Visible and Near-Infrared Calibrations for Quality

Assessment of Fresh Phase I and II Mushroom

(Agaricus bisporus) Compost

H. S. S. SHARMA,* M. KILPATRICK, G. LYONS, S. STURGEON, J. ARCHER,

S. MOORE, L. CHEUNG, and K. FINEGAN
Department of Applied Plant Science, School of Agriculture and Food Science, The Queen’s University of Belfast, Applied Plant
Science Division (H.S.S.S.); Department of Agriculture and Rural Development for Northern Ireland, Newforge Lane, Belfast BT9
5PX (G.L., S.M., L.C.); and Applied Plant Science Division, Department of Agriculture and Rural Development for Northern
Ireland, Loughgall, Co. Armagh, BT61 8JB, UK (M.K., S.S., J.A., K.F.)

Previous studies have shown that visible and near-infrared spectra
(Vis-NIR) of dry and milled compost can be used for generating
partial least squares (PLS) calibrations of phase II compost param-
eters including ammonia, nitrogen dry matter (NDM), dry matter
(DM), pH, conductivity, carbon, microbial population, and potential
productivity. The objective of this study was to develop robust cal-
ibrations for some of the key parameters from the spectra of fresh
phase I and II composts. Samples of substrates from six commercial
production yards were obtained during winter and summer months
of 2000–2004 to monitor changes in quality and were analyzed for
the test factors. Vis-NIR reflectance measurements of fresh samples
(740) were made over the range of 400–2500 nm. After mathemat-
ical pretreatments, PLS calibrations of the key parameters were
developed using the NIR (1100–2500 nm) and visible and NIR (400–
2500 nm) regions and subsequently validated using an independent
sample set of 123 phase I and II samples obtained during 2004–
2005. The phase I and II standard errors of laboratory measure-
ments of ammonia, pH, conductivity, DM, NDM, and ash were low-
er than the standard error of predictions of the same parameters,
respectively, by the best NIR or Vis-NIR models. The degree of
precision for some of the calibrations, especially ammonia, NDM,
and DM, is suitable for composters to monitor changes in quality
parameters during production. The laboratory measurement errors
for phase I samples were greater than those of the phase II samples,
except for ash, due to a higher degree of heterogeneity in the sub-
strate. The calibrations, especially for pH, conductivity, and ash,
need to be improved with new sample sets. A major advantage of
NIR spectroscopy is the ability to assess substrate quality for a
range of target parameters simultaneously, within a few hours of
receiving the samples. The main drawbacks are the expensive in-
strumentation, expertise, and training necessary for operating the
spectrometer and a dedicated chemometrician required for main-
taining the equations compared to the reference methods.
Index Headings: Visible spectroscopy; Near-infrared spectroscopy;
NIR spectroscopy; Phase I and II mushroom compost; Key quality
parameters.
INTRODUCTION
The important factors influencing substrate quality are
ammonia, dry matter (DM), nitrogen dry matter (NDM),
pH, conductivity, ash, C/N ratio, and microbial popula-
tion. Changes in these parameters during the two phases
of production are associated with microbial and biochem-
ical markers.1–6 The modifications in some of the impor-
tant characteristics can promote rapid colonization of the
Received 31 May 2005; accepted 31 August 2005.
* Author to whom correspondence should be sent. E-mail: s.sharma@
qub.ac.uk.
0003-7028 / 05 / 5911-1399$2.00 / 0
Volume 59, Number 11, 2005 q 2005 Society for Applied Spectroscopy
substrate by the mushroom mycelium, while limiting the
growth of weed mold fungi.5
Reports on the use of near-infrared spectroscopy
(NIRS) for the quality assessment of compost have been
published by researchers from the UK and the Nether-
lands.7–10 Although, Resink and van den Hurk7 have pub-
lished calibrations for predicting moisture, total nitrogen,
ash, pH, ammonia, acid detergent fiber, neutral detergent
fiber, and crude fiber from fresh samples, their report fo-
cused on the calibration transfer of the models for mois-
ture and nitrogen from a master instrument to six similar
spectrometers (Foss NIRSystem 6500 and 5000 models),
and they did not report on the performance of the re-
maining six calibrations. The samples used in that study
were obtained mainly from Dutch and Belgian producers,
where indoor composting is the dominant system for both
phase I and II production stages.11 In the UK and Ireland,
phase I consists of an outdoor bunker or windrow stage12
for a duration of 10 to 24 days and is followed by pas-
teurization and conditioning in insulated phase II tunnels
for 6 to 7 days. Our initial research was focused on the
generation of calibrations for the important parameters
including potential yield from dried and milled phase II
composts,9,10 followed by the development of preliminary
calibrations for the quality assessment of fresh materials.8
The practical problems associated with the develop-
ment of NIR calibrations for plant materials with high
moisture content, such as silage, have been reported by
various research groups. Mushroom compost can contain
70–76% moisture at phase I and 63–71% at the phase II
stage, with the moisture peaks at 1450 and 1930 nm dom-
inating the near-infrared spectra of samples.13 In addition,
sampling at the phase I stage is labor intensive due to the
large volume of materials used, typically 50–150 tonnes
per batch, and the inherent heterogeneity of the blended
materials.13 With an extensive sample population from the
industry, NIRS should be able to generate robust calibra-
tions for predicting key characteristics of fresh phase I
and II substrates.
The main advantage of calibrations that can be used
for assessing fresh compost is the need for minimal prep-
aration time by eliminating drying and milling of sam-
ples, prior to quality assessment. The aim of this study
was to expand the sample size of the two production
stages and develop robust calibrations for accurate pre-
diction of ammonia, pH, conductivity, NDM, DM, and
ash of fresh phase I and II composts that can be used for
APPLIED SPECTROSCOPY 1399
TABLE I. Comparison of minimum, maximum, mean, and standard deviation (SD) of key parameters of the phase I (312) and II (428)
calibration sample set obtained from six compost yards.
Parameter Number Minimum Maximum Mean SD
a. Phase I sample set
Ammonia (%) 201 0.15 1.14 0.64 0.16
pH 292 7.23 9.11 8.16 0.31
Conductivity (ms/cm) 292 1.28 4.60 2.94 0.55
Dry matter (%) 290 18.52 33.29 25.91 2.46
Nitrogen dry matter (%) 263 0.75 2.10 1.43 0.22
Ash (%) 157 11.55 20.66 16.1 1.52
b. Phase II sample set
Ammonia (%) 324 0.001 0.225 0.045 0.06
pH 425 6.85 8.5 7.68 0.27
Conductivity (ms/cm) 428 2.13 5.15 3.64 0.50
Dry matter (%) 427 25.32 39.18 32.25 2.31
Nitrogen dry matter (%) 363 1.87 2.79 2.33 0.15
Ash (%) 343 15.69 29.27 22.49 1.20

