ArticlePDF Available

Abstract and Figures

Agrivoltaic systems are mixed systems that associate, on the same land area at the same time, food crops and solar photovoltaic panels (PVPs). The aim of the present study is to assess whether the growth rate of crops is affected in the specific shade of PVPs. Changes in air, ground and crop temperature can be suspected due to the reduction of incident radiation below the photovoltaic shelter. Soil temperature (5 cm and 25 cm depth), air temperature and humidity, wind speed as well as incident radiations were recorded at hourly time steps in the full sun treatment and in two agrivoltaic systems with different densities of PVPs during three weather seasons (winter, spring and summer). In addition, crop temperatures were monitored on short cycle crops (lettuce and cucumber) and a long cycle crop (durum wheat). The number of leaves was also assessed periodically on the vegetable crops.
Content may be subject to copyright.
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/authorsrights
Author's personal copy
Agricultural
and
Forest
Meteorology
177 (2013) 117–
132
Contents
lists
available
at
SciVerse
ScienceDirect
Agricultural
and
Forest
Meteorology
jou
rn
al
hom
ep
age:
www.elsevier.com/locate/agrformet
Microclimate
under
agrivoltaic
systems:
Is
crop
growth
rate
affected
in
the
partial
shade
of
solar
panels?
H.
Marroua,b,,
L.
Guilionic,
L.
Dufoura,
C.
Dupraza,
J.
Weryd
aINRA,
UMR
SYSTEM,
2
place
Viala,
34060
Montpellier,
France
bSun’R,
7,
rue
de
Clichy,
75009
Paris,
France
cSupAgro,
Département
Biologie
Ecologie,
2
place
Viala,
34060
Montpellier,
France
dSupAgro,
UMR
SYSTEM,
2
place
Viala,
34060
Montpellier,
France
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
14
November
2012
Received
in
revised
form
9
April
2013
Accepted
17
April
2013
Keywords:
Agrivoltaic
Temperature
Light
Lettuce
Cucumber
Wheat
a
b
s
t
r
a
c
t
Agrivoltaic
systems
are
mixed
systems
that
associate,
on
the
same
land
area
at
the
same
time,
food
crops
and
solar
photovoltaic
panels
(PVPs).
The
aim
of
the
present
study
is
to
assess
whether
the
growth
rate
of
crops
is
affected
in
the
specific
shade
of
PVPs.
Changes
in
air,
ground
and
crop
temperature
can
be
suspected
due
to
the
reduction
of
incident
radiation
below
the
photovoltaic
shelter.
Soil
temperature
(5
cm
and
25
cm
depth),
air
temperature
and
humidity,
wind
speed
as
well
as
incident
radiations
were
recorded
at
hourly
time
steps
in
the
full
sun
treatment
and
in
two
agrivoltaic
systems
with
different
den-
sities
of
PVPs
during
three
weather
seasons
(winter,
spring
and
summer).
In
addition,
crop
temperatures
were
monitored
on
short
cycle
crops
(lettuce
and
cucumber)
and
a
long
cycle
crop
(durum
wheat).
The
number
of
leaves
was
also
assessed
periodically
on
the
vegetable
crops.
Mean
daily
air
temperature
and
humidity
were
similar
in
the
full
sun
treatments
and
in
the
shaded
situ-
ations,
whatever
the
climatic
season.
On
the
contrary,
mean
daily
soil
temperature
significantly
decreased
below
the
PVPs
compared
to
the
full
sun
treatment.
The
hourly
pattern
of
crop
temperature
during
day-
time
(24
h)
was
affected
in
the
shade.
In
this
experiment,
the
ratio
between
crop
temperature
and
incident
radiation
was
higher
below
the
PVPs
in
the
morning.
This
could
be
due
to
a
reduction
of
sensible
heat
losses
by
the
plants
(absence
of
dew
deposit
in
the
early
morning
or
reduced
transpiration)
in
the
shade
compared
to
the
full
sun
treatment.
However,
mean
daily
crop
temperature
was
found
not
to
change
significantly
in
the
shade
and
the
growth
rate
was
similar
in
all
the
treatments.
Significant
differences
in
the
leaf
emission
rate
were
measured
only
during
the
juvenile
phase
(three
weeks
after
planting)
in
lettuces
and
cucumbers
and
could
result
from
changes
in
soil
temperature.
