You are on page 1of 13

Lecture 7.

Ocean forcing
In upcoming lectures we will begin to discuss ocean circulation. The guiding principles will be
conservation of mass, momentum and energy. These conservation principles are crucial because the
ocean is forced non-uniformly, giving rise to transport, advection and fluxes that re-distribute heat,
salt and other tracers. This lecture focuses primarily on the ocean heat budget and will discuss
more briefly the total buoyancy forcing as well as wind forcing.
Ocean convection occurs at locations where density becomes sufficiently large to generate con-
vection. This may happen due to direct cooling and reduced temperatures or through brine rejection
during sea ice formation.

The ocean heat budget


The spatial distribution of sea surface temperature is dominated by incoming solar radiation, but
is also affected by ocean currents. Here we look in detail at the heat budget that quantifies this
balance. Our discussion will be in terms of heat fluxes, Q, which have units of Watts per square
meter, 1 W/m2 = 1 J/(sm2 ). To put this in perspective, a decent hairdryer produces about 1000
W, where as the typical surface heat flux is about a tenth of this per square meter. While this
seems fairly small, total ocean heat transport (depth averaged) is reported in terms of Peta-watts,
or 1012 hairdryers.
The equation governing the surface heat flux is given by
DT 1 ∂Q
=− , (1)
Dt ρcp ∂z
which shows that it is the vertical convergence or divergence of the heat flux that contributes to
changes in temperature.
The ocean heat budget has five major components:
1. Qs is the incoming solar shortwave radiation;

2. Qb is the outgoing, longwave radiation loss to the atmosphere and space from the ocean;

3. Qh is the rate of heat loss or gain due to temperature differences between the ocean and
atmosphere (sensible heat flux);

4. Ql is the latent heat flux due to heat loss/gain due to evaporation/condensation;

5. Qv is the local rate of heat loss/gain due to advection by ocean currents.


We take positive values to represent a heat flux into the ocean (warming) and negative values as
a flux out of the ocean (cooling). Note that this is not a uniformly agreed upon convention. The
total rate of change of temperature for a given water mass is then given by the sum of these five
terms
QT = Qs + Qb + Qh + Ql + Qv . (2)
Recall that the Stefan-Boltzmann Law governs the radiation of energy from a body that is
dependent on the fourth power of its absolute temperature

P = AεσT 4 , (3)

where P is the radiated power, A is area, ε is emissivity and σ is Stefan’s constant. Furthermore the
peak energy varies inversely with the temperature of the body (Wien’s Law ). Thus the sun radiates

1
energy primarily as short waves in the visible wavelength, shorter that 4µm. The ocean absorbs
this energy, converts it to heat, but radiates energy at a lower rate and at longer wavelengths,
between 3 and 80 µm (with a maximum at 10µm).
Of the five heat budget terms above, Qs is typically the only positive (source) value. Due to
absorption and scatter, only 50% of of the solar radiation incident at the top of the atmosphere
reaches the ocean surface. The shortwave radiation flux can be expressed using the following bulk
formula:
Qs = (1 − α) Qc (1 − 0.62C + 0.0019θN ) , (4)
which is useful for providing some insight into what factors affect Qs . First, is the amount of solar
energy that reaches the top of the atmosphere, Qc , which is known as the solar constant. This has
a value of roughly 1370 W m−2 perpendicular to the sun’s rays. Of course, part of the incoming
shortwave radiation is reflected upward from the Earth’s surface; this is the albedo α. The albedo
of water is roughly 0.1, but varies significantly at higher latitudes due to the formation of sea ice.
Albedo is actually very sensitive to the age of the ice and whether the ice is covered with snow.
New ice will have an albedo between 0.1 and 0.2, where as first year ice raises the albedo to 0.6,
increasing to 0.7 in later years. With the addition of snow, albedo can rise as high as 0.9. Cloud
cover is one of the most challenging aspects of the climate system because of its contribution to
albedo. Cloud cover also affects Qs through the term C in (4), as the incoming radiation is reflected,
absorbed or scattered by clouds. Finally, the term θN incorporates the effect of the sun’s elevation.
This is necessary as the sun’s elevation controls the area over which the sun’s rays reach the Earth
and also the distance it must travel through the atmosphere. As the sun’s elevation deviates from
directly overhead (θN = 90o ), Qs decreases because the area over which the radation is distributed
increases and the absorption by the atmosphere is greater. A key property of shortwave radiation
is that it penetrates to depths up to 100 m, known as the euphotic zone. The depth of penetration
depends on the wavelength of the light, the optical properties of the water, relating to its sediment
or biological concentration.
The longwave radiation term Qb actually represents the difference between energy that radiates
away from the sea surface and energy that radiates towards the sea surface by the atmosphere.
Qb is almost always a loss of heat for the ocean. This term is, to leading order, just dependent
on the water temperature, although cloud cover can also significantly impact how much of the
longwave radiation is trapped or lost to space. Unlike shortwave radiation, longwave radiation can
not penetrate very deep into the ocean. Thus the temperature that actually governs the outgoing
longwave radiation is a value known as the skin temperature, which represents the temperature
in the upper millimeter of the sea surface. Note that the skin temperature and the bulk surface
temperature (say, the temperature in the upper 1 m) are rarely the same unless the upper ocean
is very well mixed. Outgoing longwave radiation at the top of the atmosphere (dominated by the
ocean contribution) can be measured from satellites.
The term Ql is largely due to evaporation at the sea surface, which results in both a loss of
volume and a loss of heat. The latent heat flux depends on the rate of evaporation Fe as well as
the latent heat of evaporation L:

