Aquaculture 519 (2020) 734759

Contents lists available at ScienceDirect

Aquaculture
journal homepage: www.elsevier.com/locate/aquaculture

Optimizing feed automation: improving timer-feeders and on demand T

systems in semi-intensive pond culture of shrimp Litopenaeus vannamei
⁎
João Reis , Romi Novriadi, Anneleen Swanepoel, Guo Jingping, Melanie Rhodes, D. Allen Davis
School of Fisheries, Aquaculture and Aquatic Sciences, Auburn University, AL 36849,

A B S T R A C T

The continued success of shrimp farming will rely on improved feed management and reductions in labor costs. Shrimp are omnivorous, eating many small meals
with limited stomach capacity for food storage. Hence, increased performance may be obtained by spreading feed through multiple meals. Initial work has de-
monstrated that moving from two feeding per day into multiple feeding systems increases growth rate and production. Further advances have been made with on-
demand (satiation) feeding systems. The goal of this work was to continue the development of standard feeding protocol's (SFP) for automatic feeding systems to
maximize growth rates in semi-intensive pond production of shrimp, Litopenaeus vannamei. For this work a 13-week pond production trial was performed in 16, 0.1 ha
outdoors ponds, stocked at a 26 shrimp/m2, and fed 1.5-mm 40% crude protein for the first four weeks, and 2.4-mm protein soy optimized feed (35% crude protein)
for last nine weeks, both produced by Zeigler Inc. Four treatments including: three fixed feeding treatments of 130, 145 and 160% of a SFP (SFP + 30%, SFP + 45%,
SFP + 60%, respectively) were offered using automatic timer-feeders, and a fourth treatment utilized on-demand AQ1 acoustic feeding system. No statistical
differences were found between treatments for survival (ranging 75.2–81.4%) and FCR (ranging 0.96–1.11). In general, increased feed inputs resulted in higher
production. The best response was with the AQ1 system which adjusted feed inputs in real time and ended up offered higher feed inputs resulting in larger shrimp and
yields. Based on results of this work and previous trials, standard feeding protocols for automated systems can be developed but to date, automated feedback systems
which operate in real time out perform the standardized practices.

1. Introduction shrimp feeds that were previously leeched for over 0.5 h before feeding.
This confirms the hypothesis that the longer feed is in the water the
Shrimp are one of the most popular seafoods. In aquaculture, lower the nutritional value hence indicating small quickly consumed
Litopenaeus vannamei is the preferred shrimp species due to its culture meals are preferential. Multiple small and quickly consumed feedings
characteristics and consumer acceptance. The continued success of may improve nutrient delivery through reduced nutrient leaching re-
shrimp farming will rely on intensification, improved feed management sulting in improved growth and waste management. Nevertheless, of-
and reductions in labor costs. The cost of the feed is one of the most fering multiple meals can be very labour intensive and economically
important variable costs, source of nutrients and consequently biolo- impracticable in regions such as the Americas where labour cost is high
gical waste in shrimp production (Tacon and Forster, 2003). Commer- in comparison to South East Asia which tends to use more feedings per
cially available shrimp feeds are generally adequate (Quintero and Roy, day (Davis et al., 2018).
2010), but proper application is essential for maximum economic and Contrary to many fish species, shrimp feeding behaviour does not
environmental improvements in aquaculture farms (Chatvijitkul et al., allow visual perception of feed intake. Moreover, adequate estimations
2017; Van et al., 2017). of population size and biomass are essential for feed management
Shrimp are omnivorous benthic animals (Cuzon et al., 2004; Dall (NRC, 2011) which is particularly complex in non-clear water systems
et al., 1990; Varadharajan and Pushparajan, 2013) with limited capa- such as ponds. Therefore, estimating or adjusting feed inputs to meet
city to store food inside their digestive tract which results in slower the intake demands of shrimp can be very challenging. Regardless,
continued ingestion of small quantities of feed. Several studies have there are various strategies to manage feed inputs for shrimp produc-
shown enhanced growth performance for shrimp culture with multiple tion.
feedings throughout the day (Carvalho and Nunes, 2006; Jescovitch Quite often feed tables are used by farmers ((Casillas-Hernandez
et al., 2018; Ullman et al., 2019a). This is due to increasing the avail- et al., 2006 #23), 2006) which are based on previous production cycle
ability of feeds but also the time that feed is in contact with water which data and serve as a reference for future cycles regardless of feed de-
is accepted to reduce its nutritional value (Obaldo et al., 2002). Ullman livery system. Feed trays are one of the most common feed management
et al. (2019c) reported reduced growth performance and higher FCR in strategies for they allow gross estimation of feed intake (Martinez-