As
a
conclusion,
this
study
suggests
that
little
adaptations
in
cropping
practices
should
be
required
to
switch
from
an
open
cropping
to
an
agrivoltaic
cropping
system
and
attention
should
mostly
be
focused
on
mitigating
light
reduction
and
on
selection
of
plants
with
a
maximal
radiation
use
efficiency
in
these
conditions
of
fluctuating
shade.
© 2013 Elsevier B.V. All rights reserved.
1.
Introduction
Agrivoltaic
systems
(AVS)
were
defined
by
Dupraz
et
al.
(2010)
as
“mixed
systems
associating
solar
panels
and
crop
at
the
same
time
on
the
same
land
area”.
They
may
contribute
to
conciliate
food
security
and
green
energy
supply.
In
these
mixed
production
sys-
tems,
photovoltaic
panels
(PVPs)
partially
shelter
the
crop
growing
below.
PVPs
create
intermittent
shading
and
reduce
the
average
available
light
for
the
crop.
Marrou
et
al.
(2013)
showed
that
light
reduction
had
a
significant
impact
on
final
crop
yield
of
spring
and
Corresponding
author
at:
INRA,
UMR
SYSTEM,
2
place
Viala,
34060
Montpellier,
France.
Tel.:
+33
4
99
61
26
84;
fax:
+33
4
99
61
30
34.
E-mail
address:
marrou@supagro.inra.fr
(H.
Marrou).
summer
lettuces
in
AVS.
Biomass
accumulation
was
mainly
driven
by
the
capture
of
light
resource
while
other
resources
such
as
water
and
nitrogen
were
not
limiting.
By
using
a
light
driven
prediction
of
plant
biomass
accumulation
(Monteith,
1977),
it
was
showed
that
high
crop
productivities
can
be
expected
from
these
dual
pur-
pose
systems
(food
and
electricity)
(Marrou
et
al.,
2013).
Marrou
et
al.
(2013)
found
that
light
reduction
was
not
necessarily
detri-
mental
for
crop
production.
Indeed,
an
experiment
conducted
on
spring
and
summer
lettuces
in
AVS
showed
that
lettuce
yield
was
maintained,
despite
shading,
by
an
improved
radiation
intercep-
tion
efficiency
(RIE)
in
the
shade.
Enhanced
RIE
was
explained
by
an
increase
in
total
leaf
area
per
plant
despite
a
decrease
in
the
number
of
leaves.
The
aim
of
the
present
study
is
to
determine
if
other
climatic
variables
are
significantly
modified
in
the
shade
of
solar
panels
and
0168-1923/$
see
front
matter ©
2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.agrformet.2013.04.012
Author's personal copy
118 H.
Marrou
et
al.
/
Agricultural
and
Forest
Meteorology
177 (2013) 117–
132
Fig.
1.
General
map
of
the
experimental
device
during
the
wheat/lettuce
season.
In
the
bottom
right
frame,
a
zoom
on
the
two
central
crop
strips
during
wheat/lettuce
season
and
cucumber
season
is
represented.
Closed
circles
represent
radiation
sensors,
located
on
North-to-South
transects;
closed
square
symbols
represent
thermocouples
or
microthermistances.
FD
represents
the
shaded
plot
with
PVPs
at
full
density
whereas
HD
represents
the
shaded
plot
with
PVPs
at
half
density.
W-C
and
E-C
stand
respectively
for
Western
control
plot
and
Eastern
control
plot.
to
what
extent
this
could
affect
crop
temperature
and
plant
devel-
opment
rates
in
AVS.
Former
work
on
shaded
glasshouses
(Baille
et
al.,
2001;
Kittas
et
al.,
2003)
or
agroforestry
systems
(Lott
et
al.,
2009;
Monteith
et
al.,
1991)
suggested
that
air
temperature
and
vapor
pressure
deficit
at
crop
level
are
reduced
by
shading.
Effects
of
shading
on
crop
temperature
and
growth
rate
were
also
reported.
Lott
et
al.
(2009)
showed
that,
under
50%
of
available
radiation,
the
meristem
temperature
of
shaded
maize
plants
was
reduced
by
2–9 C
(depending
on
the
climatic
seasons).
At
the
same
time,
they
noticed
a
significant
delay
in
flowering
date.
Moreover,
in
the
case
of
AVS,
the
shade
pattern
under
PVPs
varies
from
one
season
to
another
and
among
different
latitudes
as
the
limits
between
shade
and
light
move
with
sun
elevation.