Ql = Fe × L = ρCe u∆q × L. (5)

Here L ≈ 2470 kJ/kg, but has a weak temperature dependence, Ce is a transfer coefficient, ∆q
represents the difference between the saturated humidity at the sea surface temperature and the
actual humidity of the atmosphere, and finally there is an important dependence on wind speed u.
Together these values provide a rough rate of evaporation of 1.2 m/yr. There are limited regions

2
Figure 1: Annual mean heat fluxes (W/m2 ) for (a) shortwave heat flux Qs , (b) longwave heat flux
Qb , (c) latent heat flux Ql and (d) sensible heat flux Qh from Grist and Josey (2003).

where atmospheric humidity is sufficiently high to enable condensation into the ocean. The latent
heat flux is typically the largest of the loss terms, but it can not be measured remotely by satellites.
The sensible heat flux Qh arises from a difference in temperature between the atmosphere and
the underlying sea surface and is simply due to conduction. A simple relationship is
dT
Qh = −Ah cp , (6)
dz
where Ah is an eddy conductivity that depends on local wind speeds. The sensible heat flux also
has a dependence on the wind and an approximate formula is often used:

Qh = ρa cp cs u∆T, (7)

where ρa is the density of the air at the sea surface, cs is a bulk transfer coefficient for heat, u is
the air speed (typically taken at 10 meters above the surface) and ∆T is the difference between the
air temperature and the sea surface temperature. This is perhaps the most intuitive component of
the heat balance, but it is also tends to be the smallest component.
Figures 1 and 2 show the annual mean spatial distribution of each of the four components
discussed above, as well as their sum. The major points of these plots are (along with some
questions to discuss in class):

1. Shortwave radiation is the only positive component. Its variation is almost completely in
the meridional direction ranging from 250 W/m2 in the tropics and subtropics to as little
as 50 W/m2 at polar latitudes. What processes might give rise to small zonal variations in
shortwave radiation?

3
Figure 2: Annual mean air-sea buoyancy flux converted to equivalent heat flux (W/m2 ) from Large
and Yeager (2009).

2. Longwave radiation is the most uniform, ∼ 50 W/m2 because it is dependent on absolute


temperature. What negative feedback helps to keep the amplitude of seasonal variations
small?

3. Latent heat shows the most zonal variability, explain why.

4. The sensible heat component is smallest of all the heat budget components. Variations in
sensible heat track variations in latent heat, why might this be the case?

5. Discuss where the ocean gains and loses heat in Figure 2.

6. Is there anything strange about the direction of the heat fluxes, shown by the black arrow in
Figure 2?

Buoyancy fluxes
An expression for the evolution of salinity similar to (1) is

DS ∂E
=S , (8)
Dt ∂z
where E is the turbulent vertical flux of freshwater and at the surface is equal to E − P . We can
define the buoyancy of the fluid (units of acceleration) using the following relationship
σ − σ0
b = −g , (9)
ρ0

4
Figure 3: Annual mean air-sea buoyancy flux converted to equivalent heat flux (W/m2 ) from Large
and Yeager (2009).

where σ0 and ρ0 are reference potential and in situ densities. With (9), the evolution equation for
buoyancy is given by  
Db g α ∂Q ∂E ∂B
=− + ρ0 βS =− . (10)
Dt ρ0 cp ∂z ∂z ∂z
Annual mean air-sea buoyancy fluxes (converted to equivalent heat fluxes (W/m2 )) are shown in
Figure 3.