monitoring changes in the quality parameters during pro-
duction.
MATERIALS AND METHODS
Sampling of Compost. The raw materials used for
commercial production of synthetic composts were wheat
straw, poultry litter, and gypsum mixed at a typical ratio
of 1000:450:10 kg.14 Three composters were selected on
the basis of differences in their production systems, such
as bunker, windrow, and duration of the phase I stage,
and further details on the yard management protocols and
sampling of phase I and II have been previously de-
scribed.14,15 In addition, five productivity trials of phase
II composts from six yards were also carried out during
2001–2004,16 generating 287 samples that were incor-
porated into the phase I and II calibration sets. Three
replicate samples (10 kg) were taken from the production
line at the end of the phase I and II stages, followed by
sub-sampling prior to analyses. The number of samples
of phase I and II composts from the three production
yards were 312 and 141, respectively, resulting in a total
population of 740 samples that included materials from
the phase II (287) compost comparative trials.16
Near-Infrared Spectroscopic Analysis. Visible and
near-infrared spectra of fresh samples were obtained in
the reflectance mode, over a wavelength of 400–2498 nm
at 2 nm intervals, set at a band pass value of 10 nm with
a wavelength accuracy of 0.5 nm, and the instrument was
calibrated before each analysis. The spectrometer (Foss
6500, Silver Spring, MD), housed in a controlled envi-
ronment maintained at 20 8C and 65% relative humidity,
was equipped with a transport device for a natural prod-
uct cell (21 3 5 3 4 cm) with a quartz glass window.
The instrument was operated by Win ISI, version 3 (In-
frasoft International Inc, Port Matilda, PA). Samples were
preconditioned for 2 to 3 h in the controlled environment
prior to analysis in order to minimize spectral differences
caused by variations in sample temperature. The test sam-
ple was packed carefully in the rectangular cup and
scanned six times, using approximately 50 g sub-samples
at a time. Reflectance measurements were converted to
absorbance (A) values with the use of the following equa-
tion: A 5 log(1/R). The spectra of all replicate scans were
examined, using the proprietary software before averag-
ing and mathematical transformation. Further evaluation
of the Vis-NIR data was achieved by principal compo-
nent analysis (PCA) to assess spectral variations within
the sample population. The relationship between the
spectrum and reference values of the test parameters was
studied using a modified partial least squares (PLS) and
a principal component regression (PCR) method. The
global H (Mahalanobis distance) for outlier detection of
spectra, T, for reference data, and PLS term were set at
2.5, 2.5, and 16, respectively.
Analysis of Test Samples for Key Parameters. Com-
post samples (1 kg) were dried in a large (2 3 2 m)
industrial-size fan oven by spreading the samples out on
a tray overnight and the dry matter (DM) content of the
materials was determined. The temperature of the oven
was set at 85 8C in order to minimize damage to the test
materials, as samples were further analyzed for fiber frac-
tions (result not reported). The dried materials were
milled using a Cyclotec mill (1093) to pass through a 0.5
mm mesh screen. The remaining fresh samples were used
for determining ammonia, pH, and electrical conductivi-
ty. Fresh compost (40 g) was shaken with 400 mL of
sterile, deionized water for one hour, followed by stan-
dard pH and conductivity measurements.16 Ammonia
content of the fresh samples was measured using standard
protocols (Tecator Application AN 101/88).17 The dried
and milled samples were used to determine the nitrogen
content by the Kjeldahl method (Tecator Application
Note AN 300)17 and the ash content and were corrected
for residual dry matter.
Validation Test. A validation test to examine the ac-
curacy of the models was set up during the summer and
winter months of 2003–2005, using 123 phase I and II
samples obtained from the commercial production yards.
Sampling protocols, preparation of the materials for Vis-
NIR scanning, and analyses of the samples for the key
parameters were as described above.
RESULTS AND DISCUSSION
Key Parameters. The phase I and II samples of the
calibration and validation sets are presented in Tables I
and II, respectively, to show range, mean, and standard
deviation (SD) of ammonia, pH, conductivity, NDM,
DM, and ash. The range, mean, and SD of the parameters
in the validation set were similar to the calibration set,
except for ammonia, which was slightly lower in the cal-

