Journal of Stored Products Research 69 (2016) 235e244

Contents lists available at ScienceDirect

Journal of Stored Products Research

journal homepage: www.elsevier.com/locate/jspr

Storage of hybrid rough rice e Consideration of microbial growth

kinetics and prediction models
Griffiths G. Atungulu*, Supriya Thote, Shantae Wilson
Department of Food Science, University of Arkansas Division of Agriculture, 2650 N Young Avenue, Fayetteville, AR 72704, USA

a r t i c l e i n f o a b s t r a c t

Article history: The objective of this study was to determine kinetics of mold growth on popularly grown hybrid long-
Received 30 August 2016 grain rough rice during storage at conditions that simulate delayed drying and prolonged storage. Three
Accepted 19 September 2016 long-grain hybrid rice cultivars, XL753 (2014) and CL XL745 (2014, 2015) and XL760 (2015) conditioned
Available online 28 September 2016
to four different moisture contents (MCs) (12.5%, 16.0%, 19.0%, and 21.0% wet basis) were stored in rough
rice form at temperatures ranging from 10
C to 40
C for a period of 16 weeks. The study was repeated
Keywords:
using rice from two consecutive crop seasons, 2014 and 2015. For all cultivars, a direct relationship
Hybrid rough rice
between mold counts and rice MC was observed e whereas more complex trends were observed for the
Storage
Microbial growth
effect of temperature and the duration of storage on mold growth. Kinetic models including Baranyi,
Prediction models Weibull, Gompertz, Richard and Buchanan were successfully modified and fitted using non-linear
regression and used to predict the mold counts (log10 CFU/g) for varying conditions
(correlations ¼ 0.65e0.76). The study concluded that long-grain hybrid rough rice could be stored at low
MC levels (17%) and moderate temperatures (27
C) for up to 6 weeks without any change in the mold
growth profile. However, storing rice at high MC (>17%) for more than 8 weeks, especially at higher
(>27
C) temperatures should be avoided due to the potential for high mold activity leading to loss of the
grain quality.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction storage ecosystem. Timely scheduling of postharvest operations by

Optimal milling yield and quality of rice is maintained at rice
harvest moisture contents (MCs) ranging between 19 and 21%, wet
basis. At such high harvest MCs, the grain is prone to heating which
arises from excessive respiration of the grain itself and associated
microorganisms and pests; this leads to rapid deterioration of the
grain quality (Harein and Meronuck, 1995; Atungulu et al., 2015).
Typically, the rate of proliferation of microorganisms and pests in
the grain storage ecosystem is dependent on the grain water ac-
tivity, grain temperature, and storage duration (Lacey and Magan,
1991; Mutters and Thompson, 2009). Therefore, timely and
proper drying of the grain to safe storage MC and temperature is
very important to minimize quality reduction (Skyrme et al., 1998;
Lee, 1999; Ranalli and Howell, 2002; Atungulu et al., 2015).
Normally, the rice harvesting “window” is relatively short. The
harvest period is also characterized by warm and humid conditions
that favor proliferation of microbes and pest within the grain
* Corresponding author.
E-mail address: atungulu@uark.edu (G.G. Atungulu).
rice producers is vital to minimize the grain quality. In the recent
past, annual rice yields per acre have been on the increase; this is
largely due to efforts by breeding programs to increase productivity
and meet food demand. Development in rice harvesting technology
has also increased the speed at which rice can be harvested; and
along with larger and faster grain carts, trucks, and trailers for
transporting rice from combines to driers, a much greater rice de-
livery rate to driers is typical within the short rice harvesting
“window”. Unfortunately, the rice drying infrastructure has not
grown at the same rate as that of delivery at commercial drying
facilities. In such instances, temporary “wet holding” (delayed
drying) of rice has become inevitable; this possess a lot of chal-
lenge, particularly for rice coming in at high initial loads of
microorganism and pest.
The goal of this research was to provide useful information that
could aid development of proper postharvest management prac-
tices for hybrid long grain (HLG) rice. Morphologically, HLG rice has
a pubescent characteristic and is very morphologically distinct
from conventional long-grain pureline and medium-grain rice. This
study hypothesized that the pubescence on the rice may predispose

