terms.
Provided data: 13 Terra Preta (TP)

24 Terra Mulata (TM)
28 Background Sands and Clays (SC)
65 Total known samples

2. Test #1: use the function developed above to discriminate between a set of 'unknown' samples
that included the original 65 samples.
Provided data: 74 TP Results received: 74 TP

49 TM a

48 TM

85 SC 85 SC

208 Total unknown 207 Total discriminated samples

3. Re-calibration process: the 207 discriminated samples were reprocessed with the calibration
programme to create a new, more accurate function with more terms and tighter tolerances.
Provided data: 74 TP

48 TM
85 SC
207 Total known samples

4. Test #2: the revised function was used to process the same set of data used in test #1.
Results received: 71 TP

44 TM
83 SC
198 Total discriminated samples

The resulting rejections were as follows:
Of 208 samples: Rejections consisted of: Providing an accuracy of:

74 TP 2 TM 1 SC 95.9% for TP

49 TM 3 TP 2 SC 89.8% for TM

85 SC 1 SC 1 TM 97.6% for SC

a One of the TM samples was classified as TP.

for the specific organic compounds that are affecting these spectra, there
has not been sufficient work to demonstrate what these may be in our soils.
Fortunately, this analysis did not appear to have been affected by the
long-standing problems resulting from particle size variation, scatter and
multi-colinearity encountered by other investigators (Barnes et al ., 1989).
We can therefore conclude then that intergroup variation is not a function
of soil texture, water content or grinding.