⁎
Corresponding author.
E-mail address: [email protected] (J. Reis).

https://doi.org/10.1016/j.aquaculture.2019.734759
Received 20 May 2019; Received in revised form 19 November 2019; Accepted 19 November 2019
Available online 21 November 2019
0044-8486/ © 2019 Elsevier B.V. All rights reserved.
J. Reis, et al. Aquaculture 519 (2020) 734759

Cordova et al., 1998). Nevertheless, being a very high labour-intensive treatments used were a standard feeding protocol (SFP) + 30%,
technique is a major setback (Bador et al., 2013; Davis et al., 2018; SFP + 45%, SFP + 60% and a passive acoustic feeding system (SF200
Ullman et al., 2019a). As a response to the necessity of the shrimp Sound feeding system, AQ1 Systems, Tasmania, Australia). SFP was
farming industry to improve its feed management protocols, some calculated based on an expected weight gain of 1.3 g/wk., a feed con-
techniques and technologies have risen to address this issue. version ratio (FCR) of 1.2, and a weekly mortality of 1.5% during
Timer feeders are not a recent technology and are extensively used growout period. The SFP used in this experiment was based on Davis
in various sectors and aquaculture production systems. These feeders et al. (2006) which was developed to optimize growth and FCR when
enable increasing the number of feedings without negatively impacting using two feedings per day, resulting in satisfactory results as reported
labor cost. Ullman et al. (2019a) has reported no significant improve- by Sookying et al. (2011). It was also used as the reference for the
ments in production for ponds fed same increasing feed amount twice a development of a protocol for timer feeders with satisfactory results as
day in contrast with ponds fed the same amount but fed six meals a day. well as reported by Sookying et al. (2011), Van et al. (2017), Jescovitch
This indicates that better productivity can be achieved by increasing et al. (2018) and Ullman et al. (2019a, 2019b). Each of the four re-
both number of meals and feed inputs. In parallel, animal feeding ac- plicates for every treatment was randomly assigned to a pond except for
tivity is also an important tool in aquaculture. Simplest feeding feed- the AQ1 system treatment due to electricity constraints. All feeders used
back in fish is visual observation which will not work in shrimp ponds for SFP treatments were BioFeeder (BioFeeder SA, Guayaquil, Ecuador)
due to both the size of the animal and poor visibility in the water. Using timer-feeders, feeding once every 20 min from 0700 to 1900. Biofeeder
a different approach, for the last decade on-demand acoustic feedback feed management (e.g. set feed amount, turn on/off) was done remotely
feeding systems have proven to be a reliable tool in shrimp farming using the feeder's specific software. AQ1 feeding system fed ad libitum
(Silva et al., 2019). These feeding systems respond to the signature using a hydrophone with computer software to monitor feeding ac-
clicking noise produced by shrimp while feeding. Previous works by tivity. All ponds under AQ1 system management were also equipped
Napaumpaipom et al. (2013) in high density, intensive systems and with an underwater DO sensor (placed approximately 10 cm off the
Jescovitch et al. (2018), Ullman et al. (2019a, 2019b) in semi-intensive pond bottom) and the system was set to only allow feeding when DO
conditions have shown improvements in growth performance by ap- levels were above 4 mg/l. In all four ponds under AQ1 system treatment
plication of acoustic demand-feeding system in comparison to hand- the main aerator was connected to the system so it could control aerator
feeding and timer feeder techniques in semi-intensive systems. As a activity based on information provided by DO sensor. All ponds were
continuation, this study aims towards improving timer-feeder protocols hand-fed a SFP-based amount twice a day for the first 30 days after
by adjusting feed amount and compare it with acoustic demand-feeding which BioFeeders were initiated. AQ1 system was started on the 34th
systems. day of pond production.