1400 Volume 59, Number 11, 2005

TABLE II. Comparison of minimum, maximum, mean, and standard deviation (SD) of key parameters of phase I (38) and II (85) validation
sample set obtained from six compost yards.
Parameter Number Minimum Maximum Mean SD
a. Phase I sample set
Ammonia (%) 18 0.28 0.899 0.567 0.15
pH 36 7.75 8.46 8.08 0.19
Conductivity (ms/cm) 36 2.46 3.90 3.21 0.399
Dry matter (%) 36 22.95 28.97 26.68 1.55
Nitrogen dry matter (%) 18 1.30 1.761 1.59 0.15
Ash (%) 38 12.88 20.21 16.36 1.85
b. Phase II sample set
Ammonia (%) 39 0.005 0.220 0.054 0.056
pH 60 7.30 8.22 7.64 0.24
Conductivity (ms/cm) 60 2.95 4.90 3.80 0.46
Dry matter (%) 60 24.53 37.42 32.12 2.49
Nitrogen dry matter (%) 41 1.93 2.70 2.33 0.19
Ash (%) 85 15.80 26.51 21.65 2.55

ibration set, indicating that better quality substrates were
sampled for the validation set, possibly due to favorable
weather conditions during production and optimum qual-
ity of raw materials used. Table I also shows changes in
substrate composition due to microbial breakdown of
straw at the two production stages, resulting in a more
uniform material at the end of phase II, as indicated by
lower SDs for all parameters. The phase I samples are
usually associated with high pH, ammonia, and moisture
content, and in contrast phase, II substrates contain high
levels of NDM, DM, ash, and conductivity.18,19 Mean pH
(8.16) and ammonia (0.64%) content of phase I samples
declined at the end of phase II, and NDM, DM, conduc-
tivity, and ash concentrations increased at phase II (Table
I); comparable changes in all parameters were also ob-
served in the validation set (Table II). The range of mea-
sured values for ammonia, pH, conductivity, DM, NDM,
and ash from the two sample sets are similar to previous
reports.7,14,20
Spectral Evaluation. Comparisons of raw spectra of
three phase I (Fig. 1a) and three phase II (Fig. 1b) com-
posts of poor, medium, and good quality substrates, as
indicated by the diversity in the composition, are pre-
sented to show differences between samples in the visible
and NIR segments. Further examination of the trans-
formed spectral overlays (Figs. 2a and 2b) of phase I and
II samples indicated that major differences in the spectra
could be detected in the 500–550, 635–760, 810–900,
1100–1200, 1350–1400, 1460–1500, 1720–1800, 2000–
2100, and 2230–2400 nm wavelengths for phase I and in