For
this
reason,
the
effect
of
the
PVP
shelter
may
have
a
different
impact
on
the
productivity
of
winter
crops
compared
to
summer
crops,
or
on
short
cycle
crops
compared
to
long
cycle
crops.
We
therefore
analyze
in
this
study
the
impact
of
PVPs
on
the
microclimate
at
crop
level
(air
temper-
ature,
air
vapour
pressure
deficit
(VPD)
and
wind)
as
well
as
crop
and
soil
temperature.
The
experiment
was
conducted
on
three
crop
species
and
cropping
seasons:
durum
wheat
(winter
to
summer),
lettuces
(spring
and
summer),
and
cucumbers
(summer).
2.
Materials
and
methods
2.1.
Experimental
device
Data
were
collected
on
an
experimental
prototype
of
agrivoltaic
systems,
in
Montpellier,
France
(43.15N;
3.87E)
from
July
2010
to
September
2011
(Marrou,
2012;
Marrou
et
al.,
2013).
In
this
exper-
imental
prototype,
photovoltaic
panels
(PVPs)
were
arranged
in
East–West
orientated
strips,
0.8
wide
and
inclined
southward
with
a
tilt
angle
of
25.
PVPs
were
hold
at
4
m
above-ground
by
wooden
pillars
spaced
on
a
6.4
m
×
6.4
m
grid,
in
order
to
allow
mechanical
cropping
of
the
plants
below,
using
tractors
(Fig.
1).
The
experimental
design
enabled
to
compare
three
shading
intensities,
corresponding
to
two
densities
of
solar
panels
and
a
full
sun
control
(FS):
(1)
the
full
density
treatment
(FD),
which
corresponds
to
the
PVPs
density
optimized
for
electricity
produc-
tion.
In
this
treatment
the
distance
between
2
strips
of
PVPs
is
1.6
m,
which
lets
an
average
of
50%
of
the
incident
radiation
to
the
crop,
(2)
the
half
density
treatment,
which
is
obtained
from
FD
by
removing
one
strip
of
PVPs
out
of
two
(distance
between
2
PVPs
strips:
3.2
m)
and
thus
letting
through
an
average
of
70%
of
incident
radiation
available
to
the
crop,
and
(3)
the
full
sun
control
plot
(100%
of
incident
radiation
available)
(Marrou,
2012).
Shading
treatments
were
applied
on
four
land
plots.
They
were
aligned
from
East
to
West
as
following:
eastern
full
sun
control
plot
(E-
FS),
FD
plot,
HD
plot,
and
western
full
sun
control
plot
(W-FS).
Each
plot
is
12
m
wide
in
the
East–West
direction
and
separated
from
each
other
by
a
buffer
zone
of
8
m
(between
control
plots
and
shaded
plots)
or
6
m
(between
FD
an
HD
plots).
In
the
North
South
direction,
each
control
plot
is
19
m
long.
In
FD
and
HD
plots,
plant
and
meteorological
measurements
were
taken
only
in
the
first
15
m
in
the
Northern
part
of
the
plots
to
avoid
a
border
effect
due
to
the
higher
incident
light
in
the
southern
side
of
the
proto-
type.
The
total
area
covered
by
PVPs
(including
FD
and
HD
plots
as
well
as
the
surrounding
buffer
zones)
is
860
m2(19.2
m
×
44.8
m)
and
corresponds
to
the
minimum
standard
size
of
an
agrivoltaic
system
adapted
for
vegetable
production.
Due
to
the
large
size
of
any
agrivoltaic
system
required
to
avoid
unwanted
border
effects
and
the
cost
of
work
and
material
to
build
an
experimental
agri-
voltaic
system,
it
was
not
possible
to
replicate
FD
and
HD
plots
and
design
a
complete
randomized
experimental
device.
However,
as
E-FS
and
W-FS
were
the
two
most
distant
plots
in
the
experi-
mental
field,
they
controlled
the
environmental
variability
in
the
East–West
direction.
Moreover,
environmental
variability
in
the
North–South
was
assessed
during
the
first
experimental
season
(2010)
and
controlled
by
subdividing
both
control
and
shaded
plots
into
three
blocks
in
the
North–South.
Variance
analysis
showed
that
Author's personal copy
H.
Marrou
et
al.