2. Material and methods 2.2. Sampling and water quality

This trial was performed at Alabama Department of Conservation After 17 days of pond culture, shrimp were sampled weekly through
and Natural Resources, Claude Peteet Mariculture Center, Gulf Shores, the remaining production cycle using a cast net (1.52 m radius and
Alabama. Pacific white shrimp L. vannamei larvae (2.3 mg) were ob- 0.96 cm mesh) to collect approximately 60 individuals per pond. Pond
tained from Shrimp Improvement Systems (Islamorada, FL, USA), ac- sampling enabled weight recording for growth assessment and inspec-
climated and nursed in a greenhouse system for 18 days. Juvenile tion for general health. Ponds were monitored (DO, temperature, sali-
shrimp (6 mg) were then stocked in outdoor ponds at a density of 26 nity, and pH) at least three times a day, at sunrise (0500–0530 h),
shrimp/m2. The production research was carried out in 16, 0.1 ha afternoon (1400–1430 h) and sunset (1900–2000 h), using a YSI
outdoor ponds over a 13 wk. production period. ProPlus meter (Yellow Springs Instrument Co., Yellow Springs, OH,
The ponds used through the growout period were approximately USA). Secchi disk readings were recorded once a week as total ammonia
0.1 ha in surface area (46 × 20 × 1.0 m) lined with 1.52 mm high- nitrogen (TAN) and chlorophyll a concentration were recorded twice a
density polyethylene with a 25 cm layer of sandy-loam soil on the week. Water samples were taken in the morning at the surface and TAN
bottom. Ponds were filled with brackish water (10.8–12.9 g/l) from was analysed with a high performance ammonia ion selective electrode
Intracoastal Canal between Mobile and Perdido Bay, Alabama, filtered (Thermo Fisher Scientific Inc., Waltham, MA, USA). Direct calibration
through a 250 μm cloth mesh filter bag. Pond primary productivity was of the electrode was conducted by preparing a serial dilution of a 100
promoted by adding inorganic fertilizers (1687 ml of 32–0-0 and +/− 1 mg/l ammonia standard (certified traceable to NIST standard
303 ml 10–34-0 for 5.70 kg/ha of N and 1.03 kg/ha of P) to the ponds reference material) to create three ammonia standards (0.1, 1.0 and
two weeks prior to stocking. The same fertilizing treatment was re- 10.0 mg/l), calibration was performed prior to each week's analysis.
peated for every pond one week after pond stocking as Secchi readings Chlorophyll samples were taken once a week by filtering a water
for all ponds was still approximate to the ponds total depth. To try to sample through glass fiber filters (47 mm diameter) using a vacuum
maintain dissolved oxygen (DO) above 3 mg/l, all ponds were supplied pump. Filters were kept in plastic 35 mm film canisters and shipped to
one 2-hp surface aerator (Aire-O2, Aeration Industries International, E.W Shell Fisheries Center at Auburn University. Analyses were per-
Inc., Minneapolis, MN, USA) as the primary source of mechanical formed according to standard analytical protocols for chlorophyll a by
aeration and one 1-hp surface aerator (Aquarian, Air-O-Lator, Kansas membrane filtration, acetone-methanol extraction of phytoplankton
City, MO, USA) for backup and/or additional aeration. No water ex- and spectroscopy (Eaton et al., 2005).
change was done throughout this trial. All AQ1 treatment ponds were provided a DO sensor with real-time
oxygen information on those ponds. All sensors were cleaned twice
2.1. Feed management daily to prevent fouling and misreading. Calibration was performed
only once through the entire cycle. Due to equipment failure near the
All ponds were offered the same two diets: 1.5-mm commercial diet end of the cycle, one of the AQ1 treatment ponds had the DO sensor and
(40% crude protein, 9% crude lipids) produced by Zeigler Inc. automatic aeration disconnected and was fed ad libitum from 0700 to
(Gardners, PA, USA) for the first four weeks, and 2.4-mm protein soy 1900.
optimized feed (35% crude protein, 8% crude lipids) produced by
Zeigler Inc. from the fourth week on, according to the treatments. Diet 2.3. Harvest and shrimp value
formulation for this experiment was the same as used by Ullman et al.
(2019a). For evaluation of the potential for automation the four The ponds were harvested over three days at the end of the 13-week

2
J. Reis, et al. Aquaculture 519 (2020) 734759

culture period. Ponds were partially drained and the night before har- 25
vest the level was reduced to about one third and aeration was provided y = 0.0254x - 3.7084