Meridional heat transport


The observed SST distribution is maintained by heat flux through the sea surface, heat flux through
the base of the mixed layer and horizontal advection. For the latter component, this consists of
a transport of heat from the tropics to polar regions. The atmosphere carries most of the heat
from the equator to the poles, but the ocean also carries a significant portion that should not be
neglected. Remember that the ocean is a huge reservoir of heat, so its distribution is very important
for climate. The heat flux is typically defined across a vertical plane extending across an ocean
basin Z Z θe z=0
H = ρ0 cp a cos φ vT dz, (11)
θw z=−h

where a is the Earth’s radius, θ is longitude and φ is latitude. A useful number to keep in mind
is that the peak heat flux integrated across the globe peaks at ±2 PW (1 petawatt = 1×1015 W)
around ±20o latitude (Figure 4). We will look at the meridional heat flux more closely when we
begin to talk about the ocean’s overturning circulation, but for the moment, it is useful to discuss
the three ways in which the meridional heat flux is measured.

1. The ocean heat transport is calculated as a residual between the total (top of the atmosphere)
transport and the atmospheric meridional heat transport.

2. Through calculation of the heat fluxes at the sea surface, which must be balanced by the
interior meridional heat transport. This provides a zonal average of the oceanic meridional
heat flux.

5
Figure 4: Ocean heat transport for the entire ocean and broken down into the contributions from
the different ocean basins. The data comes from Trenberth and Solomon (1994).

3. Through direct measurements from in situ observations.

As we can see from Figure 4 there is a northward transport of heat of roughly 1 PW in the
Atlantic that peaks around 24o N. The heat transport is a balance of the northward flux of a warm
Gulf Stream, and a southward flux of cooler thermocline and cold North Atlantic Deep Water
that is known as the meridional overturning circulation (MOC). Paleoclimate records suggest that
during the last Ice Age the MOC has undergone abrupt rearrangements that were responsible for
a cooling of European climate of between 5-10o C. Since 2004 the RAPID MOC monitoring array
has been recovered and redeployed annually. Observations have shown that even on subannual
timescales a large variability is found for the MOC. The MOCHA project and its contribution to
the MOCHA-RAPID array provides additional instrumentation to estimate the heat flux associated
with the MOC at 26o N. The heat flux is decomposed into three components: transport through the
Florida Straits, transport related to surface wind forcing (Ekman transport) and interior geostrophic
transport related to cross-basin density (pressure) gradients.

Wind forcing
The surface wind stress is largely responsible for forcing the ocean circulation. The mean wind stress
is characterized by trade winds in the tropics and westerlies poleward of 30o in both hemispheres.
The strongest wind forcing is the westerly wind belt in the southern hemisphere between 40o S and
60o S over the Antarctic Circumpolar Current. Regions of convergence and divergence due to wind
stress curl give rise to Ekman suction or Ekman pumping that stretch fluid columns and generate
interior meridional transport in order to conserve potential vorticity (Figure 5).

6
Figure 5: Annual mean wind stress curl. Ekman pumping (associated with equatorward transport)
is indicated by the blue regions and Ekman suction (associated with poleward transport) is indicated
by the red regions.

