View
1
Download
0
Category
Preview:
Citation preview
Understanding Shellfish AquacultureUnderstanding Shellfish AquacultureEnvironments in North America and Europe byEnvironments in North America and Europe by
Combining Field Measurements withCombining Field Measurements withComputational Fluid Dynamics and BioenergeticComputational Fluid Dynamics and Bioenergetic
ModelsModelsOr: You Have to Go With the FlowOr: You Have to Go With the Flow
Presented to:Presented to:
Northwest Workshop on Bivalve Aquaculture andNorthwest Workshop on Bivalve Aquaculture andthe Environmentthe Environment
By: Carter Newell and John RichardsonBy: Carter Newell and John RichardsonBlue Hill Hydraulics, Blue Hill Hydraulics, gemcart@midcoast.comgemcart@midcoast.comFor presentation only: please do not copy, cite or distributeFor presentation only: please do not copy, cite or distribute
Outline of PresentationOutline of Presentation
Background: previous studiesBackground: previous studies
General approachGeneral approach
Numerical MethodsNumerical Methods
Analysis ProceduresAnalysis Procedures
Case Studies: Maine, WashingtonCase Studies: Maine, Washington
State, Gorge Harbour, IrelandState, Gorge Harbour, Ireland
Current work on geoducksCurrent work on geoducks
GIS application layersGIS application layers
Positive Aspects of ShellfishPositive Aspects of ShellfishAquacultureAquaculture
shellfish are an economicshellfish are an economicargument for clean waterargument for clean water
shellfish improve water quality byshellfish improve water quality bygrazing on phytoplankton helpinggrazing on phytoplankton helpingto control algal blooms, removeto control algal blooms, removeNitrogen from systemNitrogen from system
structures create habitat forstructures create habitat fordozens of other species,dozens of other species,increasing marine biodiversityincreasing marine biodiversity(Murray, Seed, and Newell, 2006,(Murray, Seed, and Newell, 2006,JSR)JSR)
no feed or outside ingredients areno feed or outside ingredients areadded to the systemadded to the system
locally owned businesses createlocally owned businesses createjobs in the coastal zone for thejobs in the coastal zone for theworking waterfrontworking waterfront
high quality mussels and oystershigh quality mussels and oysterssupport Ecotourism (locally) andsupport Ecotourism (locally) andprovide exportsprovide exports
History of Maine Mussel FarmingHistory of Maine Mussel Farming
Began as a wild fishery with landingsBegan as a wild fishery with landingsup to 10,000 up to 10,000 mtmt per year. 1940 per year. 1940’’s tos topresent.present.
Wild harvests depleted mussel beds.Wild harvests depleted mussel beds.19701970’’s.s.
GEM sent fishermen in 1981 to visitGEM sent fishermen in 1981 to visitLettLett brothers in Wexford, Ireland to brothers in Wexford, Ireland tolearn about bottom culture.learn about bottom culture.
Seeding too high density, reducedSeeding too high density, reducedgrowth on farms. 1980growth on farms. 1980’’s.s.
Development of mussel model toDevelopment of mussel model tooptimize growth rates and seed tooptimize growth rates and seed toharvest yields. 1990harvest yields. 1990’’s to present.s to present.Production still limited by wild seedProduction still limited by wild seedresources.resources.
Development of more sustainableDevelopment of more sustainableraft culture and an aquacultureraft culture and an aquacultureexpert system. 2000-present.expert system. 2000-present.
Early work: Define Maine Mussel Feeding andGrowth in Context of the Carrying Capacity
Model MUSMOD
• Started with a simple formulation of energy flow, food supply and demand.Sensitivity analysis demonstrated the importance of food concentration andquality, assimilation of organic matter by the shellfish, and current speed.
• Developed the refined model at Mud Cove lease area, where sestonmeasurements were coupled with measurements of mussel growth. Model waschanged in concert with the results of the field program.
• Model was validated at two other sites.
• Final model used to manage seeding densities increased growth rates and seedto harvest yields. Growth to market size over 1 year varied from 300 musselsm-2 (Mud Cove) to 900 m-2 (Frenchman’s Bay).