DO<2.5 mg/L Occurrences

20 R² = 0.0944
using the surface aerator. On the day of harvest, the remaining water
was drained and the shrimp were pumped out of the catch basin using a
15
hydraulic fish pump equipped with a 25 cm diameter suction pipe
(Aqua-Life pump, Magic Valley Heli-arc and Manufacturing, Twin-Falls, 10
Idaho, USA). The pump was placed in the catch basin and shrimp were
SFP+30% SFP+45%
pumped, de-watered, and collected into a hauling truck. Shrimp were 5
then rinsed, weighed in bulk, and 150 were randomly selected to SFP+60% AQ1
measure individual weights and determine the size distribution. A 0
450 500 550 600 650 700 750
subsample of these shrimp were collected and frozen for subsequent
Total Feed Input (kg/pond)
analysis. Whole body proximate with minerals analysis of the shrimp
was performed by Midwest Laboratories (Omaha, NE, USA). Fig. 1. Relationship between total low oxygen occurrences (< 2.5 mg/l) per
Shrimp prices used were the three year average (2014–2016) as treatment and total feed input.
reported by Urner Barry (Urner Barry, Toms River, NJ, USA) for Latin
American Farmed white shrimp, whole. The partial value was calcu- Production data is summarized in Table 2 with final weights and
lated by subtracting the feed costs from the production value as cal- yield generally following the level of feed input. The mean final in-
culated from the Urner Barry prices and the size distribution of shrimp dividual weights of shrimp were significantly different between timer
produced. The feed prices were $1.72/kg for the starter diet and $1.09/ feeder treatments and AQ1 but not among timer feeder treatments.
kg for the grower diet. Weekly growth and yield were significantly different between the two
treatments with lower feed inputs (SFP + 30% and SFP + 45%) and
2.4. Statistical analysis the highest feed input treatment (AQ1). Survival ranged between 72.5
and 81.4% and FCR between 0.96 and 1.11 but no statistical differences
Statistical analysis of the growth data was conducted with SAS 9.4 were found among these parameters. Figs. 3 and 4 present average
(SAS Institute, Cary, NC, USA) to perform a one-way analysis of var- treatments feed inputs and average individual weight throughout the
iance to determine significant difference (p-value < .05) among treat- production cycle. Feed inputs (kg/ha) were different among treatments,
ments, the assumptions for ANOVA were met. Student-Newman-Keuls as shown in Table 2.
multiple range test was used to determine differences among treat- Results for feed input analysis are summarized in Fig. 4. Data
ments. Effect of feed inputs in low DO occurrences was analysed summarized in Fig. 4 did not include data until day 17 due to lack of
through a regression analysis. sampling although feed amount was adjusted on day 10 based on ex-
pected growth and survival. Combined analyses of data revealed in-
3. Results creasing differences in size as previously indicated by Figs. 2 and 3.
Proximate whole body composition analysis are summarized in
During this trial, main water quality parameters were kept within Table 3. SFP + 60% produced shrimp with significantly lower ash%
typical range for shrimp production (Boyd and Tucker, 1992) (Table 1). than SFP + 45% but no other statistical differences were found in any
To evaluate the effects of nutrient loading on oxygen demand. The of the other parameter evaluated in whole body composition analysis.
occurrences of DO reading below 2.5 were summer across time. Figure1 Feed costs and economic value of shrimp produced is summarized in
shows the number of low DO occurrences for each pond identified by Table 2. Significant differences were found for all treatments in both
treatments. Regression analysis was conducted on the whole data set feed inputs and feed cost. However, for shrimp value and partial income
regressed against final feed input. Although feed inputs affect the ponds statistical difference were only found between both SPF + 30% and
biological oxygen demand (BOD) and most occurrence were registered SFP + 45% in comparison to AQ1 treatment.
at dawn, there is no linear correlation (R2 = 0.0944) (Fig. 1) between
the number of low oxygen occurrences (< 2.5 mg/l) in DO readings and
the feed input for each pond.

Table 1
- Summary of water quality parameters for the four treatments over the 13 wk. culture period. Values are presented as mean ± standard deviation and maximum
and minimum value are presented in parenthesis.
SFP + 30% SFP + 45% SFP + 60% AQ1

a
Morning DO (mg/l) 3.81 ± 1.14 3.95 ± 1.33 3.66 ± 1.04 3.66 ± 1.11
(1.65, 9.90) (0.82, 13.81) (1.77, 7.93) (0.78, 7.02)
a
Afternoon DO (mg/l) 10.68 ± 2.78 10.48 ± 2.81 10.69 ± 2.73 10.60 ± 2.99
(4.32, 18.05) (2.71, 21.36) (3.38, 16.97) (2.94, 10.02)
Night DOa (mg/l) 9.73 ± 2.70 9.34 ± 2.97 9.31 ± 2.69 9.35 ± 3.04
(3.56, 18.5) (3.17, 24.11) (2.77, 16.89) (1.87, 18.36)
Temperature (°C) 31.8 ± 1.7 31.7 ± 1.6 31.6 ± 1.7 31.4 ± 1.6
(27.4, 36.3) (27.5, 38.1) (24.6, 35.4) (27.3, 35.0)
pH 8.48 ± 0.79 8.45 ± 0.75 8.39 ± 0.76 8.33 ± 0.70
(6.81, 10.01) (6.8, 9.81) (6.87, 9.87) (6.95, 10.18)
Salinity 9.27 ± 1.35 10.71 ± 2.58 9.68 ± 1.42 10.28 ± 1.25
(g/l) (7.13, 12.09) (7.73, 11.41) (6.72, 12.36) (8.03, 12.88)
TANb 0.4 ± 0.7 0.5 ± 1.0 0.6 ± 1.0 0.7 ± 1.9
(mg/l) (< 0.001, 3.0) (< 0.001, 4.0) (< 0.001, 4.0) (< 0.0001, 6.0)
Chrolorophyll a 307 ± 213 363 ± 202 396 ± 325 318 ± 203
(μg/l) (3.7, 990) (71, 745) (25, 1742) (35, 1044)

a
DO - Dissolved Oxygen.
b
TAN - Total Ammonia Nitrogen.