/
Agricultural
and
Forest
Meteorology
177 (2013) 117–
132 119
there
was
no
significant
difference
in
lettuces’
dry
matter,
neither
between
blocks
within
the
same
full
sun
plot,
nor
between
the
W-FS
and
the
E-FS.
Identical
soil
type
(loamy
clayish
deep
alluvial
soil)
and
field
history
(10
years
of
homogenous
non
tillage
cropping)
also
contributed
to
homogenous
soil
conditions
between
treatment
plots.
Moreover,
the
homogeneity
of
soil
properties
was
checked
across
the
treatment
plots
using
soil
hygrometry
measurements
on
samples
evenly
collected
over
the
area
of
the
experimental
field
(Marrou,
2012).
2.2.
Crop
management
Three
different
species
were
tested
on
the
prototype
for
various
cropping
seasons
spanning
from
July
2010
to
September
2011:
let-
tuces
Lactuca
sativa
spp.
(two
cropping
cycles,
five
subspecies),
durum
wheat
(Triticum
durum
L.),
and
cucumbers
(Cucumis
sativus
L.).
Thus,
a
wide
range
of
growing
conditions
in
terms
of
tempera-
ture,
VPD,
and
sun
elevation
was
covered.
Lettuces
were
planted
during
one
summer
and
one
spring
sea-
son.
Firstly,
lettuces
were
grown
from
July
21,
2010
(day
of
the
year,
DOY
202)
to
September
6,
2010
(DOY
249).
Two
varieties
were
then
planted
at
the
same
time:
one
variety
of
crisphead
lettuce,
called
“Kiribati”
(noted
FC+),
and
one
butterhead
lettuce,
called
“Tour-
billon”
(noted
B0).
Secondly,
lettuces
were
planted
on
March
22,
2011
(DOY
81)
and
harvested
on
May
24,
2011
(DOY
144).
For
this
second
cropping
season,
two
varieties
of
crisphead
lettuces
(FC+,
and
a
second
one
called
“Bassoon”
and
noted
FC)
and
two
new
varieties
of
butterhead
lettuces
(variety
“Model”
noted
B+
and
vari-
ety
“Emocion”
noted
B)
were
tested
simultaneously.
Cucumbers
(variety
“Marketmore”)
were
planted
on
June
25,
2011
(DOY
178)
and
fruits
were
picked
from
August
8,
2011
(DOY
220)
to
August
31,
2011
(DOY
239),
twice
a
week.
Durum
wheat
(variety
“Clau-
dio”)
was
sown
at
a
density
of
150
kg/ha
on
November
26
(DOY
331),
2010,
and
harvested
at
maturity
on
June
17,
2011
(DOY
168)
(Fig.
2).
Lettuces
and
cucumbers
were
planted
in
lines.
Planting
rows
were
parallel
to
PVP
strips.
The
distance
between
two
planting
rows
was
0.33
m
for
lettuces
and
3
m
for
cucumbers.
To
allow
simulta-
neous
cultivation
of
wheat
and
vegetable
crops
in
2011,
the
entire
experimental
field
was
divided
into
3.2
m
wide
block
strips
in
the
North–South
direction.
Blocks
were
dedicated
alternatively
to
the
cultivation
of
cereals
or
vegetables.
In
spring
2011,
three
blocks
of
wheat
were
intercalated
with
two
strips
of
6
rows
of
lettuces.
Each
block
of
lettuces
was
replaced
by
one
row
of
cucumber
in
June
2011.
Vegetables
were
irrigated
with
sprinklers
(summer
2010)
or
drip
lines
(spring
and
summer
2010),
in
the
day-time.
Irrigation
was
monitored
with
tensiometers
(SDEC,
Reignac
sur
Indre,
France).
In
order
to
avoid
plant
water
stress,
soil
water
potential
at
0.3
m
depth
was
kept
above
0.02
MPa
(Gay,
2002).
2.3.
Plant
development
rate
Development
rate
of
vegetable
crops
was
assessed
by
count-
ing
the
number
of
leaves,
over
1
cm
long,
that
were
emitted
every
3
weeks
for
lettuces
and
twice
a
week
for
cucumbers.
For
lettuces,
the
number
of
leaves
per
plant
was
assessed
through
destructive
samp-
ling
at
three
dates
(DOY
223,
236,
and
249
in
summer
2010,
and
DOY
104,
125,
and
144
in
spring
2011.
