Sensing Organic Resource Quality

 

Portable reflectance spectroscopy can be used to predict N concentration in a wide range of organic inputs for soil fertility management, and to screen materials into management categories based on resource quality and decomposition characteristics.

Over 500 plant samples with diverse origins and quality characteristics were selected from the Organic Resource Database System (TSBF - Imperial College Wye). The materials included leaf, stem and bark materials from a range of tree species, crop residues, composts and animal manures. Samples were oven-dried at 60 degrees C and ground to pass a 1 mm sieve. Diffuse reflectance spectra were collected from samples placed in glass petri dishes using a high intensity reflectance probe, as shown on the Sensing Soil Quality page.

 

Plant properties in the ICRAF Spectral Library

Property

n

Min

Max

N†

319

0.13

6.6

N‡

514

0.13

6.6

P

351

0.01

1.2

K

306

0.04

6.6

Ca

229

0.01

6.9

Mg

220

0.02

2.4

Lignin

319

1.9

44

Soluble polyphenols

319

0.01

15

† Samples for which lignin and polyphenol were also determined.
‡ All samples for which N was determined.

Lignin, soluble polyphenol, and N concentrations showed a wide range, and were not strongly correlated in this data set (see below).

 

Correlation coefficients (upper triangle) and partial correlation coefficients (lower triangle) for lignin, soluble polyphenol and nitrogen concentrations in the organic resource materials (n=319).

Property†

Lignin

Sol phenol

N

Lignin

–

0.00

-0.34

Sol phenol

-0.10

–

 0.26

N

-0.36

0.27

–

† Transformations were applied prior to analysis to obtain approximately multivariate normally
distributed values: ln lignin; ln Sol phenol.

 

The plant properties were calibrated to reflectance spectra using the 1000 to 2500 nm wavelength region using partial least squares regression. Prediction performance was assessed using full cross-validation and also on an independent test of 32 samples chosen to represent a wide range of organic resource qualities.

Nitrogen and soluble polyphenols gave reasonably good calibrations, but the calibration for lignin was only moderate.

 

Predicted values from spectral reflectance versus actual values for N concentration on an independent validation set of 32 samples from the Organic Resource Library.

Spectral screening tests were developed for high and low levels of N, lignin and soluble polyphenol (see below). The tests achieved moderately high levels (>80%) of sensitivity and specificity in cross-validated runs.

 

 

A decision tree for selecting organic materials as a N source in biomass transfer systems has been proposed (Palm et al., 2001) based on the above cut-off limits on N, lignin and soluble polyphenol concentrations.  A spectral test achieved a high success rate in screening samples into the four management categories (see below).