http://dx.doi.org/10.1016/j.jspr.2016.09.003
0022-474X/© 2016 Elsevier Ltd. All rights reserved.
236 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244
it to harbor and support microbial and pest growth in a manner
different from non-pubescent rice (Atungulu et al., 2015); this
morphological distinctiveness necessitated, as a first step, an
investigation of the interactions among storage microorganisms,
storage pests and the grain within the storage ecosystem.
In the last decade, HLG rice has become very popular among
United States rice growers. The primary reasons behind this surge
in popularity include the potential for high rice yields, high resis-
tance to disease, and short growing seasons (Nalley et al., 2014). A
Significant amount of research effort has been dedicated to
studying insect infestations and control for rough rice (RR) storage.
Skyrme et al. (1998) and Ranalli and Howell (2002) characterized
molds and APCs on rice during storage, but largely for long- and
medium-grain varieties. By and large, the greatest challenge to rice
safety during storage a rises from the potential of toxigenic mold
contamination which under certain conditions may produce sub-
stances called mycotoxins. Some mycotoxins pose a significant
public health risk and lead to huge economic losses.
Few studies, if any have focused on identifying and enumerating
mold counts or determining aflatoxin growth kinetics for stored
HLG rice (Jayas et al., 1994; Wawrzyniak et al., 2013; Atungulu et al.,
2015). With the recent rise in the use of in-bin drying systems for
HLG rice, there is increased need to generate relevant data to
support development of models for predicting microbial kinetics
during the rice drying and storage; this is particularly important to
ensure that critical drying and storage regimes that ensure safety of
rice are maintained. An immediate application of such information
is in optimization of on-farm in bin drying of rice using recently-
introduced “cabling and sensing technology”. The new technol-
ogy uses the concept of equilibrium moisture content to monitor
grain MC and temperature and achieve automated fan control
during natural air in-bin drying and storage of the rice. The dura-
tion required for natural air in-bin drying, generally 2e4 weeks or
even longer, is directly affected by local weather conditions, which
sometime may not be conducive for complete drying, especially for
upper layers of rice (Atungulu et al., 2014). When drying is
incomplete (delayed), rice, especially that in the upper layers, re-
mains at high MCs for prolonged durations leading to excessive rice
respiration, microbial activity, dry matter loss, and hence reduction
in overall rice quality. At present, the allowable duration rice can
remain at high MC during on-farm, in-bin drying is under-
researched and is vaguely determined, in part based on dry mat-
ter loss guidelines from corn-related research — typically recom-
mended not to exceed 0.5% (Reed et al., 2007). What is not known is
the extent to which using the dry matter loss guideline, which does
not account for rice types, effectively prevents mold proliferation or
production of mycotoxin on rice during natural air in-bin drying
and storage. In the absence of such knowledge, on-farm rice drying
and storage with natural air will likely remain susceptible to
contamination with toxigenic fungi and their associated myco-
toxins, many of which are carcinogenic to humans.
Based on literature survey, Buchanan three-phase linear, Gom-
pertz, Richards, Weibull and Baranyi model for microbial growth
have been used to successfully predict mold growth kinetics (Lo
pez
et al., 2004). New data describing microbial growth kinetics for HLG
rice could be fitted to these existing models to determine suitability
for predicting microbial growth kinetics over a wide range of
storage conditions encountered in actual field scenarios. Such in-
formation will be vital in development of post-harvest processing
and storage guidelines for safe storage of rice, especially for the HLG
rice cultivars.
The specific objectives for this study were to determine (1) ki-
netics of mold growth on HLG rice at different MCs and storage
temperatures; (2) mathematical models that best describe the ki-
netics of the mold growth; and (3) safe storage conditions for HLG
rice by considering the activity of mold growth during storage.
2. Materials and methods
2.1. Samples
Freshly-harvested rice was procured from rice fields located in
Arkansas. Two HLG rice cultivars, CL XL745 from Running Lake
Farms near Pocahontas, AR and XL753 from Keiser, AR (University
of Arkansas, Research Station) were harvested in 2014 (Year 1,
Table 2) at approximately 22% MC wet basis (w.b.). All MCs for this
study are expressed on a w.b. unless otherwise stated. For the year
2015 (Year 2, Table 2.), HLG cultivars, CL XL745 and XL760 were
harvested from Running Lake Farms near Pocahontas, AR at
approximately 22% MC w.b. The rice samples were then cleaned to
remove chaff and foreign matter using a dockage tester (Model XT4,
Carter-Day, Minneapolis, MN) and then conditioned to four initial
MC levels (12.5%, 16%, 19% and 21%) by placing the rice on a tarp at
room temperature. The MC was monitored periodically during this
drying step by a moisture tester (AM 5200, Perten Instruments,
H€agersten, Sweden). After conditioning the rice to the set MC
levels, the rice samples were immediately stored in individual
quart-sized, glass containers (mason jars) to prevent significant
alterations of their initial MCs and then transported to five separate
temperature environments 10
C, 15
C, 20
C, 27
C and 40
C. The
chosen environments consisted of a combination of three Param-
eter Generation and Control (PG&C, Black Mountain, NC)) units,
including one PG&C unit maintaining the environment of a walk-in
equilibrium moisture content (EMC) chamber, one walk-in refrig-
erator and, one incubator (BINDER Inc., Bohemia, NY). The jars were
stored for a period of 16 weeks and sampled every two weeks
except after week 12, when the rice samples were stored for a
continuous period of four weeks. The same sampling protocol was
followed during 2015. A total of 56 jars (4 MC  2 cultivars  7
storage durations) were placed in each of the environmental units,
resulting in a total of 280 experimental units for each of the two
years. The experimental design is illustrated by Table 1.
2.2. Temperature monitoring
The temperature of the chosen storage environments was
monitored by placing two temperature sensors (HOBO™ Pro v2,
Onset Computer Corp., Bourne, MA) in each of the five environ-
mental units. These sensors recorded the temperature every 5 min
throughout the storage period.
2.3. Moisture content measurement
To determine the rice sample MC, the HLG RR stored in jars was
removed from the storage environments and placed at room tem-
perature conditions where they were allowed to equilibrate for at
least 24 h before removing samples for MC determination. The MC
of each sample was measured in duplicate by placing a 15 g sample
into a 130
C convective oven (Shellblue, Sheldon Mfg., Inc., Cor-
nelius, OR) for 24 h followed by cooling in a desiccator for at least
half an hour (Jindal and Siebenmorgen, 1987). The measured MC
values were within 1.5 percentage point of the set value. Table 2
provides the MC values obtained using the oven method.
2.4. Water activity measurement
The water activity for the 2015 cultivars (CL XL745 and XL760)
was measured using a water activity meter (AquaLab Pre, Decagon
Devices, Pullman, WA). The measurements were performed in
duplicate. Fig. 1 provides the linear regression for the measured
 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244 237

Table 1
Conceptual illustration of the experimental design for enumerating mold counts on hybrid long-grain rice cultivars. The experiment was set up as a full-factorial design with
Week 0 and MC ¼ 12.5% serving as the controls.