12–15
plants
per
treatment
and
per
variety
were
collected
at
each
sampling
date.
Samples
were
stratified
to
warrant
that
each
sample
contained
the
same
num-
ber
of
plants
collected
from
each
planting
rank,
and
to
explore
the
intra-treatment
variability.
Leaf
emission
rates
()
were
calculated
by
fitting
linear
models
between
the
number
of
leaves
measured
on
each
sampled
lettuce
plant
and
the
thermal
time
(calculated
from
air
temperature
at
2
m
above-ground
with
a
base
tempera-
ture
of
3C,
Thicoïpé,
1997)
for
each
period
between
two
sampling
Fig.
2.
Hourly
(a,
for
DOY
128)
and
daily
(b,
from
DOY
115
to
DOY
230)
incident
radiation
in
the
FS
treatment
(white
background
boxes),
in
HD
(gray
background
boxes)
and
FD
(dark
gray
background
boxes).
The
boxes
feature
the
spatial
variability
of
the
incident
radiation
(radiation
was
recorded
at
the
same
time
by
sensors
settled
at
different
locations
on
the
North–South
axis).
dates.
Analysis
of
covariance
(ANCOVA),
using
shading
treatment
as
a
factor,
was
performed
to
determine
whether
a
single
linear
model
could
be
fit
for
all
the
treatments.
If
not,
development
rate
were
considered
as
significantly
different
between
treatments,
for
the
corresponding
time
period.
Concerning
cucumbers,
the
number
of
leaves
on
the
main
stem
and
on
each
secondary
branch
was
counted
in
the
field
twice
a
week
on
16
plants.
These
plants
had
been
randomly
chosen
at
plant-
ing
date
(3
plants
both
in
E–C
and
in
W–C
and
5
plants
in
FD
and
in
HD).
After
plotting
the
dynamic
of
the
total
number
of
leaves
per
plant
over
the
entire
cropping
period,
three
time
periods
with
nearly
constant
were
identified,
and
a
mean
was
calculated
in
the
three
treatments
for
each
period.
To
do
so,
we
applied
the
same
methodology
as
for
lettuces:
linear
models
were
fitted
to
predict
the
number
of
leaves
as
a
function
of
thermal
time
(calculated
from
air
temperature
at
2
m
above-ground
with
a
base
temperature
of
15 C,
Perry
et
al.,
1986)
for
each
time
period.
Shading
effect
on
was
tested
with
ANCOVAs.
For
wheat,
phenological
stages
such
as
tillering,
flowering,
and
maturity
were
determined
according
to
the
Zadoks
scale
(Zadoks
et
al.,
1974)
in
each
plot.
A
given
phenological
stage
was
reported
as
attained
when
50%
of
the
plant
population
had
reached
this
stage
(visual
assessment).
Author's personal copy
120 H.
Marrou
et
al.
/
Agricultural
and
Forest
Meteorology
177 (2013) 117–
132
2.4.
Microclimate
monitoring
2.4.1.
Air
temperature,
air
humidity
and
wind
speed
at
standard
height
Air
temperature
(±0.2 C)
and
air
relative
humidity
(±2%)
at
2
m
above-ground
were
measured
respectively
with
a
capacitive
thermohygrometer
(HMP
45,
Vaisala,
Helsinki,
Finland)
placed
in
radiation
screens
with
natural
ventilation
(MET20,
Campbell
Sci-
entific,
Inc.,
USA)
and
connected
to
dataloggers
(CR1000
Campbell
Scientific,
Inc.,
USA).
Three
probes
were
set
in
the
middle
of
each
treatment
plot
from
DOY
222
to
249
in
2010
(for
second
half
of
the
summer
season
for
lettuces)
and
from
DOY
36
to
239
in
2011
(for
the
complete
cycle
of
summer
lettuces
and
cucumber,
and
from
“3
leaves”
stage
to
maturity
for
the
wheat
cycle).
Wind
speed
(u,
m
s1)
was
measured
at
2
m
above-ground
with
a
mechanical
wind
moni-
tor
(05103-5,
Young,
Traverse
City,
MI,
USA),
in
the
E-FS
plot
and
in
the
FD
treatment
from
DOY
36
to
239
in
2011.
All
these
data
were
collected
every
5
s
on
a
datalogger
(CR1000,
Campbell
Scientific
Inc.,
Logan,
UT,
USA)
and
averaged
over
1
h.