Mud Cove, Maine, approx. 15 hectares, farmed since 1982, over 5000 tons lbs.production and lots of happy Eider ducks
Factors Affecting Shellfish GrowthFactors Affecting Shellfish Growth
Food concentrationFood concentration and andquality (phytoplankton,quality (phytoplankton,detritus, dissolveddetritus, dissolvedorganic matter,organic matter,particulate inorganicparticulate inorganicmatter)matter)
Shellfish density andShellfish density andbiomassbiomass
HydrodynamicsHydrodynamics of the of theculture system (tidal andculture system (tidal andwind-driven current speedwind-driven current speedand direction, waves)and direction, waves)
Water temperatureWater temperature
Campbell and Newell,1998, JEMBE 219: 171-204
Mud Cove model vs data
0
0.5
1
1.5
125155 185215 245275 305335365 395425 455485515
Day of the Year
g dry wt.
Mussel Growth vs DensityMussel Growth vs Density Shellfish reach a Shellfish reach a maximum biomassmaximum biomass (g dry tissue (g dry tissue
weight per square meter) based on the carryingweight per square meter) based on the carryingcapacity of the site based on capacity of the site based on seasonal foodseasonal foodconcentration and site-specific hydrodynamicsconcentration and site-specific hydrodynamics..
Growth to market size is also determined on a Growth to market size is also determined on a locallocalscalescale by local mussel density (mussel bottom patch by local mussel density (mussel bottom patchsize or number of mussels per m of culture rope)size or number of mussels per m of culture rope)and and farm scalefarm scale total mussel density (total biomass total mussel density (total biomasson the farm).on the farm).
It is easy to determine the It is easy to determine the appropriate seedingappropriate seedingdensitydensity for a for a targeted market sizetargeted market size (i.e. 1 g dry (i.e. 1 g dryweight, 60 mm shell length) by looking at theweight, 60 mm shell length) by looking at themaximum biomass at a site during harvest: i.e.maximum biomass at a site during harvest: i.e.500 g m500 g m-2-2 , mussels will be either .5 grams at a , mussels will be either .5 grams at adensity of 1000mussels mdensity of 1000mussels m-2-2 or 1 g at 500 mussels or 1 g at 500 musselsmm-2-2..
Similarly for rope culture, maximum biomass of 500Similarly for rope culture, maximum biomass of 500g mg m-1-1 of rope could grow 250 mussels m of rope could grow 250 mussels m-1-1 at 2 g at 2 geach or 1000 mussels meach or 1000 mussels m-1-1 at .5 g each. at .5 g each.
Once the local mussel density is controlled throughOnce the local mussel density is controlled throughgrowing practices to an optimum level, growth isgrowing practices to an optimum level, growth isthen controlled by the then controlled by the farm scalefarm scale food supply and food supply anddemand, which can be coupled with estuary scaledemand, which can be coupled with estuary scalemodels such as ECOWIN2000.models such as ECOWIN2000.
0
500
1000
1500
2000
2500
0 500 1000 15002000 2500
Density (no. m- 1)
Measuring Mussel Feeding RatesMeasuring Mussel Feeding Rates Benthic ecosystem tunnel (Newell andBenthic ecosystem tunnel (Newell and
Shumway, 1993, NATO ASI Series, R. DameShumway, 1993, NATO ASI Series, R. DameEditor,. 87-148).Editor,. 87-148).
Profiling of seston in the benthic boundary layerProfiling of seston in the benthic boundary layer(Muschenheim and Newell, 1992, MEPS 85: 131-(Muschenheim and Newell, 1992, MEPS 85: 131-136). Showed the importance of resuspended136). Showed the importance of resuspendedbenthic diatoms, and defines feeding zonebenthic diatoms, and defines feeding zone(bottom 10 cm). At edge of bed, phytoplankton(bottom 10 cm). At edge of bed, phytoplanktonand organic detritus enhanced 10 fold. At middleand organic detritus enhanced 10 fold. At middleof bed, vertical mixing supplies food fromof bed, vertical mixing supplies food fromsurface water to the bottom.surface water to the bottom.
Using particle counts and oxygen concentrationUsing particle counts and oxygen concentrationin flow-through feeding chambers to estimatein flow-through feeding chambers to estimatescope for growth (Newell et al., 1998 JEMBE 219:scope for growth (Newell et al., 1998 JEMBE 219:143-169).143-169).