Lecture 8. Ocean vertical structure and property distributions


The ocean is an interesting fluid system because it is a three-dimensional body of water that
is forced almost exclusively at a single two-dimensional surface, the air-ocean interface. This
has important implications for how changes in surface forcing are communicated throughout the
ocean. Most importantly, though, it means that ocean properties are predominantly set by surface
dynamics. The process by which the ocean communicates surface properties to the interior is
known as ventilation. We have already discussed that motion in the ocean is predominantly along
isentropic or isopycnal surfaces, yet the ocean is vertically stratified. Thus, the ocean is relatively
inefficient at transferring heat into the interior. This is because there are very few places in the
ocean that generate sufficient density anomalies that allow convection to set in. We will discuss
these unique situations later in the term. However, there are also benefits associated with the fact
that the ocean is only forced at the sea surface. Because the interior flow is largely adiabatic, water
properties are generally useful for identifying flow paths from the surface to the interior. Changes
in properties then identify regions of strong mixing or water mass modification, since there are
very small interior sources and sinks (at least with respect to tracers that are not strongly affected
by biology). In this lecture we will consider vertical and horizontal distributions of temperature,
salinity and density properties.
One of the most important concepts in physical oceanography is that of a water mass, which
refers to a body of water whose properties have been set by a single, identifiable physical process.
Almost all water masses are set by processes at the sea surface, although, identifiable features
of these water masses may also rely on non-conservative tracers, i.e. those that have internal
sources and sinks, due to biological processes, for instance, that acquire their distinctive properties
over time. Water masses are typically distinguished by extrema in one or more properties, such
as temperature, salinity, oxygen, etc. The second key part of water mass identification is also

7
Surface temperature (°C)

Note total range and general distribution of temperature!


Figure 6: Annual mean sea surface temperature.
Talley SIO 210 (2012)!
DPO Figure 4.1: Winter data from Levitus and Boyer (1994)!
determining the physical process that gives rise to these characteristics. In terms of water masses,
it is most convenient to divide the vertical water column into four main layers: the upper layer, the
intermediate layer, the deep layer and the bottom layer. As this division is based on water mass
properties, it is actually quite distinct from how one might choose to divide up a single vertical
property profile. For instance, with a typical temperature profile, much of the structure would be
contained in the upper layer, while, intermediate, deep and bottom layers would all be associated
with the abyssal portion of the profile where there is little variation in temperature. We now
consider a typical temperature profile in more detail, and largely focus on how salinity and density
differ from this.

Temperature
Surface distribution
Water has an albedo of only 0.1 (this should be compared with an Earth average of roughly 0.3 that
is largely determined by cloud cover; the albedo of sea ice ranges from 0.6 to 0.7). Thus the ocean
is very efficient at absorbing solar radiation. It is perhaps not surprising then that contours of sea
surface temperature (SST) are largely zonal (extend in the east-west direction), with warmest values
near the equator and cooler values at the poles (Figure 6). It should be noted that observations of
SST are amongst our most reliable ocean observations. These values are obtained remotely from
satellites based on infrared radiation that is emitted from the ocean surface; these values typically
represent temperature in the upper 0.01 mm, which may be significantly different from the bulk
temperature in the upper meter. New microwave techniques allow us to “see” through clouds,
providing broad spatial coverage in all weather conditions.
The more interesting aspect of the SST distribution is where the contours deviate from zonality.
This zonal variability arises from the effects of ocean circulation and advection. This is most
apparent in the tropics with relatively cool temperatures in the east and the warmest temperatures
being found in the western Pacific. The warm pool is located in the western tropical Pacific and

8
Figure 7: Maximum (winter) mixed layer depth in meters for the world ocean.

represents the location of greatest atmospheric convection. The equatorial cold tongue is the narrow
tongue of colder water along the equator in the eastern equatorial Pacific and Atlantic. This is
a result of a shoaling of the thermocline from west to east, bringing up colder water to the east.
Away from the equator, variations in SST are due to boundary currents (in the west) and coastal
upwelling (in the east). SST is also influenced by the gyre circulations in the ocean basins. Finally,
it is also insightful to compare time-averaged maps of SST to snapshots. These highlight the
prevalence of mesoscale eddy and wave processes in the ocean, particularly in the Southern Ocean,
western boundary current regions and wave-like features along the equator (Tropical Instability
Waves).

Vertical structure
A typical vertical temperature profile can be broken up in to three major zones: (1) the mixed
layer, (2) the thermocline and (3) the abyssal layer, which again, are distinct from the layering
that arises from different water masses. This structure may differ at polar latitudes, where a
subsurface temperature maximum may often develop due to capping of cold winter water overlying
the traditional thermocline and abyssal layers.
The mixed layer refers to a region in the near-surface layer where turbulent processes keep
properties vertically well-mixed. Mixing arises due to the effects of wind and buoyancy forcing.
Cooling and evaporation tend to lead to enhanced mixing, while heating and precipitation act to
stratify the upper layer and reduce the depth of the mixed layer. The mixed layer typically has a
depth of 100 to 150 m at the end of winter, but there is substantial seasonal variability. In isolated
regions of deep convection, the “mixed layer” can extend to 1000 m. Figure 7 shows the distribution
of global winter mixed layer depths. Note that mixed layers are the principal means by which the
atmosphere communicates with the ocean interior.
Radiation entering the ocean surface is absorbed mostly in the top few meters, while heat loss,
which occurs through IR radiation, sensible heat loss to the atmosphere and latent heat loss through