Analyses
of
variance
(ANOVA)
and
least
significant
difference
(LSD)
tests
were
performed
successively
on
TAand
VPD,
using
the
shading
level
as
treatment
to
determine
the
effect
of
the
shading
treatments
on
mean
daily
air
temperature
and
VPD.
These
statisti-
cal
analyses
were
done
over
the
whole
the
measurement
period
at
a
daily
(to
compare
mean
daily
values)
or
hourly
(to
compare
daily
microclimatic
patterns)
time-step.
2.4.2.
Soil
temperature
Soil
temperature
(±0.4 C)
was
measured
below
the
wheat
crop
in
2011
(from
DOY
36
to
165)
at
0.05
m
(TS5)
and
0.25
m
(TS25)
below
the
ground
surface
with
thermistors
(107
thermistors,
Campbell
Scientific,
Inc.,
USA)
at
2
positions
along
a
North
South
transect
between
two
strips
of
solar
panels.
Soil
temperature
at
0.25
m
depth
was
also
measured
in
irrigated
soil
during
the
crop
cycle
of
summer
lettuce
(from
DOY
220)
until
after
the
harvest
of
the
lettuces
(DOY
279).
6
probes
were
dug
both
in
the
FD
and
in
the
HD
plots,
at
different
positions
on
North
South
transects,
and
3
probes
were
dug
in
the
control
plots.
All
probes
were
connected
to
dataloggers
(CR1000,
Campbell
Scientific,
Inc.,
USA),
and
data
were
recorded
with
a
time
step
of
5
s,
and
then
averaged
over
1
h
for
storage
in
the
datalogger
memory.
Soil
temperature
was
not
collected
during
the
spring
lettuce
cycle,
because
it
would
have
required
a
higher
number
of
probes
as
the
crop
was
irrigated
with
drip
lines.
2.4.3.
Crop
temperature
Crop
temperatures
(TL)
were
measured
on
spring
lettuces
from
DOY
117
to
149
in
2011,
with
copper-constantan
thermocouples,
inserted
between
the
lettuces
leaves,
close
to
the
central
axis
of
the
plant.
Leaf
temperatures
were
measured
on
cucumber
(DOY
178–244)
and
durum
wheat
(DOY
36–168)
with
microthermis-
tors
taped
on
the
bottom
side
of
the
leaves
(Thorpe
and
Butler,
1977).
Thermistors
were
moved
from
one
leaf
to
another
during
the
crop
cycle
so
that
they
would
measure
the
temperature
of
a
non-
senescent
leaf.
We
chose
leaves
that
were
located
in
the
middle
of
the
main
stem
for
cucumber,
or
at
mid-height
of
the
plant
cover
for
wheat.
Measurements
from
thermocouples
and
microthermis-
tors
were
recorded
on
dataloggers
(CR10X
and
CR1000,
Campbell
Scientific,
Inc.,
USA)
with
the
same
time
steps
as
mentioned
above.
The
precision
of
the
data
was
estimated
at
±0.4 C
in
the
worst
case.
2.4.4.
Incident
radiation
2.4.4.1.
Incident
global
radiation
received
on
an
horizontal
surface
at
crop
height.
Incident
solar
radiation
(Rs,
W
m2)
was
measured
during
the
wheat
and
cucumber
cycles
with
pyranometers
(SKS
1110,
Skye
Instruments,
Powys,
UK)
set
horizontally
at
the
height
of
the
crop.
For
lettuces,
photosynthetically
active
radiation
(Rp,
W
m2)
was
measured
with
PAR
sensors
(spring
and
summer).
Rs
and
Rpmeasurements
were
treated
equally
after
conversion,
as
Rp/Rsis
constant,
equal
to
0.48
in
outdoors
conditions,
provided
all
measurements
are
taken
above
plant
foliage.
In
the
FD
and
the
HD
treatments,
several
sensors
were
set
along
North–South
transects,
perpendicularly
to
the
PVP
strips.
We
thus
captured
the
hourly
and
daily
intra-treatment
variations
below
the
panels.
One
sensor
acquired
data
above
each
row
of
lettuce,
while
5
sensors
were
regularly
placed
(spacing
of
40
cm)
on
a
diagonal
transect
centered
on
the
cucumber
rows.
As
for
the
wheat
experi-
ments,
two
pairs
of
sensors
were
set
on
a
transect
perpendicular
to
the
crop
strip:
one
was
just
below
a
PVP
strip
and
the
other
was
in-
between
two
PVP
strips.