FIG. 1. (a) An overlay of raw visible and near-infrared spectra of poor,
medium, and good quality phase I composts showing differences in the
spectral segments. (b) An overlay of raw visible and near-infrared spec-
tra of poor, medium, and good quality phase II compost showing dif-
ferences in the spectral segments.
FIG. 2. (a) An overlay of transformed visible and near-infrared spectra
of poor, medium, and good quality phase I composts showing differ-
ences in the spectral segments (arrows). (b) An overlay of transformed
visible and near-infrared spectra of poor, medium, and good quality
phase II compost showing differences in the spectral segments (arrows).

APPLIED SPECTROSCOPY 1401

TABLE III. Calibration statistics for NIR and Vis-NIR segments of phase I and II samples with first- and second-derivative mathematical
transformations (MT) showing R2 for calibration with standard error for calibration (SEC), R2 for cross-validation with standard error for
cross-validation (SECV), and terms used.
Sample R2 for R2 for
Key parameter number SEC calibration SECV cross-validation Terms
a. 1100–2500 nm, 1441 MT
Ammonia (%) 587 0.038 0.982 0.041 0.979 13
pH 728 0.160 0.833 0.167 0.817 14
Conductivity (ms/cm) 728 0.219 0.888 0.233 0.874 14
DM (%) 770 0.739 0.965 0.780 0.961 14
NDM (%) 631 0.107 0.949 0.114 0.941 12
Ash (%) 481 1.161 0.862 1.239 0.842 13
b. 400–2500 nm, 1441 MT
Ammonia (%) 585 0.038 0.981 0.041 0.979 10
pH 732 0.150 0.852 0.162 0.829 15
Conductivity (ms/cm) 727 0.209 0.898 0.226 0.881 14
DM (%) 775 0.758 0.963 0.803 0.959 12
NDM (%) 630 0.102 0.953 0.109 0.946 11
Ash (%) 477 1.211 0.851 1.272 0.836 11
c. 1100–2500 nm, 2861 MT
Ammonia (%) 586 0.039 0.981 0.042 0.978 9
pH 730 0.159 0.834 0.198 0.74 13
Conductivity (ms/cm) 724 0.213 0.894 0.350 0.870 14
DM (%) 766 0.726 0.966 0.779 0.961 13
NDM (%) 621 0.109 0.946 0.115 0.940 9
Ash (%) 489 1.24 0.843 1.299 0.827 8
d. 400–2500 nm, 2861 MT
Ammonia (%) 583 0.385 0.981 0.041 0.979 9
pH 725 0.144 0.865 0.155 0.843 14
Conductivity (ms/cm) 724 0.206 0.894 0.233 0.875 14
DM (%) 768 0.726 0.966 0.774 0.961 11
NDM (%) 629 0.103 0.952 0.109 0.947 7
Ash (%) 479 1.149 0.865 1.240 0.844 9