Year Cultivar Moisture content (% wet basis) Temperature (
C) Storage duration

1 (2014) CL XL745 12.5 10 Week 0

XL753 16 15 Week 2
2 (2015) CL XL745 19 20 Week 4
XL760 21 27 Week 6
40 Week 8
Week 10
Week 12
Week 16

Table 2 mold counts, 104 to 107 times dilutions were plated, respectively.
Difference between moisture content values set during the design stage and the For total APCs, 104 to 107 dilutions were plated, respectively.
actual, experimentally-measured oven moisture content (OMC) values wet basis
(w.b.).
Preliminary tests had been done to determine the appropriate
dilution for each kind of microbial population analysis. The inocu-
Year Cultivar Oven moisture content (% OMC w.b.) lated plates were stacked up to a maximum of 20 units and
Set value for moisture 12.5 16 19 21 incubated.
content (% MC w.b.) Aerobic Count Plates were incubated (VWR General Purpose
2014 CL XL 745 12.7 16.3 19.1 21.3 Incubator 1536, Sheldon Manufacturing Inc., Cornelius, OR) at 37
C
XL753 12.4 16.7 20.6 21.6 for 48 h before counting. Mold Count Plates were incubated (Thelco
2015 CL XL 745 12.4 16.3 19.6 21.1
Model 4, Precision Scientific Instruments, Inc., Chicago, IL) at 25
C
XL 760 12.2 16.4 20.2 21.7
for 120 h before counting. After incubation, the Colony Forming
The bold signifies the case where the set value and oven moisture contents were one Units (CFU) on each plate were counted. Aerobic bacteria colonies
percentage point different.
appeared red in color on Aerobic Count Plate. Yeast colonies
appeared blue-green or off-white in color and had non-diffusive
edges. Mold colony colors were blue, black, yellow, or green.
Mold colonies tended to be larger and more diffusive than yeast
colonies. The appropriate dilution factor, volume, and sample
weight were taken into account to obtain the total CFU/g of each
sample:

PCFU
Tcfu ¼ (1)
Dr

Fig. 1. Relationship between water activity (aw) and measured moisture content for
rough rice samples conditioned to set moisture contents ¼ 12.5%, 16%, 19% and 21% and
stored over a period of 16 weeks at temperatures ¼ 10
C, 15
C, 20
C, 27
C and 40
C
for cultivars (a) CL XL745 and (b) XL760 for the year 2015.
where, Tcfu is the total colony forming units per gram of RR (CFU/g),
PCFU are the colony forming units counted on each plate per gram of
rice (CFU/g), and Dr is the dilution factor (103 to 108 times).
For the studied rice samples, preliminary results showed that
yeast counts were very low with nearly none detected even with
1010 dilution. Therefore, yeast count was not reported in this
research. All samples were plated and analyzed for mold counts in
triplicate. For APC, only the 2015 rice samples were plated, for up to
8 weeks; after 8 weeks, the counts were too low to detect using the
previously described dilutions.

2.6. Data fitting for kinetic growth models

water activity and MC (R2 ¼ 0.965). Since the correlation between
water activity and MC is strong, trends and analysis are presented Five microbial growth models - Buchanan three-phase linear,
using MC. Gompertz, Richards, Weibull and Baranyi e were identified from
literature and modified to predict mold counts. The models were fit
to the natural log of experimental mold counts using the non-linear
2.5. Microbial analysis platform on JMP Pro 12 (SAS Institute Inc., Cary, N.C.).Where,
L0¼log10 of initial mold counts
The AOAC methods 990.12 (2005) and 997.02 (2005) for the 3M t¼time, number of weeks
Petrifilm Aerobic Count Plates and 3M Petrifilm Mold Count Plates tlag¼lag time in number of weeks
(3M Microbiology Product, Minneapolis, MN) were used to deter- r¼specific growth rate (per week)
mine the RR total (ground sample) microbial counts. Suspensions tmax¼time, in number of weeks to reach maximum log10
containing the microbes were prepared by masticating the rice microbial population
samples by using a masticator (Silver Panoramic, iUL, S.A., Barce- Lmax¼ maximum log10 mold population
lona, Spain) set at 240 s and 0.5 stroke/s, allowing the rice samples a¼dimensionless temperature adjustment parameter
to be pulverized into powder for total microbial load analysis. The k¼dimensionless curvature parameter
successive dilutions were made by mixing 1 mL of the original rmax¼maximum growth rate (per week)
mixture with 9 mL of phosphate-buffered dilution water. For total L1¼ simplifying term 1 for Baranyi equation
238 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244

L2¼simplifying term 2 for Baranyi equation
3. Results
3.1. Statistical analysis
Linear and non-linear regression, analysis of variance, Student's
t-test (least significant difference test), and the Tukey HSD tests
were performed with statistical software (JMP version 12.0.0, SAS
Institute Inc., Cary, N.C.) to determine significant differences within
and among sample means. Level of significance (a) was set at 5% for
comparing means of APC and mold counts. Table 4 gives the P
values and degrees of freedom (df) associated with the three factors
involved in the study tracking mold counts and APC found on the
stored RR samples.
3.2. Frequency distribution of microbial loads by cultivar and year
3.2.1. Mold counts
For Year 1, the overall mean value of mold counts for the CL
XL745 (2014) samples was M ¼ 6.01, SD ¼ 0.82 and for XL753
(2014) was M ¼ 6.16, SD ¼ 0.76 log10 CFU/g. Whereas, for Year 2,
overall mean of the mold counts (log10 CFU/g) for the CL XL745
(2015) samples was M ¼ 5.72, SD ¼ 0.77 and for XL760 (2015) was
M ¼ 5.94, SD ¼ 0.78. The mean values for the two years were
significantly different (p < 0.001, for two-tailed t-test, assuming
unequal variances). Within each year, the mold counts for the two
cultivars - CL XL745 and XL753 for 2014 - and - CL XL745 and XL760
for 2015 - were also significantly different from each other
(p ¼ 0.0034 (2014) and p < 0.001 (2015) for two-tailed t-test,
assuming unequal variances). Therefore, further analyses are
separated by cultivar type and year.
The medians of the samples for year 1 were slightly lower than
the corresponding means, 5.87 and 5.95 log10 CFU/g for CL XL745
(2014) and XL753 2014), respectively. The same trend was observed
for the median values for year 2 cultivars. The median for CL XL745
(2015) was 5.60 log10 CFU/g and for XL760 (2015) the median value
was 5.70 log10 CFU/g. Fig. 2(a), (b), (c) and (d) provide histograms
displaying the frequency distribution of mold counts (log10 CFU/g)
for CL XL745 (2014), XL753 (2014), CL XL745 (2015) and XL760
(2015) respectively.
From Fig. 2(a) and (b), it was noted that nearly 80% of the CL
XL745 and XL753 samples were between 5 and 7 log10 CFU/g for
year 1 (2014) with the mold counts ranging from 4.0 to 8.5
log10 CFU/g. The range for sets of samples for 2014 was over 4
log10 CFU/g, which means that there were 104 times more colony
forming units in the sample with the most mold growth as
compared to the sample with the least mold growth. Fig. 2(c) and
(d) plots the range of mold counts for year 2015 (year 2). The range
for year 2 samples was lower than the range for the year 1 samples -
from 4.0 to 7.5 log10 CFU/g, with about half of the samples having
mold counts between 5.0 and 6.0 log10 CFU/g.
3.2.2. Aerobic plate counts
The APC analysis was performed solely for the year 2 (2015)
samples. The log10 of the overall mean APC for cultivar CL XL745
(2015) was M ¼ 7.27, SD ¼ 0.76 CFU/g; for cultivar XL760 (2015) the
log10 of the overall mean was M ¼ 7.47, SD ¼ 0.63. The overall mean
APC for XL760 was found to be significantly higher (p < 0.0001, for
two-tailed t-test, assuming unequal variances) than the overall
mean APC for CL XL745. The median values of APC for both cultivars
were slightly higher than the corresponding means, 7.54 and 7.61
log10 CFU/g for CL XL745 (2015) and XL760 (2015), respectively.
Fig. 3(a) and (b) depict the frequency distribution of APC (log10 CFU/
g) for cultivars CL XL745 (2015) and XL760 (2015) respectively.
From Fig. 3(a) and (b), it was noted that APC distribution even with
the log10 transform, is negatively skewed for both cultivars. Nearly
80% of the CL XL745 and 85% XL760 samples have APC values over 7
log10 CFU/g. These APC values provide a measure for the prevalence
and trends of bacteria on stored RR.
3.3. Factors affecting microbial growth kinetics
3.3.1. Mold counts
3.3.1.1. Initial mold counts. The mean of initial mold counts for CL
XL745 (2014) was M ¼ 5.97, SD ¼ 0.13 log10 CFU/g, whereas the
mean of initial mold counts for the same cultivar, CL XL745 (2015),
for the second year was M ¼ 5.68, SD ¼ 0.07 log10 CFU/g. Cultivar