3
J. Reis, et al. Aquaculture 519 (2020) 734759

Table 2
- Summary of Pacific white shrimp response to different feed management protocols.
Treatment IndW (g) Survival Weekly Growth (g) Yield (kg/ha) Total Feed Input (kg/ha) FCR Feed Cost ($/ha) Shrimp Value ($/ha) Partial Income ($/ha)

a a a a a a
SFP + 30% 26.29 77.6 1.97 5226 4933 0.99 5592 43,490 37,898a
SFP + 45% 26.87a 75.2 2.04a 5115a 5332b 1.11 6026b 42,468a 36,442a
SFP + 60% 29.04a 80.7 2.21ab 6128ab 5844c 0.96 6585c 52,623ab 46,039ab
AQ1 32.53b 81.4 2.49b 6869b 6984d 1.02 7828d 60,723b 52,896b
P-value 0.0096 0.9083 0.0091 0.0274 < 0.0001 0.7313 < 0.0001 0.0073 0.0164
PSE 1.18 6.52 0.093 39.62 5.07 0.097 55.3 3362 3380

1
PSE: Pooled Standard Error.

90 automatic feeders on 4. Discussion

80
Commercial shrimp feeds are considered nutritionally appropriate
70
Feed Inputs (kg/pond)

and are one of the primary operating costs of most farms. To ensure the
60 investment in high quality feed is maximized it is important to focus on
50 feeding protocols Shrimp have been traditionally fed 2 to 4 meals a day
40 either by hand-dispersion or through the use of feed trays. However,
30 shrimp can be described as grazers in that they have evolved to find
SFP+30% SFP+45% small patches of food with high frequency indicating that feed fre-
20
SFP+60% AQ1 quency is an important driver of nutrient intake. Ullman et al. (2019a)
10 reported a significant increase in final weights of shrimp reared with 6
0 feeding/day as compared to those fed a similar amount of feed over two
0 10 20 30 40 50 60 70 80 90 feedings per day. The use of automation to increase the number of
Time (days) feedings not only favors shrimp growth but also improves economic
Fig. 2. Weekly feed inputs (kg/pond) through production cycle as average per
balance as labor requirement is reduced and feed efficiency is improved
treatment. Feed inputs were equivalent for the first 30–34 day. Timer feeders (Davis et al., 2018). Application of automatic feeders has shown many
were initiated on day 30 and AQ1 feeders on day 34. advantages in comparison with traditional methods. Within automatic
feeders, on-demand acoustic feedback systems have shown improved
performance over simpler timer-feeders (Jescovitch et al., 2018;
35
Napaumpaipom et al., 2013; Ullman et al., 2019a, 2019b) and in some
Average individual weight (g)

30 SFP+30 cased improved water quality has been reported.

SFP+45 During this entire production cycle water quality management
25 aimed towards keeping dissolved oxygen levels above 3 mg/l. Given the
SFP+60
20 variability between ponds as well as the variation in feed management
AQ1
it is difficult to make conclusions of the water quality data. Jescovitch
15 et al. (2018) reported increased levels of TAN associated with increased
10
feed inputs using the AQ1 system. However, our feed loading was
considerably higher than the previously mentioned study yet there were
5 minimal differences in water quality. The lack of differences across feed
input levels would indicate that we were within the processing capacity
0
of the pond based ecosystem. Under our conditions, aeration was
0 10 20 30 40 50 60 70 80 90
managed either using automated set points AQ1 system or through
Time (day)
manual management. Although managed we counted the days for
Fig. 3. Weekly average individual weight (g) as average per treatment. which DO dropped below 2.5 mg/l. This data was plotted against feed
inputs for each pond and presented in Fig. 1. This regression has a very
weak fit (R2 0.0944) and further statistical tests showed no significant

Table 3
Means of whole body composition for each treatment as analysed by Midwest Laboratories (Omaha, NE, USA).
Treatment SFP + 30% SFP + 45% SFP + 60% AQ1 P-value PSE