the 522–540, 588–678, 796–1234, 1248–1360, 1248–
1360, 1472–1528, 1870–1920, 1942–2128, 2248–2258,
and 2294–2494 nm wavelengths for phase II samples.
The majority of the spectral segments listed above are
linked to C–H, O–H, and N–H stretching and bending, 20
and the differences could be due to breakdown of the
fiber fractions, moisture content, pH, ammonia, and lig-
nin-humus formation. The main spectral differences be-
tween fresh phase I and II samples could be identified in
the following segments: 672–932, 1110–1430, 1594–
1664, 1948–1990, 2042–2072, 2120–2206, and 2432–
2468 nm. Sharma et al. 21 have previously reported that
the spectral segments listed above are key indicators of
high-yielding phase II substrate.
Principal component analysis of the Vis-NIR spectra
of the calibration samples (740) was conducted and the
results showed that 12 phase I and 15 phase II outlier
samples were grouped separately from the main popula-
tion, indicating significant differences possibly due to
over- or under-composting, use of poor quality raw ma-
terials, or unfavorable weather conditions at the outdoor
stage (results not presented). One of the six composters
followed a 17–24 day phase I stage, resulting in a more
homogeneous but over-composted substrate. Consequent-
ly, the samples originating from that production yard
showed a higher degree of similarity in the replicate spec-
tral scans compared to spectra of substrates produced by
a shorter 10–14 day phase I stage. 21
Calibration. Since fresh composts, especially phase I,
are variable in straw length and degree of breakdown,
pretreatment of the spectral data with standard normal
variate-detrend (SNV-D) was necessary to minimize scat-
ter effects and possible co-linearity between chemically
unrelated wavelengths. In addition, first- (1441) and sec-
ond- (2861) derivative mathematical transformations
(MT) of the 400–1100, 1100–2500, and 400–2500 seg-
ments were examined to determine the best calibration
equations with the lowest SECV values of the parame-
ters. The 400–1100 nm segment performed poorly com-
pared to the other spectral segments (results not present-
ed). PCR equations were found to be inferior to models
developed using PLS regression methods (results not pre-
sented) and only PLS equations generated from 1100–
2500 nm and 400–2500 nm bands are presented in this
report. The equations developed from the combined
phase I or II calibration sets were better than models de-
veloped from separate phase I and II sets (results not
presented) most likely due to the extended range of val-
ues for each parameter listed in Table I.
Table III shows the performance of PLS equations gen-
erated from the calibration set consisting of phase I and
II samples. The impact of employing the 1100–2500 nm
segments on the cross-validation error of equations was
assessed for all parameters. Whenever the Vis-NIR spec-
tral data (400–2500 nm) was used for calibration, the
standard errors for cross-validation (SECV) were margin-
ally reduced for parameters such as pH, conductivity,
NDM, and ash content; in contrast, a lower SECV value
for DM was obtained only from the 1100–2500 nm seg-
ment (Table III). Comparison of the equations developed
from the first- and second-derivative transformations
showed insignificant differences in the R 2 and SECV val-

1402 Volume 59, Number 11, 2005

TABLE IV. Validation statistics for all calibration equations using an independent set of 123 phase I and II samples showing the number
of samples for each parameter, R2 for correlation between predicted and measured values, slope, and bias of the NIR and Vis-NIR segments
mathematically transformed (MT) using first and second derivatives.
Key parameter model No. of samples R2 SEP Slope Bias
a. 1100–2500 nm, 1441 MT
Ammonia (%) 57 0.916 0.078 0.927 0.015
pH 96 0.546 0.251 0.782 20.007
Conductivity (ms/cm) 96 0.803 0.242 0.868 20.125
DM (%) 96 0.967 0.666 0.940 0.094
NDM (%) 59 0.919 0.113 1.079 20.005
Ash (%) 123 0.733 2.448 0.686 21.862
b. 400–2500 nm, 1441 MT
Ammonia (%) 57 0.907 0.080 0.942 0.003
pH 96 0.491 0.274 0.867 20.166
Conductivity (ms/cm) 96 0.735 0.271 0.912 0.019
DM (%) 96 0.964 0.693 0.941 0.147
NDM (%) 59 0.905 0.125 1.12 0.003
Ash (%) 123 0.771 1.668 0.826 21.097
c. 1100–2500 nm, 2861 MT
Ammonia (%) 57 0.921 0.074 0.940 20.001
pH 90 0.704 0.189 0.886 20.087
Conductivity (ms/cm) 96 0.798 0.243 0.876 0.002
DM (%) 96 0.960 0.749 0.956 0.260
NDM (%) 59 0.902 0.127 1.091 0.028
Ash (%) 123 0.731 1.931 0.760 21.288
d. 400–2500 nm, 2861 MT
Ammonia (%) 57 0.913 0.078 0.939 20.011
pH 96 0.621 0.227 0.995 20.129
Conductivity (ms/cm) 96 0.768 0.274 0.909 0.102
DM (%) 96 0.952 0.879 0.924 0.368
NDM (%) 59 0.877 0.145 1.108 0.038
Ash (%) 123 0.852 1.010 0.957 20.354