Table 3
Modified Buchanan Three-Phase Linear, Modified Gompertz, Modified Richards, Modified Weibull and Modified Baranyi Equations used to fit experimental mold counts
(natural log, CFU/g).

Mathematical model Equation

Buchanan Three-Phase Linear L0 t  tlag L0 þ rt tlag < t < tmax Lmax t  tmax
Gompertz aLmax ðExpð  kExpðrtÞÞ
Richards
aLmax
1
1þk*fExp½r*ðttmax Þg k

Weibull a½Lmax  ðLmax  L0 ÞExpððrtÞk Þ

!
Baranyi rmax *ðttlag Þ rmax *tlag
þe Þ
L0 þ rmax t þ L1  L2 L1 ¼ lnðermax *t  ermax *ðtþtlag Þ þ ermax *tlag ÞL2 ¼ ln 1 þ e e Lmax L0

Table 4
Summary of the P values from analysis of variance for the effect of storage duration, temperature and moisture content by cultivar mold counts (2014, 2015) and aerobic plate
counts (2015).

Factor Prob > F Prob > F

df MOLD counts df APC counts

2014 2015 2015

CL XL 745 XL753 CL XL 745 XL 760 CL XL 745 XL 760

Moisture Content (MC% w.b.) 3 <0.0001 <0.0001 <0.0001 <0.0001 3 0.0920 <0.2187
Storage duration (number of weeks) 7 <0.0001 <0.0001 <0.0001 <0.0001 4 <0.0001 <0.0001
Temperature 4 0.0036 0.2822 <0.0001 <0.0001 4 <0.0001 <0.0001
 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244
Fig. 2. Frequency distribution for mold counts (log10 CFU/g) detected over a storage
period of 16 weeks for rough rice moisture contents ¼ 12.5%, 16%, 19% and 21% and
temperatures ¼ 10
C, 15
C, 20
C, 27
C and 40
C for cultivars (a) CL XL745, (b)
XL753, (c) CL XL745 and (d) XL760 during the years 2014 and 2015.
XL753 (2015) had the highest mean of initial mold load - M ¼ 6.0,
SD ¼ 0.25 log10 CFU/g; the XL760 (2015) cultivar had the lowest
mean for initial mold counts, M ¼ 5.61, SD ¼ 0.11 log10 CFU/g. The
initial mold counts on HLG rice for Year 1 (2014) were significantly
higher (p < 0.001) than the counts obtained for Year 2 (2015). These
initial microbial loads affected the mold growth profiles and trends.
3.3.1.2. Moisture content. The difference between the mold counts
detected for the different MC levels for each of the two rice cultivars
studied over the two years was significant (Table 5). The standard
deviations for these means were relatively high because the data
encompasses the range of storage durations and temperature
conditions for the entire experiment. For the two cultivars studied
in each of the two years, the mold counts for the two lower MC
239
Fig. 3. Frequency distribution for aerobic plate counts (log10 CFU/g) detected over a
storage period of 8 weeks for rough rice stored at moisture contents ¼ 12.5%, 16%, 19%
and 21% and temperatures ¼ 10
C, 15
C, 20
C, 27
C and 40
C for cultivars (a) CL
XL745 and (b) XL760 in the year 2015.
levels are significantly lower than the two higher MC levels
(p < 0.0001 for both). Additionally, for cultivar CL XL745 (2014), the
mold counts for MC ¼ 21.3% is significantly higher than MC ¼ 19.1%.
Also, the mold counts for MC ¼ 21.6% were significantly higher than
MC ¼ 20.6% for cultivar XL753 (2014). Conversely, for cultivar CL
XL745 (2015), no difference was detected between the mold counts
for MC ¼ 19.7% and MC ¼ 21.1%, and between the mold counts for
MC ¼ 21.7% and MC ¼ 20.2% for cultivar XL760 (2015). For all cul-
tivars, a direct relationship between mold counts and MC was
observed e as the MC level increased, the mold counts increased.
3.3.1.3. Storage duration. The effect of storage duration on mold
counts was studied after separating the MC levels into two levels e
LOW for MC  17% and HIGH for MC>17%. Fig. 4 gives twenty
separate line graphs for each cultivar studied, separated by year,
showing the distribution of mold counts (log10 CFU/g) (y-axis) over
the storage duration of 16 weeks (x-axis). For the LOW MC level, the
mold counts did not fluctuate significantly e staying either close to
the initial mold counts or decreasing slightly as storage duration
increased. For the HIGH MC level, the growth curves depicting the
effect of storage duration on mold counts (log10 CFU/g) follow a
similar pattern observed for microbial growth curves. The curves
can be divided into three major categories e the lag phase, the
exponential growth phase and the decay phase (Fig. 5). During the
lag phase, the mold counts remain almost the same as the initial
mold counts. The exponential growth phase is characterized by a
near linear increase in mold counts (log10 CFU/g). Finally, the decay
phase is the stage when the mold counts begin to decline. The start
and end-points of each of these three phases varied depending on
cultivar-year combination, MC level and temperature.
3.3.1.4. Temperature. The effect of temperature on mold counts was
highly dependent on the storage MC and the initial mold loads on
the HLG RR. As listed in Table 5, the low MC samples (MC  17%), no
significant differences were observed between the mold counts for
T ¼ 10
C, 15
C and 20
C for all four cultivar-year combinations; on
the contrary, for all four cultivar-year combinations, the mold
counts for the 40
C were significantly lower than the other four
240 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244