Biodeposition chambers: PIM in available foodBiodeposition chambers: PIM in available foodused as a tracer in biodeposits to calculate waterused as a tracer in biodeposits to calculate waterfiltration rates. Newell et al. , 2005 JEMBE 321:filtration rates. Newell et al. , 2005 JEMBE 321:109-124.109-124.
Flume and field work determinining optimumFlume and field work determinining optimumflows and food concentration for mussel feedingflows and food concentration for mussel feedingutilizing video of mussel exhalant siphon area.utilizing video of mussel exhalant siphon area.Newell et al., 2001, JEMBE 262: 92-111, Ph.D.Newell et al., 2001, JEMBE 262: 92-111, Ph.D.thesis UNBSJ 2005).thesis UNBSJ 2005).
Effects of water velocity and particleEffects of water velocity and particleconcentration on mussel filtration rates: can beconcentration on mussel filtration rates: can be
used to identify optimum conditions for field grow-used to identify optimum conditions for field grow-out of mussels on rafts (from Newell et al., 2001,out of mussels on rafts (from Newell et al., 2001,
JEMBE 262)JEMBE 262)
0
20
40
60
80
100
120
0 20 40
Velocity U (cm s-1
)
% M
axim
um
Exh
ala
nt
Sip
ho
n A
rea
A
0
25
50
75
100
0 10000 20000 30000 40000 50000
Concentration (number ml-1
)P
erc
en
t m
axim
um
exh
ala
nt
sip
ho
n a
rea
A
Field Data CollectionField Data Collection
Profiling CTD and Multi-parameter Sonde(courtesy of Seabird Electronics, Inc. and SonTek/YSI, Inc.)
Water Quality ParametersWater Quality Parameters
CTDsCTDs and multi-parameter and multi-parameter sondessondes
are used to measure water qualityare used to measure water quality
parameters.parameters.
Seabirds may be used to profile Seabirds may be used to profile
in suspended systems or inin suspended systems or in
moored mode for bottommoored mode for bottom
cultures.cultures.
Field Data CollectionField Data CollectionVelocitiesVelocities
Acoustic Doppler CurrentAcoustic Doppler Current
Profilers (Profilers (ADCPsADCPs) and) and
electromagnetic current meterselectromagnetic current meters
are used to monitor velocitiesare used to monitor velocities
at aquaculture site.at aquaculture site.
Boundary layer physics may be Boundary layer physics may be
estimated from data at specifiedestimated from data at specified
heights off the bottom (i.e. 1 m) andheights off the bottom (i.e. 1 m) and
measurements of bottom roughnessmeasurements of bottom roughness
(i.e. shell length).(i.e. shell length).
ADCPs and S4 Current Meter(courtesy of SonTek/YSI, Inc. and Interocean Systems, Inc.)
Principle Objectives of ModelingPrinciple Objectives of Modeling
Growers and regulators want to optimize shellfish Growers and regulators want to optimize shellfish
aquaculture methods to:aquaculture methods to:
Maximize Maximize –– Production. This is a function of the site Production. This is a function of the site
specific flux of particulates. Matching biomassspecific flux of particulates. Matching biomass
distribution with site characteristics can reduce grow-distribution with site characteristics can reduce grow-utut
time and increase harvest/seed yields.time and increase harvest/seed yields.
Minimize Minimize –– Impact. This is also a function of water Impact. This is also a function of water
velocity, due to the dispersal and velocity, due to the dispersal and resuspensionresuspension of organic of organic
matter and the oxygen flux to the benthos (Panchang,matter and the oxygen flux to the benthos (Panchang,
Cheng and Newell, 1997, Estuaries 20: 14-41; Dudley,Cheng and Newell, 1997, Estuaries 20: 14-41; Dudley,
Panchang and Newell, 2000, Aquaculture 187, 319-349)Panchang and Newell, 2000, Aquaculture 187, 319-349)
Numerical MethodsNumerical MethodsOptimization of aquacultureOptimization of aquaculture
methods involves hydraulic responsemethods involves hydraulic response
on different length scales:on different length scales:
Large Scale (L = 10Large Scale (L = 1022 –– 10 1044 meters) meters)
Advection and dispersion of foodAdvection and dispersion of food
particles and waste materialsparticles and waste materials
b)b) Small Scale (L = 10Small Scale (L = 1000 –– 10 1022 meters) meters)
Movement and consumption ofMovement and consumption of
food particles within aquaculturefood particles within aquaculture
systemssystemsGorge Harbour (top), Desolation Sound Oyster Co. (bottom)
Cortes Island, BC
Analysis Procedure: Large ScaleAnalysis Procedure: Large ScaleKillary Harbor, Co. Galway, IrelandKillary Harbor, Co. Galway, Ireland
Killary Harbour(Digital Terrain Model and Tidal Record)
Preliminary ModelingPreliminary Modeling ( (i.e., i.e., screeningscreening
analysis) requires the following data:analysis) requires the following data:
BathymetryBathymetry
Boundary Flow ConditionsBoundary Flow Conditions
and suggests solutions to perceivedand suggests solutions to perceived
problems.problems.