ues for all parameters. The best models explained 98, 84,
88, 96, 95, and 87% of the variations in ammonia, pH,
conductivity, DM, NDM, and ash content, respectively,
of the sample population, using 7–14 terms (Table III).
The slope and bias of the equations were, respectively,
1.02 and 0.005 for ammonia, 0.99 and 20.012 for pH,
1.01 and 0.002 for conductivity, 0.99 and 0.043 for DM,
1.00 and 20.001 for NDM, and 1.00 and 20.005 for ash.
Resink and van den Hurk7 have reported robust models
for some of the parameters listed above, but their mea-
surements were based on wet weight analysis and results
cannot be directly compared to the performance of the
equations from this study.
During composting, the rate of color change in sub-
strates can be influenced by variations in raw material
quality, outdoor weather conditions, and the degree of
microbial breakdown,9 suggesting that the visible part of
a spectrum may contribute significantly to the overall
changes in quality characteristics. The use of the visible
segment could enhance accuracy of calibrations for some
of the parameters likely to be linked to a change in color,
such as pH, conductivity (Table III), water-soluble car-
bohydrates and polyphenols, rate of fiber breakdown, or
the formation of lignin-humus fractions. Resink and van
den Hurk7 reported that accurate calibrations for assess-
ing moisture, pH, ash, ammonia, NDF, ADF, and crude
fiber have been developed from the 1100–2500 nm spec-
tral segment using a large sample population, but they
did not report the full validation results of the equations,
such as SEP or bias of the models, using an independent
sample set.
Validation. The calibration models were validated
with a phase I and II sample set (123), obtained from the
industry during 2003–2005. The validation statistics for
each parameter, showing R 2 for correlation between the
predicted and measured values, standard error of predic-
tion (SEP), slope, and bias, is presented in Table IV. The
slope and bias values for the best ammonia, DM, and
NDM models were close to the optimum values of 1.0
and 0.0, respectively, for a majority of the equations, but
not for pH, conductivity, and ash. This could be due to
the fact that the latter models predicted phase II samples
better than phase I composts (results not presented) and
also due to higher SEP values, confirming lack of ro-
bustness. The high SEP values for pH and conductivity
also indicated that the two factors, although easy to mea-
sure, could be influenced by errors in the repeatability of
the measurement protocols. The accuracy of a reference
method is the most important factor affecting the perfor-
mance of the NIR regression equation to predict a target
parameter. The inorganic constituents in compost cannot
be directly detected by the instrument, as minerals do not
absorb radiation in the NIR region. 22 However, minerals
associated with an organic compound, such as humus,
could be indirectly calibrated, but with a lower degree of
accuracy, as indicated by the high SECV value for ash
(Table III).
In order to improve the accuracy of equations, the cal-
ibration and validation sets were combined and recali-
brated to generate optimized models of the parameters.
Consequently, improvements in R 2 and SECV of all pa-
rameters including pH, conductivity, and ash content
have been observed (results not presented). However, fur-
ther sampling, selection of suitable samples, and re-cali-