Table 5
Mean values of mold counts for rough rice stored at different moisture content levels for the cultivars studied for each year. Mean values differing by a lower case letter are
significantly different at a ¼ 0.05 using the Tukey HSD test for means at differing moisture content levels (% wet basis, w.b.); Mean values differing by an upper case letter,
across a row are significantly different at a ¼ 0.05 using the Tukey HSD test for means at different cultivar-year combinations.

Set value for Moisture content (% MC w.b.) 2014 2015

CL XL 745 XL753 CL XL 745 XL 760

21 6.78a, A 6.68a, A 6.19a, B 6.72a, A

19 6.08b, B 6.44b, A 6.04a, B 6.31b, A
16 5.62c, B 5.82c, A 5.30b, C 5.36c, C
12.5 5.55d, A 5.68d, A 5.34b, B 5.37c, B

Fig. 4. Mold counts (log10 CFU/g) detected over a storage period of 16 weeks for rough rice moisture contents ¼ LOW (17%) and HIGH (>17%) and temperatures ¼ 10
C, 15
C,
20
C, 27
C and 40
C for cultivars CL XL745 (2014), CL XL745 (2015), XL753 and XL760 (2014).

cultivars are provided in Table 6.

3.3.2. Aerobic plate counts

3.3.2.1. Initial APC. The initial aerobic bacteria load in terms of APC
for the two rice cultivars studied were significantly different
(p < 0.001, for two-tailed t-test, assuming unequal variances). The
mean of initial APC for CL XL745 (2015) (M ¼ 7.82, SD ¼ 0.07
log10 CFU/g) was significantly higher than the mean of initial XL760
(2015) (7.62, SD ¼ 0.14 log10 CFU/g). The initial APC values did not
affect overall trends for the range of storage durations, tempera-
tures and MC levels studied.

Fig. 5. Typical microbial growth curve (McKellar and Lu, 2003).
temperatures. For the high MC samples (MC>17%), cultivar CL
XL745, both years, no significant differences were observed for
mold counts at different temperatures e except for the CL XL745
(2015) 40
C samples, where the mold counts were significantly
lower than the mold counts for the other temperatures. For the
remaining two lots, XL753 (2014) and XL760 (2015), the trends
observed were more complex; the average mold counts for the two
3.3.2.2. Temperature and moisture content. The effect of tempera-
ture on APC was highly dependent on the storage MC and storage
duration of the HLG RR, however, overall as storage temperature
increased APC decreased. As listed in Table 7, for the 12.5% MC
samples for both cultivars, no significant differences were observed
between the APC values for T ¼ 10
C, 15
C, 20
C and 27
C but the
APC values for the 40
C were significantly lower than the other
four temperatures. For the high MC samples (19 and 21%), the APC
generally decreased as the temperature increased. Overall, no sig-
nificant differences were detected between APC for the four
 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244 241

Table 6
Mean values of log10 CFU/g mold counts on rough rice stored at different temperatures for the rice cultivars studied in 2014 and 2015..Mean values differing by a lower case
letter are significantly different at a ¼ 0.05 using the Tukey HSD test for means at differing temperature levels; Mean values differing by an upper case letter, across a row are
significantly different at a ¼ 0.05 using the Tukey HSD test for means at different cultivar-year combinations.

Moisture content level Temperature (
C) CL XL 745 (2014) CL XL 745 (2015) XL753 (2014) XL 760 (2015)
a, A a, B a, A
Low 10 5.68 5.51 5.93 5.56a, B
15 5.78a, A 5.50a, B 5.88a, A 5.57a, B
20 5.73a, B 5.52a, BC 5.82a, A 5.51a, C
27 5.43b, A 5.23b, B 5.67ab, A 5.20b, B
40 5.29b, A 4.84c, B 5.51b, A 4.87c, B
High 10 6.13a, A 6.29a, A 6.09b, A 6.36bc, A
15 6.52a, A 6.20a, A 6.37ab, A 6.57ab, A
20 6.49a, AB 6.22a, A 6.41ab, AB 6.67ab, B
27 6.44a, AB 6.21a, A 6.70a, A 6.77a, A
40 6.43a, A 5.65b, B 6.64a, A 6.19c, A