9
evaporation, happens within a few mm of the surface. The turbulent motions within the mixed layer
may entrain cold water from below across the mixed layer base, while wind stress at the surface
drives turbulent motions within the mixed layer that both mix fluid vertically and entrain fluid from
below. Another key process that allows communication between the mixed layer and the ocean
interior is due to the fact that the depth of the mixed layer is spatially non-uniform. Therefore
horizontal currents can cause subduction of mixed layer properties into the ocean interior. Because
the mixed layer represents that body of water that is in direct contact with the ocean surface, and
thus can exchange properties with the atmosphere, understanding how the mixed layer varies with
time is critical for understanding Earth’s climate.
Below the mixed layer, T and S vary rapidly over the upper 500 to 1000 m. This change tends to
be dominated by a particularly strong gradient known as the thermocline, beneath which T maps
on to abyssal properties that vary very little below 1000 m. In most places, the thermocline is
associated with a strong pycnocline since temperature makes the dominant contribution to density.
Throughout low and middle latitudes a thermocline is always present. There are two different
processes that give rise to this permanent thermocline, one arising largely from vertical processes
and another from horizontal processes. In the former case, the ocean is heated at the sea surface
and this heat is transferred slowly through the water column either by molecular diffusion or via
vertical turbulent mixing. This turbulent mixing can be parameterized as an eddy diffusivity. If a
motionless ocean was constantly heated, over time heat would be “diffused” away from the surface,
heating the water column and leading to an increase in the temperature of the abyssal ocean.
This is not what is observed however, because abyssal regions in the tropics and subtropics are
constantly being supplied by cold water from the poles. As this cold water enters at depth, it
displaces resident fluid, leading to a positive vertical velocity. The fact that the large-scale vertical
distribution of temperature is relatively constant implies that these two processes are in balance.
Thus the advection of temperature T by the vertical velocity w is balanced by the flux of heat
acting on a background temperature gradient or
∂T
wT = κeddy . (12)
∂z
The actual value of this vertical velocity is extremely small, typically on the order of 1 cm/day. This
is frustrating because it means that the vertical velocity is too small to measure directly and must
be inferred by other means. If we solve (12) for T , we find that this balance predicts an exponential
decay of temperature with depth, which is actually in remarkably good agreement with observed
profiles of temperature.
The second process that gives rise to the thermocline/pycnocline was first suggested by Iselin
in 1939. He made the interesting observation that the meridional surface temperature and salinity
distribution in the North Atlantic is very similar to the vertical variation of T and S. This led him
to propose that waters in the subtropical thermocline originated from surface waters further north
and had subducted along isopcynals as they were advected equatorward. Subduction of many of
these layers then leads to the formation of a thermocline. Note that this process is distinct from the
one discussed above, as it is a completely adiabatic process (i.e. flow occurs along neutral density
surfaces), whereas the former requires dia-neutral, or diapycnal, mixing to keep the thermocline in
large-scale balance with vertical advection.
Finally, there are some locations where double thermoclines form, which are separated by a
region of weaker vertical temperature gradient or a thermostad. These thermostads are typically
referred to as mode water. Mode water arises from the subduction of thick mixed layers. In the
region where the Mode Water outcrops as a thick mixed layer, the overlying thermocline is actually
a seasonal thermocline that disappears in late winter. After Mode Water subducts, its thermostad

10
is embedded in the permanent thermocline. Mode Water appears as a mode in temperature and
salinity space, they are also typically associated with low values of potential vorticity.
Below the thermocline, temperature slowly decreases with depth, but variability is weak. Po-
tential temperature shows a monotonic decrease with depth, although in situ temperature will
increase due to the effects of pressure. The temperatures at the very bottom of the ocean are set
by two sources, one from around Antarctica and a second due to overflows around the Nordic Seas.
Waters originating from around Antarctica are colder, ranging from just above freezing, while wa-
ter originating from the Atlantic are considerably warmer around 2o C. Roughly 80% of the water
found at the very bottom of the water column originates from around Antarctica.