APPLIED SPECTROSCOPY 1403

TABLE V. Comparison of errors measured as standard error of laboratory (SEL) measurements and 3% mean reference value (MRV)
of the phase I and II calibration samples with standard errors of prediction (SEP) of the validation set (38 phase I and 85 phase II samples)
to show accuracy of predicting the test parameters by the best equations listed in Table IV.
SEL phase I SEL phase II
Parameter samples samples
Ammonia (%) 0.022 0.006
pH 0.116 0.083
Conductivity (ms/cm) 0.173 0.147
DM (%) 0.696 0.518
NDM (%) 0.101 0.051
Ash (%) 0.669 0.751
bration will be required to improve and maintain the ac-
curacy of the models.
There are several ways of assessing the precision of
the calibrations, for example, Williams 23 recommended
that SEP should not be more than 3% of the mean ref-
erence value (MRV) for an analyte, and alternatively,
Westerhaus 24 suggested that an SEP should be no greater
than twice the standard error of laboratory (SEL) mea-
surement. The SEL, MRV, and SEP values for phase I
and II samples of the validation set are listed in Table V.
The prediction errors for phase I and II samples were
calculated separately to show differences in the accuracy
of prediction instead of the mean SEP values listed in
Table IV. The SEL values for all parameters of phase I
samples, except for ash, were higher than phase II ma-
terials, due to substrate heterogeneity during the first
stage of production (Table V). The SEP for all parameters
except for ammonia were within the SEL limits recom-
mended by Westerhaus. 24 The SEL values for some of
the parameters were high and further improvement in
protocols will be required to reduce the measurement er-
rors. With the 3% MRV values of the key parameters, 23
all SEP values were outside the limits except for DM and
pH (Table V).
During substrate production, composters need to know
changes in ammonia, pH, DM, NDM, and ash to evaluate
the progress of microbial breakdown of straw, which will
determine an optimal duration and quality of the phase I
and II stages. 21 In addition, the interactions of these pa-
rameters along with microbial population, carbon, ash,
and minerals can influence the productivity of phase II
substrate19 and accurate measurement of the parameters
is important for assessing quality. Currently, intervention
of the production protocols cannot be implemented sat-
isfactorily, as the wet chemistry tests require a minimum
of 24 h to measure ammonia, NDM, and DM. With the
NIRS technique, provided that the temperature of the test
materials is near 20 8C, a number of samples can be
scanned and predicted within an hour and test reports can
be sent to the composters electronically. When required,
changes in the production protocol could be implemented
within a day by the yard managers, based on the pre-
dicted results and recommended intervention steps. 21
Warm (25–30 8C) or cold (7–15 8C) phase I or II samples
(2 kg sample size) will require a conditioning time of 1
h or more to optimize substrate temperature to near 20
8C before NIR assessment can be carried out and this will
minimize effects of temperature variation on the perfor-
mance of equations.
The results of this study have confirmed that NIRS
could be used to predict key parameters of substrates pro-
1404 Volume 59, Number 11, 2005
SEP phase I SEP phase II MRV phase I MRV phase II
samples samples samples samples
0.097 0.042 0.019 0.001
0.186 0.162 0.245 0.230
0.249 0.235 0.088 0.109
0.876 0.705 0.777 0.968
0.132 0.102 0.043 0.070
1.010 1.023 0.483 0.675
duced under different production systems, such as bunker
or windrow phase I. The equations reported by Resink
and van den Hurk7 may not be suitable for assessment of
commercial substrates prepared in the UK and Ireland
due to significant differences in the production proto-
cols.11
CONCLUSION
This study has confirmed that accurate calibrations for
assessing fresh phase I and II composts can be developed
from the visible and near-infrared spectra of the samples
and the SEPs of a majority of parameters are within the
acceptable error range, indicated by SEL values. Since
the equations have been developed and validated with
commercial samples obtained during winter and summer
months, the accuracy of the selected models is likely to
be maintained. However, the performance of a calibration
equation can drift, due to changes in instrument sensitiv-
ity, e.g., new lamp, standard operating protocol for ref-
erence analysis, and new samples different from the ex-
isting population caused by changes in the production
system. Therefore, equations should be regularly moni-
tored for performance and if necessary, re-calibrated with
the new samples. This could be carried out easily with
protocols available in the Win ISI III software. The cal-
ibration equations developed in this study may be able to