different MC levels for both cultivars. phases: the lag phase, exponential growth phase and stationary
phase as illustrated in Fig. 5 (McKellar and Lu, 2003). The Gompertz
equation is the most frequently used sigmoidal function to describe
3.3.2.3. Storage duration. Ten distinct line graphs in Fig. 6 depict
microbial growth (McKellar and Lu, 2003). The other three models
the relationship between APC (y-axis) and storage duration (x-axis)
e Richards, Weibull and Baranyi e are also commonly used to
for each cultivar studied. As seen in Fig. 6, APC was highly depen-
describe microbial kinetics in literature (Lo
pez et al., 2004).
dent on temperature, storage duration and MC and the three-way
interaction between these three terms was significant
(p < 0.001). For the lower two temperatures (10
C and 15
C), the
APC did not fluctuate significantly e generally staying either close 3.5. Model comparison
to the initial APC for the eight-week storage duration. For the two
high storage temperatures (27
C and 40
C), overall, APC decreased Fig. 7 shows a plot comparing the fit of all five models for
as the storage duration decreased; the trend varied considerably cultivar CL XL745 (2015) at MC ¼ 21% for 20
C and 27
C to illus-
depending on MC. For the storage temperature 20
C, the trend trate the relative abilities of the models to predict experimentally
observed over storage varied depending on MC level. measured mold counts. For this particular combination of MC,
temperature and cultivar, the three phases e the lag phase, the
exponential growth phase and the decay or death phase can be
3.4. Kinetic models for mold counts
observed clearly. A similar trend was observed for the other three

pez et al. (2004) compared the statistical robustness of several cultivar-year combinations (CL XL745 (2014), CL XL753 and XL760)
Lo
for the two high MC set points e 19% and 21%, although some
previously published mathematical models to describe microbial
combinations lacked the lag phase or the death or decay phase
growth kinetics. The authors included 21 growth curves for
(Fig. 4). The mold counts for the lower MC levels, 12.5% and 16%,
different fungal and bacterial species but expressed the growth in
either remained relatively constant or declined over the storage
terms of optical density units. The study recommended five models
duration; the prediction models simulated this trend well.
e Buchanan three-phase linear, Gompertz, Richards, Weibull and
First, the fitted Baranyi, fitted Weibull, fitted Gompertz, fitted
Baranyi. These models were modified to include a dimensionless
Richards and fitted Buchanan models were estimated for each
parameter accounting for the change in growth kinetics depending
storage temperature, hence generating 5 sets of parameters (a, rmax,
on storage temperature. For the purpose of this study, the models
tlag for each storage temperature). The parameter “a” was included
were adopted to predict mold counts (Table 3).
in every model as a dimensionless parameter to account for the
The first model, Buchanan three-phase linear model was
change in trend behavior due to each temperature. The effect of MC
developed by Buchanan et al. (1997) and describes three growth
and cultivar-year combination is accounted for in the initial and
maximum mold count parameters L0 and Lmax used in the modified
Table 7 equations because both MC and mold counts are directly correlated.
Mean values of log10 CFU/g aerobic plate counts (APC) on rough rice at for different The parameter rmax represents the maximum rate of mold growth
storage temperatures for the cultivars studied in 2015. Mean values differing by a
in terms of CFU/g per week and lastly, tlag is the parameter that
lower case letter are significantly different at a ¼ 0.05 using the Tukey HSD test for
means at differing temperature levels; Mean values differing by an upper case letter,
defines the storage duration, in weeks, the mold counts stay con-
across a row are significantly different at a ¼ 0.05 using the Tukey HSD test for at stant and do not fluctuate.
differing moisture content levels (% wet basis, w.b.). Next, the Weibull, Gompertz, Baranyi, Richards and Buchanan
Temperature (
C) 12.5 16 19 21 were fitted over the entire temperature range and therefore, only
generated one set parameters (a, rmax, tlag over the entire temper-
CL XL745 (2015)
10 7.62a, B 7.73ab, B 7.74a, A 7.69a, A
ature range). Therefore, for the second simulation, the parameter
15 7.57a, B 7.38ab,B 7.55a, A 7.48ab, A “a” did not change in value for the different temperatures. Except
20 7.66a, B 7.93a, B 7.25ab, A 7.29ab, A for Buchanan and Richards, all equations had R2 higher than 0.6 for
27 7.62a, B 7.23b, B 6.80b, A 6.92bc, A temperatures 15
C, 20
C and 27
C. Most models performed worse
40 6.46b, C 6.21c, BC 6.62b, AB 6.65c, A
for the two extreme temperatures 10
C and 40
C. From the R2
XL760 (2015)
10 7.68a, C 7.83a, C 7.85a, B 7.80a, A values in Table 8, the fitted Gompertz model performed best at
15 7.81a, C 7.78a, C 7.67a, B 7.68a, A 20
C and 27
C. However, over the entire temperature range, the
20 7.76a, B 7.72a, B 7.56ab, A 7.04b, A Baranyi model gives the best overall predictions followed by the
27 7.61a, C 7.54a, C 7.26bc, B 6.99b, A fitted Weibull model (Table 9). Table 9 gives the parameter esti-
40 6.81b, C 6.50b, C 7.10c, B 7.30ab, A
mates for the Modified Baranyi and Weibull equations.
242 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244

Fig. 6. Aerobic Plate Counts (log10 CFU/g) detected for rough rice during a storage over a period of 8 weeks for moisture contents ¼ 12.5, 16, 19 and 21% and temperatures ¼ 10
C,
15
C, 20
C, 27
C and 40
C for cultivars XL753 (2015) and XL760 (2015).

Fig. 7. Comparison of the model fitted to describe the kinetics of mold growth - fitted Richard, Weibull, Baranyi, Gompertz and Buchanan e with the experimental mold counts
(natural log, CFU/g) for rough rice (cultivar CL XL745 (2015)) at moisture content ¼ 21% for temperatures ¼ 20
C and 27
C stored for 16 weeks.