Moisture % 74.9 75.0 75.2 74.1 0.5042 1.95

Dry Matter % 25.08 25.05 25.93 24.85 0.5042 1.95
Protein % 74.5 73.2 78.1 76.3 0.4336 8.79
Fat % 4.16 3.89 4.94 5.66 0.3793 2.75
Ash % 11.67ab 12.83a 10.29b 11.03ab 0.0250 2.06
Sulfur % 0.84 0.80 0.85 0.82 0.1278 0.04
Phosphorus % 1.62 1.61 1.43 1.57 0.2436 0.26
Potassium % 1.27 1.24 1.27 1.27 0.8351 0.09
Magnesium % 0.35 0.37 0.30 0.35 0.2459 0.09
Calcium % 3.62 4.10 3.09 3.69 0.3103 1.46
Sodium % 0.73 0.73 0.73 0.69 0.5147 0.08
Iron (ppm) 152.8 161.6 101.2 202.8 0.4222 173.41
Manganese (ppm) 7.6 7.1 3.6 6.6 0.3363 57.07
Copper (ppm) 137.5 136.0 125.8 136.8 0.0640 27.94
Zinc (ppm) 78.3 73.4 75.5 75.2 0.3439 6.18

4
J. Reis, et al. Aquaculture 519 (2020) 734759

differences between treatments (p = .2469) ultimately confirming that based on the amount of food with higher feed inputs resulting in larger
the ponds were able to process the nutrients load. shrimp. Supporting this conclusion is the fact that ponds fed
During the first 30 days feeding program for all treatments was SFP + 60% also registered numerically higher average survival. Also,
preprogramed following the previously described SFP which assumes although feed inputs were only differentiated from day 30 on (Fig. 2), it
estimates for the population as well as growth. Although this is not an is possible to identify larger individuals (Fig. 3) in SFP + 60% ponds at
optimized protocol it is assumed that primary productivity is con- the same time as feed inputs by percentage body weight (Fig. 4) remain
siderable portion of nutrient intake and that feed inputs must be sys- similar. This is likely a consequence of numerically higher survival
tematically increased to allow conditioning of the pond to the high feed (Table 2) in this treatment regardless of higher feed input and shrimp
loads. Also, as shrimp feed lower in food chain ponds primary pro- adjusting their growth to feed input as well. In short, combined analysis
ductivity is more than likely one of the main sources of nutrients at this of data summarized in Figs. 3 and 4 indicates that shrimp are able to
stage and uneaten feed will trigger phytoplankton growth as well (NRC, adjust their growth based on feed availability it also suggests that there
2011). After 30 days of culture, treatments were initiated and feed was may be a threshold for feed input over which relative growth does not
dispersed using timer feeders. At day 34 the AQ1 system was initiated. increase. Consequently, from an economical and water quality man-
Total feed inputs (kg/ha) were significantly different for every treat- agement perspective our data suggests that shrimp could have been be
ment (Table 2). By evaluating feed inputs through the production cycle fed SFP + 30% until individual sizes reach about 18 g (~day 50) and
(Fig. 2) and comparing this to the average individual growth (Fig. 3) it then feed inputs would be increased to SFP + 60% until the end of
is possible to discern some feeding differences. Between days 38 and 45 production. Possibly even further increase feed inputs for the last two
there was a substantial reduction in feed input for AQ1 feeding system. weeks of production, as was seen in the AQ1 system, was responsible
There are two possible interpretations of this: the small size of shrimp for further increased shrimp size (Table 2).
producing a minimal acoustic signal resulting in low feed inputs or Feed management and nutrient composition of the diet is known to
primary productivity remains a sufficient food source for shrimp within influence proximate composition of the animal albeit shrimp seem to be
that size class. As there no differences in mean weights it would appear less responsive than other animals. To evaluate possible shifts in nu-
reduced feed inputs did not affect growth. From this point forward AQ1 trient content, proximate analysis of whole shrimp body composition
feed inputs steadily increase up to day 59 where it peaked. From (Table 3) were determined. Significantly higher ash content of shrimp
50 days to the end of production, feed inputs were highest for the AQ1 fed in SFP + 45% in comparison to SFP + 60%. Ullman et al. (2019b)
treatments. Based on sample weight it is apparent that up to 45 days of has reported differences in several compounds between treatments,
culture the lowest level of feed input was acceptable. However, after namely higher fat content for higher feed input treatments. However, in
this point SFP + 30% and SFP + 45% feed treatment resulted in this research no differences were found in any components except for
smaller shrimp or a reduced growth rate. Shrimp fed using the ash. In our work ash was significantly higher in shrimp reared on the
SFP + 60% level maintained similar growth as the AQ1 system through SFP + 60% treatment as compared to those on the SFP + 60% feeding
day 73 after which it appeared that growth was reduced. This data leads regime. Variation in ash content was not consistent across feed inputs;
us to believe it is possible to obtain high growth rate with lower feed hence, it may simply be due to natural variation in the data or possibly
inputs than AQ1 although at some point feed will become a limiting small changes macro minerals such as Ca and P.
factor for growth. Regardless no differences in FCR among treatments
were registered and reported values are more than acceptable 5. Conclusion
throughout all treatments.
Shrimp are not sampled during the first week as representative The results of this study underline the results achieved in similar
samples are difficult to obtain with small shrimp in ponds. Hence, with studies by Jescovitch et al. (2018) and Ullman et al. (2019a, 2019b),
the exception of the first few week of production the collected data can indicating that higher production and value of L. vannamei produced in
be used to develop a feed curve. To do this, final survivals are used to semi-intensive pond culture can be achieved through application of on-
back calculate the estimated number of shrimp at any given time point demand acoustic feedback systems. This study also shows that it is
and the percent body weight calculated. This data is presented in Fig. 4 possible to establish an efficient feeding protocol for timer feeders.
which does not include data from the first 17 days of production. This Therefore, reducing the performance differences between the two
data can then be used as a recommended feed rate for shrimp produced technologies. Nevertheless, efficient use of timer feeders heavily relies
under similar conditions. in adequate feeding plans based on previous production cycles as well
Combined analyses or data also suggest that shrimp adjusted growth as post feeding observations. Poor estimations of survival, growth and
feed response are likely to negatively affect growth, environmental
conditions (water quality) and financial performance.
4.5
For the intrinsic nature of a feedback technology is to feed on de-
4.0 mand in real time, it is virtually impossible that any timer feeder will be
y = -9E-05x3 + 0.0091x2 - 0.2983x + 4.4767 as efficient as a real-time passive feedback system. However, our results
3.5 R² = 0.8937
confirm that a standard feeding protocol can be developed for auto-
% Body weight fed