predict the test parameters of phase I and II substrates
prepared in other countries where synthetic composts are
produced using broadly similar recipes and production
protocols.
In addition to ammonia, pH, conductivity, DM, NDM,
and ash reported in this study, other parameters, such as
microbial population, carbon, C:N ratio, fiber fractions,
water soluble carbohydrate and polyphenols, and certain
minerals are important for monitoring substrate quality
during production.3,5,6,18,19,25 The changes in some of these
factors can reveal valuable information on the rate of sub-
strate breakdown by microorganisms and ultimately pro-
ductivity of phase II compost.
We are in the process of transferring the calibrations
to a new instrument. However, transferring such calibra-
tions to other instruments will require an environmentally
controlled room. If the test conditions, especially tem-
perature, are different from the original parameters, a bias
could be induced by this disparity to the calibrations. All
relevant factors will be taken into account during transfer
of the calibrations and this aspect along with additional
parameters, including fiber fractions and carbohydrate
and polyphenol content of phase I and II compost will
form a part of future reports.
ACKNOWLEDGMENTS
We wish to thank all support staff members involved in the experi-
mental trials and Dr. S. Watson for analysis of results from the compost
comparative trials. Financial support by the Department of Agriculture
and Rural Development for Project 0083 and assistance provided by
the major composters of Northern Ireland and Ireland are gratefully
acknowledged.
1. S. Burrows, J. Sci. Food Agric. 2, 395 (1959).
2. S. Burrows, J. Sci. Food Agric. 2, 403 (1959).
3. B. P. Eddy and L. Jacobs, Mushroom J. 38, 56 (1976).
4. J. W. Sinden and E. Hauser, Mushroom Sci. 2, 123 (1953).
5. R. C. Ross and P. J. Harris, Sci. Horticul. 19, 223 (1983).
6. J. P. G. Gerrits, H. C. Bels-Koning, and F. M. Muller, Mushroom
Sci. 6, 225 (1967).
7. J. W. Resink and M. Van den Hurk, in Science and Cultivation of
Edible Fungi, L. J. L. D. Van Griensven, Ed. (Balkema, Rotterdam,
2000), pp. 349–353.
8. H. S. S. Sharma and G. Lyons, in Mushroom Biology and Mush-
room Products, A. Broderick and T. Nair, Eds. (University of West
Sydney, Sydney, Australia, 1999), pp. 481–489.
9. H. S. S. Sharma, G. Lyons, and J. Chambers, Ann. Appl. Biol. 136,
59 (2000).
10. H. S. S. Sharma and M. Kilpatrick, Appl. Spectrosc. 54, 44 (2000).
11. J. P. G. Gerrits and L. J. L. D. Van Griensven, Mushroom J. 205,
21 (1990).
12. H. S. S. Sharma, M. Kilpatrick, and L. Burns, J. Near Infrared
Spectrosc. 8, 11 (2000).
13. H. S. S. Sharma, in Near Infrared Spectroscopy: Proceedings of
the 11th International Conference, A. M. C. Davis and A. Garrido-
Varo, Eds. (NIR Publications, Chichester, 2004), pp. 327–332.
14. H. S. S. Sharma, M. Kilpatrick, G. Lyons, J. Murray, S. Moore, L.
Cheung, K. Finnegan, S. Sturgeon, and R. Mellon, in Science and
Cultivation of Edible and Medicinal Fungi, P. Romaine, C. B. Keil,
D. L. Rinker, and D. J. Royse, Eds. (Pennsylvania State University,
Pennsylvania, 2004), pp. 229–239.
15. H. S. S. Sharma, in Science and Cultivation of Edible and Medic-
inal Fungi, P. Romaine, C. B. Keil, D. L. Rinker, and D. J. Royse,
Eds. (Pennsylvania State University, Pennsylvania, 2004), pp. 221–
228.
16. H. S. S. Sharma, M. Kilpatrick, and G. Lyons, Appl. Spectrosc. 59,
1054 (2005).
17. The Analysis of Agricultural Materials, Reference Book 427 (Min-
istry of Agriculture, Fisheries and Food, HMSO, London, 1986).
18. H. S. S. Sharma, in Science and Cultivation of Edible Fungi, M.
Maher, Ed. (Balkema, Rotterdam, 1991), pp. 169–179.
19. H. S. S. Sharma and M. Kilpatrick, Can. J. Microbiol. 46, 515
(1999).
20. B. G. Osbourne, T. Fearn, and P. H. Hindle, Practical NIR Spec-
troscopy with Applications in Food and Beverage Analysis (Long-
man Scientific Technical, Harlow, 1992).
21. H. S. S. Sharma, M. Kilpatrick, and G. Lyons, in Mushroom Bi-
ology and Mushroom Products, Q. Tan, J. Zhang, M. Chen, H. Cao,
and J. A. Buswell, Eds. (Acta Edulis Fungi, Shanghai, 2005), pp.
221–235.
22. D. H. Clark, H. F. Mayland, and R. C. Lamb, Agron. J. 79, 485
(1987).
23. P. C. Williams, in Near Infrared Technology in the Agricultural
and Food Industries, P. C. Williams and K. H. Norris, Eds. (Amer-
ican Association of Cereal Chemists, St. Paul, Minnesota, 1987),
pp. 143–167.
24. M. O. Westerhaus, in Near Infrared Reflectance Spectroscopy
(NIRS): Analysis of forage quality (Agriculture Handbook 643), G.
C. Marten, J. S. Shenk, and F. E. Barton II, Eds. (USDA-ARS,
Washington D.C., 1989), pp. 39–40.
25. P. Flegg, D. M. Spencer, and D. A. Wood, The Biology and Tech-
nology of the Cultivated Mushroom (John Wiley and Sons, Chich-
ester, England, 1985).
APPLIED SPECTROSCOPY 1405