All
sensors
were
connected
to
dataloggers
(CR1000
Campbell
Scientific,
Inc.,
USA),
and
data
were
recorded
with
a
time
step
of
5
s,
and
then
averaged
over
1
h.
2.4.4.2.
Spatial
distribution
of
the
incident
beam
rays.
The
spatial
distribution
of
the
incident
radiation
was
assessed
through
mod-
eling
and
field
measurements
in
the
FS
treatment
as
well
as
in
the
two
shaded
treatments
(FD
and
HD)
in
order
to
test
whether
there
were
significant
scattering
effects
below
the
PVPs
and
to
charac-
terize
the
proportion
of
diffuse
and
direct
radiation
in
the
different
treatments.
2.4.4.2.1.
Field
measurements.
Complementary
measurements
were
conducted
in
spring
2012,
from
DOY
124
to
134
to
quantify
incoming
radiation
from
different
directions,
using
a
turtle
PAR
sensor.
This
device
has
been
described
by
Chenu
et
al.
(2008).
It
is
made
of
six
faces
with
equal
solid
angles,
so
that
the
entire
sky
hemisphere
is
covered
without
overlapping.
Each
face
of
the
tur-
tle
sensor
is
a
pentagonal
PAR
sensor.
Face
1
was
horizontal,
while
faces
2–6
were
inclined
with
a
tilt
angle
of
63.4,
and
directed
to
5
different
azimuth
directions.
Face
4
was
oriented
northwards.
The
directional
measurement
of
incident
radiation
was
repeated
for
dif-
ferent
locations
within
the
FD
and
HD
plots:
the
sensor
was
set
for
2
days
of
acquisition
at
each
location
where
the
PAR
sensors
had
been
placed
in
2011
above
each
lettuce
planting
row.
The
sensors
were
connected
to
a
datalogger
(CR10X
Campbell
Scientific,
Inc.,
USA),
and
data
were
recorded
with
a
time
step
of
5
s,
then
averaged
over
10
min.
2.4.4.2.2.
Simulation
of
the
light
captured
by
an
inclined
surface.
The
spatial
distribution
of
incoming
radiation
measured
below
the
PVPs
needed
to
be
compared
that
in
the
full
sun
conditions
(FS).
To
do
so,
the
solar
energy
that
would
have
been
captured
by
each
face
of
the
turtle
PAR
sensor
in
full
sun
conditions
(FS)
for
the
same
days
of
the
year
was
simulated.
The
spatial
distribution
of
solar
energy,
in
the
absence
of
sheltering,
abides
to
astrological
laws
and
can
be
pre-
dicted
with
precision
according
to
existing
models
(Liu
and
Jordan,
1960;
Spitters
et
al.,
1986;
Allen
et
al.,
1998;
Bindi
et
al.,
1992;
Anderson,
1966).
A
model
was
coded,
and
implemented
with
the
R-cran
software
(http://cran.r-project.org/),
to
simulate
the
energy
captured
by
each
face
of
the
turtle
sensor
in
FS,
with
a
time
step
of
10
min,
for
a
given
day
of
the
year
(DOY).
Model
algorithm
uses
astrology
equations
currently
in
use
in
existing
astrological
models
(Marrou
et
al.,
2013),
with
a
few
adaptations.
Firstly,
extraterres-
trial
radiation
was
calculated
with
the
De
Jong
formula
(Bindi
et
al.,
1992),
which
is
more
suitable
at
infra-daily
time
steps.
Secondly,
the
calculation
step
integrating
the
incident
radiation
over
one
day
was
removed.
Inputs
of
the
model
were
latitude
of
the
site,
orien-
tation
of
the
sensor
and
incident
global
radiation
measured
on
a
horizontal
surface
at
the
model
time
step.
Radiation
data
were
col-
lected
at
an
hourly
time
step
from
a
weather
station
located
400
m
from
the
experimental
field
(INRA,
Lavalette
weather
station).
Lin-
ear
interpolation
was
performed
between
hourly
data
in
order
to
get
a
dataset
with
a
time
step
of
10
min
that
can
be
used
as
an
input
for
the
radiation
model.
Model
quality
was
verified
by
comparing
Author's personal copy
H.
Marrou
et
al.
/
Agricultural
and
Forest
Meteorology
177 (2013) 117–
132 121
Fig.