3.0
mated feeding system that will support the enhanced growth rates seen
2.5 when using these systems. Thus, providing guidance for this level of
2.0 technology. Increased product value may also offset the installation and
1.5
running cost of any of these technologies. However, as reported by
Ullman et al., (2019a), it is not possible to accurately provide im-
1.0 plementation costs due to a lack of linearity inherent to the facility and
0.5 production setup.
0.0
0 5 10 15 20 25 30 Acknowledgements
Individual weight (g)
SFP+30% SFP+45% SFP+60% The authors would like to express our gratitude to those who have
reviewed this manuscript and the students and staff who participated in
Fig. 4. Back calculated feed inputs expressed as percentage body weight for the this project from Auburn University and Claude Peteet Mariculture
various sizes of shrimp. Regression represents the results of pooled data. Center. Special thanks to Zeigler Bros, Inc., BioFeeders and AQ1 for

5
J. Reis, et al. Aquaculture 519 (2020) 734759

support of the research. This research was supported by United Soybean Eaton, A.D., Clesceri, L.S., Greenberg, A.E., Franson, 2005. Standard methods for the
Board Project #1810-353-0536 as well as the Hatch Funding Program examination of water and wastewater. Washington, DC, USA: American Public Health
Association, 21, 1600.
of Alabama Agriculture Experiment Station. The mention of trademarks Jescovitch, L.N., Ullman, C., Rhodes, M., Davis, D.A., 2018. Effects of different feed
and proprietary products does not constitute endorsement by Auburn management treatments on water quality for Pacific white shrimp Litopenaeus van-
University and is not intended to exclude other products or services that namei. Aquac. Res. 49, 526–531.
Martinez-Cordova, L.R., Porchas-Cornejo, M.A., Villarreal-Colemnares, H., Calderon-
may be suitable. Perez, J.A., Naranjo-Paramo, J., 1998. Evaluation of three feeding strategies on the
culture of white shrimp Penaeus vannamei Boone 1931 in low water exchange ponds.
Declaration of Competing Interest Aquac. Eng. 17, 21–28.
Napaumpaipom, T., Chuchird, N., Taparhudee, W., 2013. Study on the efficiency of three
different feeding techniques in the culture of Pacific white shrimp (Litopenaeus van-
All authors of this research project would like to disclose the ab- namei). Kasetart Univers. Fis. Res. Bull 37 (2), 8–16.
sence any personal or financial relationship with people or organiza- NRC, 2011. Nutrient Requirements of Fish and Shrimp. National Academic Press,
Washington D.C (376pp).
tions that may inappropriately influence our work. This naturally in-
Obaldo, L.G., Divakaran, S., Tacon, A.G., 2002. Method for determining the physical
cludes all grant and sources of funding described in the paper. stability of shrimp feed in water. Aquac. Res. 33 (5), 369–377.
Quintero, H., Roy, 2010. In: Alday-Sanz, V. (Ed.), Practical feed management in semi-
References Intensive systems for shrimp culture. The shrimp book. Nottingham University Press,
Nottingham, UK, pp. 443–454.
Silva, J.F., Hamilton, S., Rocha, J.V., Borie, A., Travassos, P., Soares, R., Peixoto, S.J.A.,
Bador, R., Blyth, P., Dodd, R., 2013. Acoustic control improves feeding productivity at 2019. Acoustic characterization of feeding activity of Litopenaeus vannamei in cap-