4. Discussion Wawrzyniak et al., 2013). Therefore, the challenges of “wet hold-

ing” experienced with other rice cultivars remain relevant to HLG
The two year study (2014 and 2015) demonstrated that HLG rice, rice producers. Atungulu et al. (2015) studied and compared the
despite being bred for disease and pest resistance, still had elevated relative prevalence of microorganisms on freshly-harvested HLG,
levels of both mold and APCs at harvest (Jayas et al., 1994; long-grain pureline, and medium-grain rice types. The authors
concluded from their two-year study (2013 and 2014) that HLG rice
generally had significantly lower APC and mold levels compared to
Table 8
R2 values for Modified Buchanan Three-Phase Linear, Modified Gompertz, Modified
Richards, Modified Weibull and Modified Baranyi Equations used to fit experimental Table 9
mold counts (natural log, CFU/g). Parameter estimates for Modified Baranyi and Modified Weibull Equations used to
fit experimental mold counts (natural log, CFU/g).
Predictor R2 values
Temperature (
C) a rmax tlag
Temperature (
C)
Baranyi 10 0.91 0.42 3.45
10 15 20 27 40
15 0.89 0.58 2.29
Fitted Baranyi 0.71 0.69 0.71 0.74 0.68 20 0.89 0.66 1.8
Fitted Weibull 0.69 0.71 0.72 0.73 0.64 27 0.94 1.73 1.2
Fitted Gompertz 0.48 0.65 0.73 0.76 0.68 40 0.84 1.07 2.33
Weibull 0.57 0.65 0.7 0.7 0.57
Temperature (
C) a r k
Gompertz 0.49 0.62 0.7 0.76 0.63
Baranyi 0.53 0.65 0.69 0.7 0.62 Weibull 10 0.97 0.1 1.13
Fitted Buchanan 0.54 0.83 0.71 0.76 0.35 15 0.97 0.22 0.27
Fitted Richards RCH 0.44 0.62 0.71 0.74 0.64 20 0.96 0.68 0.33
Richards RCH 0.44 0.6 0.69 0.74 0.6 27 0.93 0.62 3.06
Buchanan 0.44 0.36 0.41 0.28 0.29 40 0.9 0.42 1.46
 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244
long-grain pureline and medium-grain rice. In their 2013 study,
HLG rice had the lowest APC levels (M ¼ 8.10, SD ¼ 0.09) (M and SD
symbolize mean and standard deviation, respectively); and in 2014,
HLG rice had the lowest mold count levels (M ¼ 5.97, SD ¼ 0.27)
(Atungulu et al., 2015). The foregoing authors also noted that rice
growing geographic locations played a role on the level of microbial
contamination of the rice at harvest, however, their observed
trends varied from year to year. Based on our study, it may be
recommended that long-grain pureline and medium-grain rice,
harvested at optimum MC and grown in the same field be accorded
faster attention to drying compared to HLG rice. Otherwise, the
long-grain pureline and medium-grain rice cultivars might be
subjected to more rapid dry matter loss, mold related staining,
problems of mycotoxin development and reductions in head rice
yields. In general, the levels of mold observed on HLG rice at harvest
were comparable to those reported in literature (Atungulu et al.,
2015).
Our study recommended that HLG rice could be stored at low
MC levels (17%) and moderate temperatures (27
C) for up to 6
weeks without any change in the mold growth profile. However,
various studies in literature have demonstrated that it is necessary
to evaluate postharvest rice handling as a system in order to pro-
vide suitable storage recommendations (Meullenet et al., 1999;
Daniels et al., 1996, 1998). Therefore, it is also important to look
at the implication of the storage conditions (duration, grain MC and
temperature) in the light of other quality characteristics as well. The
effects of temperature and heating duration on medium-grain
(Bengal variety) and long-grain (Cypress variety) rough rice at
high (~21%) and low (~18%) harvest MC levels were investigated by
Dillahunty et al. (2001); results showed that as exposure duration
and temperature increased, the occurrence of yellow rice also
increased. Daniels et al. (1996, 1998) and Meullenet et al. (1999)
designed experiments to simulate extended delays in drying, as
might occur during the peak of the harvest season. The authors
(Daniels et al., 1998) showed that head-rice yield was not affected
by wet-holding conditions (20.5% MC for 86 h at 20
C) for long-
grain rice. But, for samples dried at low temperature, delayed-
drying condition, resulted in significantly lower water absorption
and volume expansion compared to the samples that were dried
immediately after harvest (Daniels et al., 1996). From a rice sensory
standpoint, the same holding conditions described by Daniels et al.
(1998) resulted in slight but significant effects of wet holding on the
sensory quality of long-grain Cypress rice (Meullenet et al., 1999). In
particular, wet holding significantly affected both the clumpiness
and hardness of cooked rice; samples for which drying was delayed
yielded a significantly clumpier and less firm cooked rice. These,
results emphasized the necessity of evaluating postharvest
handling as a system. Each postharvest operation may have po-
tential to influence the functionality and quality of the final prod-
uct. In order to provide suitable storage recommendations for the
HLG rice, it is suggested that further investigations to document
corresponding milled rice yield, sensory attributes, changes in rice
functional characteristics and kinetics of mycotoxin formation on
rice be conducted. The forgoing are subjects of our current
investigations.
5. Conclusions
The goal of this study was to simulate conditions of delayed
drying and prolonged storage of RR and to characterize mold
growth on HLG RR. The study characterized the growth of mold on
HLG rice cultivars CL XL745, XL753 and XL760 over a storage period
of 16 weeks for RR at four MC levels (12.5%, 16%, 19% and 21%) and
five temperatures (10
C, 15
C, 20
C, 27
C and 40
C). The findings
from the study provide baseline information which is very helpful