In this case, a mussel grower wasIn this case, a mussel grower was
getting slow growth in an area ofgetting slow growth in an area of
the lough relative to other sitesthe lough relative to other sites
Preliminary ModelingPreliminary Modeling
Calculated Tidal Response
Large scale model: calibrationLarge scale model: calibration
Calibration Results
Preliminary Modeling: Results showed that the new musselPreliminary Modeling: Results showed that the new musselfarm was in a farm was in a ““bad areabad area”” with low flows and with low flows and recirculatingrecirculating
eddies.eddies.
Calculated Tidal Response(detail)
Location ofproblem musselfarm site
Large-scale model as a screeningLarge-scale model as a screeningtool for site selectiontool for site selection
Bottom culture Raft culture
Small-Scale Modeling TechniquesSmall-Scale Modeling Techniques
Wake Formation Downstream of Mussel Rafts
The use of Computational Fluid Dynamics (CFD) modeling techniques toThe use of Computational Fluid Dynamics (CFD) modeling techniques to
model small-scale flow patterns within aquaculture systems was model small-scale flow patterns within aquaculture systems was
investigated by Newell and Richardson (USDA Funding to Great Eastern Mussel Farms) for investigated by Newell and Richardson (USDA Funding to Great Eastern Mussel Farms) for
Maine Mussel Rafts, Oyster rafts in B.C. and Washington State Mussel Rafts.Maine Mussel Rafts, Oyster rafts in B.C. and Washington State Mussel Rafts.
CFD Modeling of Floating Rafts:CFD Modeling of Floating Rafts:Calibration of model with field dataCalibration of model with field data
61117371317Calculated Data
4.79.716.63.16915.4Measured Data
7654321Location
Relative Fluorescence Measurements
1
6
4 3 2
7
5
Comparison of Calculated and Measured Velocities
Taylor rafts at Totten InletTaylor rafts at Totten Inlet
Chl a depletion through a section of 6 mussel rafts: Model ResultsChl a depletion through a section of 6 mussel rafts: Model Results
Supported by the Sea Grant National Marine Aquaculture Initiativein cooperation with the Pacific Shellfish Institute
120%0.054‘\\\’ Gap
111%0.050‘\\’ Gap
107%0.048‘+’ Gap
100%0.045As-built
% of As-builtAvg Vel (m/s)Configuration
What-if ScenariosWhat-if Scenarios
Field MeasurementsField Measurements
Field Data Collection
Does this sound familiar to you?Does this sound familiar to you?
““We donWe don’’t have enought have enoughequipmentequipment””
““Why did you put the velocityWhy did you put the velocitymeter in the shadow of thatmeter in the shadow of thatislandisland””
““If we only had another data pointIf we only had another data pointover thereover there…”…”
Analytical models can be used toAnalytical models can be used tooptimize field programs; optimize field programs; e.g.,e.g.,minimize data requirements andminimize data requirements andreduce errorreduce error
For example, current data collection for controlflows are best taken to the sides of mussel raftsrather than upstream or downstream of them..
Contribution of CFD Modeling:Contribution of CFD Modeling:Gorge Harbor, B.C. exampleGorge Harbor, B.C. example
Oyster Trays(Desolation Sound Oyster Co, Cortes Island, BC)
The flow through different types ofThe flow through different types of
aquaculture rafts and process unitsaquaculture rafts and process units
can be modeled with the samecan be modeled with the same
computational tools.computational tools.