Salinity
Salinity has a much more complicated horizontal and vertical structure than temperature, largely
because it makes a smaller contribution to density. Since the ocean is largely thermally stratified
throughout the ocean, the vertical structure of salinity may undergo inversions without destabilizing
the water column. Also, whereas the amplitude of solar radiation is largely monotonic with latitude,
evaporation-precipitation and river run-off have more complicated spatial distributions that impact
surface and interior salinity values.
Surface salinity values range between 33 and 37, with low and high values dominated by river-
run off and evaporation respectively. The surface distribution is dominated by salinity maxima in
the subtropics with minima in the tropics and subpolar regions. The meridional salinity maxima
are in the trade wind regions and subtropical high pressure regions where the annual E − P (evapo-
ration minus precipitation) value is positive. Salinity minima are associated with the Intertropical
Convergence Zone (ITCZ). Surface salinity is higher in the Atlantic than the Pacific making water
masses in this ocean basin more susceptible to convection.
In the subtropics, salinity is high throughout the upper few hundred meters, largely due to
evaporation. Below this salinity will decrease to a minimum between 600 and 1000 m, at which
point it begins to increase again towards a maximum at 2000 to 2500 m depth. Thus there are
typically up to four distinct water masses associated with the salinity distribution. In the tropics,
the surface water is fresh due to precipitation, but just below this layer there is a salinity maximum
due to subducted waters from the subtropics being advected towards the equator. Lower salinity
further down in the water column may also arise from water that has been subducted from further
north in the subtropical gyre. This is known as Subarctic Intermediate Water. In subpolar and
polar regions salinity tends to be low at the surface with a rapid increase in salinity known as a
halocline.
At intermediate depths, there are multiple vertically broad layers of either low or high salinity.
Most intermediate water masses originate at the sea surface at subpolar latitudes where surface
waters are relatively fresh. The major low salinity intermediate water layers are North Pacific Inter-
mediate Water (NPIW), Labrador Sea Water (LSW) and Antarctic Intermediate Water (AAIW).
Ventilated intermediate waters spread towards the equator bringing their low salinity signature
with them. There are two high salinity intermediate water layers associated with outflow from
the Mediterranean and Red Seas, where evaporation is particularly high. Within these seas, high
salinity water goes into the marginal seas where salinity is enhanced through evaporation and the
water is also cooled. This leads to dense water formation and convection, such that when the water
flows back into the basins is equilibrates at intermediate depths.
Finally, deep waters also exhibit salinity variations depending on their sources. Deep water
originating in the north Atlantic are associated with a high salinity signature, a water mass known
as North Atlantic Deep Water (NADW). Dense Antarctic waters are fresher than their northern

11
hemisphere counterparts and can be tracked by their relatively fresh signature in Antarctic Bottom
Water (AABW). This difference is seen most clearly in a vertical section of salinity from the
Atlantic. Note that in both temperature and salinity, deep waters tend to be relatively uniform.
This is a result of the small number of dense water sources as well as the long residence times in
the ocean during which these water masses are subjected to mixing and diffusion.
Potential temperature-salinity-volume diagrams are a nice way of quickly visualizing differences
in T /S properties in the different basins (Figure 8). The fingers extending from left to right in the
diagram indicate the north Pacific (B), the north Atlantic (F) and the three southern hemishere
basins (C-E). The latter here indicates the importance of the Southern Ocean in modifying water
mass properties. Of the deep waters, the largest peak by far is Pacific Deep Water. This is the
oldest water mass and due to the influences of mixing is represented as a single peak in T /S space.
Ridges extending away from this peak indicate different source waters in the Atlantic and around
Antarctica that have not been homogenized by mixing.

Density
Since the water column must be stably stratified, potential density must increase with depth and
therefore is not allowed to show the same structure and inversions seen in salinity profiles. Inversions
in potential density are often seen in individual profiles from CTD casts, but these are recording
transient overturning events often arising from internal wave breaking.

12
Figure 8: Potential temperature, salinity, volume distribution for (a) the whole ocean and (b)
waters with temperatures less than 4o C. (Worthington 1981).

13

You might also like