243
in understanding the activity of microbes on HLG rice and modeling
kinetics of microbial growth during rice storage:
The initial mold counts for Year 1 (2014) HLG rice cultivars were
significantly higher than the initial mold counts for HLG rice
cultivars obtained for Year 2 (2015).
The initial microbial load detected on the HLG RR affected the
mold growth profiles and trends significantly.
The overall mean value of the log10 CFU/g for the CL XL745
(2014) samples was M ¼ 6.0, SD ¼ 0.87 and for XL753 (2014) was
M ¼ 6.15, SD ¼ 0.78. Whereas, overall mean of mold counts
(log10 CFU/g) for the CL XL745 (2015) samples was M ¼ 5.7,
SD ¼ 0.77 and for XL760 (2015) was M ¼ 5.9, SD ¼ 0.80.
For all cultivars, a direct relationship between mold counts and
MC was observed e as the MC level increased, the mold counts
increased.
More complex trends were observed for the effect of tempera-
ture and the duration of storage on mold growth.
Microbial growth equations: the Fitted Baranyi, fitted Weibull,
fitted Gompertz, fitted Richards and Fitted Buchanan models
were successfully modified to estimate kinetics of mold growth.
The Baranyi and Weibull equations performed the best
(R2 > 0.64) over the temperature range of 10
Ce40
C.
The information may be useful to guide decisions on drying and
storage conditions, especially of HLG rice cultivars, to avoid
mold growth leading to mycotoxin contamination.
In summary, this study finds that HLG RR may be stored at low
MC levels (<17% w.b.) and moderate temperatures (27<
C) for up to
six weeks without any change in the mold growth profile. However,
storing RR at high MC (>17% w.b.) for more than eight weeks,
especially at higher temperatures (27>
C) should be avoided.
References
Atungulu, G.G., Zhong, H., Siebenmorgen, T.J., 2014. Microbial Population on
Freshly-harvested Rice and Factors Affecting Prevalence. ASABE Paper No.
141914312. ASABE, St. Joseph, Mich.
Atungulu, G.G., Zhong, H., Thote, S., Okeyo, A., Couch, A., 2015. Microbial prevalence
on freshly-harvested long-grain pureline, hybrid, and Medium-grain rice cul-
tivars. Appl. Eng. Agric. 31 (6), 949e956.
Buchanan, R., Whiting, R., Damert, W., 1997. When is simple good enough: a
comparison of the gompertz, baranyi, and three-phase linear models for fitting
bacterial growth curves. Food Microbiol. 14 (4), 313e326.
Daniels, M.J., Marks, B.P., Meullenet, J.-F., Siebenmorgen, T.J., 1996. In: Norman, R.
(Ed.), Effects of Rice Storage History on The End-Use Quality Of Long-Grain Rice.
B.R.Wells Arkansas Rice Research Studies, Fayetteville, AR, pp. 152e157. Univ.
Ark. Agric. Exp. Stn.
Daniels, M.J., Marks, B.P., Siebenmorgen, T.J., McKnew, R.W., Meullenet, J.-F., 1998.
Effects of long-grain rough rice storage history on end-use quality. J. Food Sci.
63, 832e835.
Dillahunty, A.L., Siebenmorgen, T.J., Mauromoustakos, A., 2001. Effect of tempera-
ture, exposure duration, and moisture content on the yellowing of rice. Cereal
Chem. 78, 559e563.
Harein, P., Meronuck, R., 1995. Stored grain losses due to insects and molds and the
importance of proper grain management. Stored Product. Manag. 29e31.
Jayas, D.S., White, N.D., Muir, W.E., 1994. Stored-grain Ecosystems. CRC Press.
Jindal, V., Siebenmorgen, T., 1987. Effects of oven drying temperature and drying
time on rough rice moisture content determination. Trans. ASAE 30 (4),
1185e1192.
Lacey, J., Magan, N., 1991. Fungi in Cereal Grains: Their Occurrence and Water and
Temperature Relationships. Developments in Food Science, Netherlands.
Lee, S., 1999. Low-temperature Damp Corn Storage with and without Chemical
Preservatives (Masters).
Lo
pez, S., Prieto, M., Dijkstra, J., Dhanoa, M.S., France, J., 2004. Statistical evaluation
of mathematical models for microbial growth. Int. J. Food Microbiol. 96 (3),
289e300.
McKellar, R.C., Lu, X., 2003. Modeling Microbial Responses in Food. CRC Press.
Mutters, R.G., Thompson, J.F., 2009. Rice quality Handbook. UCANR Publications.
Meullenet, J.-F., Marks, B.P., Griffin, V.K., Daniels, M.J., 1999. Effects of rough rice
drying and storage conditions on sensory profiles of cooked rice. Cereal Chem.
76, 483e486.
Nalley, L.L., Dixon, B.L., Brye, K., Rogers, C.W., Myrteza, H., Norman, R.J., 2014.
Estimating cultivar effects on water usage and greenhouse gas emissions in rice
244 G.G. Atungulu et al. / Journal of Stored Products Research 69 (2016) 235e244
production. Agron. J. 106 (6), 1981e1992.
Official Methods of Analysis of AOAC INTERNATIONAL, 2005. Official Method
997.02. Yeast and Mold Counts in Foods, 18th Ed. AOAC INTERNATIONAL, Gai-
thersburg, MD, USA.
Official Methods of Analysis of AOAC International, 2005. Official Method 990.12,
Aerobic Plate Count in Foods, 18th Ed. AOAC International, Gaithersburg, MD,
USA.
Ranalli, R., Howell, T., 2002. Effects of storage conditions on the total aerobic and
Yeast/Mold bacterial counts of rough rice during On-farm storage. J. Food Sci. 67
(2), 807e810.
Reed, C., Doyungan, S., Ioerger, B., Getchell, A., 2007. Response of storage molds to
different initial moisture contents of maize (corn) stored at 25 C, and effect on
respiration rate and nutrient composition. J. Stored Prod. Res. 43 (4), 443e458.
Skyrme, D.S., Marks, B.P., Johnson, M.G., Siebenmorgen, T.J., 1998. Distribution of
total aerobic and coliform bacterial counts among rice kernel components.
J. Food Sci. 63 (1), 154e215.
Wawrzyniak, J., Ryniecki, A., Gawrysiak-Witulska, M., 2013. Kinetics of mould
growth in the stored barley ecosystem contaminated with aspergillus west-
erdijkiae, penicillium viridicatum and fusarium poae at 23e30
C. J. Sci. Food
Agric. 93 (4), 895e901.