Tray Style Tray Style –– Oyster Raft Oyster Raft Velocity (ft/s)
0.0 1.0 1.5
(a)
(b)Calculated Flow Pattern around Tray Raft
(a)Plan-view colored by speed, (b) Side-view colored by speed(dimensions are in feet, locations of highest and lowest speed are noted)
Representative Fluorescence Data
Diamonds – Approach Flow (~100 ft upstream)Boxes – Upstream Edge of Raft
Triangles – Center of RaftCrosses – Downstream Edge of Raft
Astericks – Departure Flow (~100 ft downstream)
Vertical Variation of ChlorophyllVertical Variation of Chlorophyll
ConcentrationsConcentrations
Maximum concentrations ofMaximum concentrations of
chlorophyll were consistentlychlorophyll were consistently
measured at depths between 6 and 8measured at depths between 6 and 8
meters.meters.
Existing aquaculture rafts onlyExisting aquaculture rafts only
extend to a depth of 4.5 meters.extend to a depth of 4.5 meters.
Profile 9 tray rafts
0
10
20
30
40
0 5 10 15
Depth (m)
Ch
l a
flu
ore
sc
en
ce
Station 1
Station 2
Station 3
Station 4
Station 5
Site VisitSite Visit
Site VisitSite Visit
Velocity MeasurementsVelocity Measurements
Flow speeds within the aquacultureFlow speeds within the aquaculture
rafts are about 10 times less than therafts are about 10 times less than the
flow speeds measured around theflow speeds measured around the
periphery of the rafts.periphery of the rafts.
Representative Velocity Data
Diamonds – Approach Flow (~100 ft upstream)Boxes – Upstream Edge of Raft
Triangles – Center of RaftCrosses – Downstream Edge of Raft
Astericks – Departure Flow (~100 ft downstream)
Angle of Incidence – 45 degrees
Stick rafts profile 2
0
5
10
15
20
25
30
35
40
45
50
0 5 10 15
DEPTH (M)
U C
M S
-1
pro1
pro2
pro3
pro4
pro5
Problem SolvingProblem Solving
Calculated Flow beneath Stick Rafts and Tray Rafts
Orientation & Placement ofOrientation & Placement of
Aquaculture RaftsAquaculture Rafts
Flow accelerates beneath theFlow accelerates beneath the
aquaculture rafts and brings wateraquaculture rafts and brings water
with high concentrations of chlorophyllwith high concentrations of chlorophyll
to the surface in the wake of the rafts.to the surface in the wake of the rafts.
Rows of rafts should be alignedRows of rafts should be aligned
perpendicular to the predominateperpendicular to the predominate
direction of flow and separated by adirection of flow and separated by a
distance of 4 distance of 4 –– 5 raft widths. 5 raft widths.
Downstream wake of chl a depletion for W.Downstream wake of chl a depletion for W.Coast mussel rafts at 2 velocitiesCoast mussel rafts at 2 velocities
5 cm s-1approachvelocity
15 cm s-1approachvelocity
0
5
10
15
20
25
17:30
18:30
19:30
20:30
21:30
22:35
23:35
0:35
1:35
2:35
3:40
4:40
5:40
6:40
7:40
8:45
9:45
Time
Velo
cit
y (
cm
s-1
)
0
0.5
1
1.5
2
2.5
3
3.5
17:30
18:30
19:30
20:30
21:30
22:35
23:35
0:35
1:35
2:35
3:40
4:40
5:40
6:40
7:40
8:45
9:45
De
pth
(m
)
0
50
100
150
200
250
300
350
400
17:30
18:30
19:30
20:30
21:30
22:35
23:35
0:35
1:35
2:35
3:40
4:40
5:40
6:40
7:40
8:45
9:45
Dir
ec
tio
n (
de
gre
es
)
Intertidal Culture of Manila Clams Hydrodynamics and CTDmoorings: Eld clams under nets June 8-9, 2005
Eld clam nets
0
1
2
3
4
5
6
7
8
17:30 19:00 20:30 22:00 23:30 1:00 2:30 4:00 5:30 7:00 8:30 10:00
Time
ch
l a
(y
ell
ow
no
rth
, re
d s
ou
th,
blu
e m
idd
le)
Geoducks Thorndyke June 7, 2005
0
1
2
3
4
5
6
7
8
upstream dow nstream
location
ch
l a
ug
/l
Thorndyke geoducks June 7, 2005 17:30
-1
0
1
2
3
4
5
6
7
8
1 21 41 61 81 101 121 141 161 181 201 221
Time (18 hours)
Ch
l a
(u
g l
-1)
depth (m)
chl a west
Thorndyke ducks
0
5
10
15
20
25
1 95 189 283 377 471 565 659 753 847 941 1035 1129
Time
Cu
rren
t sp
eed
(cm
s-1
)
speed
depth (m)
0
50
100
150
200
250
300
350
400
1 93 185 277 369 461 553 645 737 829 921 1013 1105
dir
ec
tio
n (
de
gre
es
)
Hydrodynamics, CTD moorings and diver samples, Thorndyke geoducks June 7-8, 2005
Washington State: siteWashington State: sitecomparisonscomparisons
0
1
2
3
4
5
6
7
8
Eld bags Eld nets Thorndyke
ducks
Thorndyke
clams
velocity (cm s-1)
chl a (_g l-1)
Percent depletion of chl a vs culture
system of Manila clams: middle vs
outside control
0
5
10
15
20
25
30
35
Eld nets Eld bags Thorndyke Bags
Velocities beneath the buoyed net in the clam bed are about ½ of the mean flow speed(note: the amount of biofouling modeled is the same as that modeled for the dirty bag).