shrimp farms. Global Aquacul. Advocate 77–78. tivity. Aquaculture 501, 76–81.
Boyd, C.E., Tucker, C.S., 1992. Water Quality and Pond Soil Analyses for Aquaculture. Sookying, D., Silva, F.S.D., Davis, D.A., Hanson, T.R., 2011. Effects of stocking density on
Alabama Agricultural Experiment Station. Auburn University. the performance of Pacific white shrimp Litopenaeus vannamei cultured under pond
Carvalho, E.A., Nunes, A.J.J.A., 2006. Effects of feeding frequency on feed leaching loss and outdoor tank conditions using a high soybean meal diet. Aquaculture 319,
and grow-out patterns of the white shrimp Litopenaeus vannamei fed under a diurnal 232–239.
feeding regime in pond enclosures. Aquaculture 252 (2–4), 494–502. Tacon, A.G.J., Forster, I.P., 2003. Aquafeeds and the environment: policy implications.
Casillas-Hernandez, R., Magallon-Barajas, F., Portillo-Clarck, G., Paez-Osuna, F., 2006. Aquaculture 226, 181–189.
Nutrient mass balances in semi-intensive shrimp ponds from Sonora, Mexico using Ullman, C., Rhodes, M., Hanson, T., Cline, D., Davis, D.A., 2019a. Effects of four different
two feeding strategies: trays and mechanical dispersal. Aquaculture 258, 289–298. feeding techniques on the pond culture of Pacific white shrimp, Litopenaeus vannamei.
Chatvijitkul, S., Boyd, C.E., Davis, D.A., McNevin, A.A., 2017. Embodied resources in fish J. World Aquacult. Soc. 50, 54–64.
and shrimp feeds. J. World Aquacult. Soc. 48, 7–19. Ullman, C., Rhodes, M.A., Davis, D.A., 2019b. Feed management and the use of automatic
Cuzon, G., Gaxiola, G., Rosas, C., Guillaume, J., 2004. Nutrition of Litopenaeus vannamei feeders in the pond production of Pacific white shrimp Litopenaeus vannamei.
reared in tanks and ponds. Aquaculture 235, 513–551. Aquaculture 498, 44–49.
Dall, W., Hill, J., Rothlisberg, P.C., Staples, D.J., 1990. The biology of Penaeidae. Ullman, C., Rhodes, M., Davis, D.A., 2019c. The effects of feed leaching on the growth of
Advances in Marine Biology. Vol. 27 Academic Press, New York, NY. Pacific white shrimp Litopenaeus vannamei in a green-water system. Aquac. Res. 50,
Davis, D.A., Amaya, E., Venero, J., Zelaya, O., Rouse, D.B., 2006. A Case Study on Feed 3074–3077.
Management to Improving Production and Economic Returns for the Semi-Intensive Van, T.P.T.H., Rhodes, M.A., Zhou, Y., Davis, 2017. Feed management for Pacific white
Pond Production of Litopenaeus vannamei. Avances en Nutrición Acuícola VIII. shrimp Litopenaeus vannamei under semi-intensive conditions in tanks and ponds.
Memorias del Octavo Simposio Internacional de Nutrición Acuícola. Universidad Aquac. Res. 48, 5346–5355.
Autonoma de Nuevo Monterrey, Nuevo León, México, pp. 282–303 (970-694-333-5). Varadharajan, D., Pushparajan, N., 2013. Food and feeding habits of aquaculture candi-
Davis, D.A., Ullman, C., Rhodes, M., Novriadi, R., Swanepoel, A., 2018. Automated date a potential crustacean Pacific white shrimp Litopenaeus vannamei. J Aquac Res
Feeding Systems in Pond Production of Pacific White Shrimp. Global Aquaculture Development 4, 1–5.
Advocate. https://www.aquaculturealliance.org/advocate/automated-feeding-
systems-in-pond-production-of-pacific-white-shrimp/.