Velocities in the middle of the bag are about 10 times less than the mean flow speed.
Velocities in the middle of the bag are 100 times less than mean flow speeds whenthe bag is fouled.
Water velocity inside Geoduck tunnel and outside Aug. 15-Water velocity inside Geoduck tunnel and outside Aug. 15-16, 2007 Thorndyke Bay, Hood Canal16, 2007 Thorndyke Bay, Hood Canal
Aug. 15-16 Thorndyke Tunnel
-20.00
-15.00
-10.00
-5.00
0.00
5.00
10.00
19:19
20:19
21:19
22:19
23:19
0:19
1:19
2:19
3:19
4:19
5:19
6:19
7:19
8:19
9:19
Cu
rre
nt
sp
ee
d (
cm
s-1
)
Wa
ter
de
pth
(m
)
Aug. 15-16, Thorndyke S4
0
5
10
15
20
25
18:30
19:20
20:10
21:00
21:50
22:40
23:30
0:20
1:10
2:00
2:50
3:40
4:30
5:20
6:10
7:00
7:50
8:40
Cu
rre
nt
sp
ee
d (
cm
s-1
)
Wa
ter
de
pth
(m
)
Aug. 15 - 16, Thorndyke S4
0
50
100
150
200
250
300
350
400
18:30:00
19:15:30
20:05:00
20:50:30
21:40:00
22:25:30
23:15:00
0:00:30
0:50:00
1:35:30
2:25:00
3:10:30
4:00:00
4:45:30
5:35:00
6:20:30
7:10:00
7:55:30
8:45:00Cu
rren
t d
irecti
on
(d
eg
rees)
Cooperative studies withPSI and Joth Davis utilizingestimates of particle fluxand consumption bygeoducks. ONGOING!
Aquaculture GIS applications: STEM GISAquaculture GIS applications: STEM GIS
Sample Screen Capture(bathymetry, flood tide vectors, existing aquaculture and potential run-off
pollution site layers active)
SummarySummary Bivalve shellfish carrying capacityBivalve shellfish carrying capacity
may be estimated bymay be estimated byunderstanding of understanding of food supply andfood supply anddemanddemand, which is site specific., which is site specific.
The primary drivers, The primary drivers, water velocitywater velocityand primary productionand primary production, interact, interactwith the biomass distribution of thewith the biomass distribution of thebivalves to determine growthbivalves to determine growthrates, particle depletion andrates, particle depletion andecological interactions of theecological interactions of theshellfish populations.shellfish populations.
The hydraulic characteristics ofThe hydraulic characteristics ofculture structures and systemsculture structures and systemscan be can be measured, measured, modelledmodelled and andoptimizedoptimized relative to business and relative to business andenvironmental objectives.environmental objectives.
Incorporation of modelling andIncorporation of modelling anddata results in a data results in a GIS formatGIS format allows allowsthe integration of shellfishthe integration of shellfishaquaculture into a multiple useraquaculture into a multiple usercoastal zone management tool.coastal zone management tool.
The authors thank Washington Sea Grant for theinvitation to participate in the conference
Recommended