3.
Mean
daily
temperature
(TA)
measured
at
2
m
above-ground
during
every
cropping
cycles
from
July
2010
to
August
2011.
Zooms
on
6-day
periods
(6
days
in
winter
and
6
days
in
summer)
are
provided.
FS
is
represented
with
shaded
intervals
on
the
main
graph
(95%
confidence
interval
for
TAmeasured
in
FS
with
three
different
probes)
and
with
open
circles
()
in
the
zoom
areas,
while
FD
and
HD
are
represented
respectively
with
open
triangles
()
and
closed
squares
().
In
the
zoom
areas,
vertical
error
bars
feature
standard
errors
for
all
the
treatments.
simulation
with
measurements
taken
in
FS
from
DOY
196
to
199
in
2012.
A
linear
model
was
fitted
for
each
sensor
face
with
a
cor-
rection
coefficient
higher
than
0.94,
and
the
slope
coefficient
was
always
between
0.88
and
1.
It
was
concluded
that
the
model
was
a
good
predictor
and
that
simulated
data
in
FS
needed
no
correction
coefficient
to
match
measured
values.
More
details
on
the
model
specifications
and
validation
are
given
in
Appendix
A.
2.4.4.3.
Long
wave
radiations.
In
spring
2012
(DOY
174–202)
upwards
(R)
and
downwards
(R)
longwave
radiation,
and
upward
(Rs)
and
downward
(Rs)
short
wave
radiation
were
measured
for
each
treatments.
The
four
types
of
radiations
were
measured
successively
in
the
FD,
HD,
and
FS
treatments
with
two
net
radiometers
(Q-7.1,
Campbell
Sci,
USA)
and
2
pyranometers
(SKS
11110,
Skye
Instruments,
Powys,
UK).
All
the
sensors
were
set
on
an
East–West
orientated
line,
below
a
PVP
strip,
for
FD
and
HD.
The
ground
was
then
covered
with
spring
barley
sown
on
DOY
89,
with
an
almost
closed
canopy.
2.5.
Theoretical
background:
crop
energy
balance
and
consequences
on
crop
temperatures
The
temperature
of
any
plant
organ
results
from
the
balance
between
incoming
energy
and
energy
loss.
The
energy
balance
of
a
plant
organ
(or
a
canopy)
can
be
written
as:
Rn
H
E
G
=
0
(1)
where
Rnis
the
net
radiation,
H
and
E
are
the
sensible
and
latent
heat
fluxes
between
the
considered
vegetation
and
the
surrounding
air,
and
G
is
the
rate
of
heat
storage
in
the
vegetation
and
soil.
During
daylight,
the
main
energy
input
is
radiation,
both
solar
and
longwave
radiation
(Eq.
(2)).
If
crop
or
leaf
temperature
is
dif-
ferent
from
air
temperature
at
the
same
height
(TA(z)),
energy
may
be
absorbed
or
released
as
sensible
heat
by
convection
(Eq.
(3)).
Part
of
this
energy
can
also
be
released
by
evaporation
through
stomata
(Eq.
(4a)).
Moreover,
energy
may
be
transferred
to
and
from
storage
in
plant
canopies
and
in
the
soil
by
conduction
(Eq.
(5)).
During
the
night,
the
sign
of
must
fluxes
change
to
the
opposite.
The
radiation
balance
is
negative
then,
and
water
vapor
may
condense
on
plants
(Eq.
(4b)).
Rn=
(1
˛)
·
Rs+
ε
·
R
ε
·
·
T4
L(2)
H
=
·
Cp
ra
(TL
TA(z))
(3)
·
E
=
·
Cp
·
(ra+
rs)(e(TL)
ea(z))
for
transpiration
(a)
·
E
=
·
Cp
·
ra
(e(TL)
ea(z))
for
dew
deposition
or
evaporation
(b)
(4)
G
=
·
Cp
r0(T0
Tz0) (5)
where
Ris
the
downwards
incoming
longwave
radiation,
eais
the
vapor
pressure
of
the
air
surrounding
the
plant
organ,
e*(TL)
is
the
saturated
vapor
pressure
at
leaf
temperature,
rais
the
aerody-
namic
resistance
to
water
vapor
and
sensible
heat
mainly
function
of
wind
speed,
rsthe
stomatal
resistance,
is
the
air
density,
Cpis
the
specific
heat
of
air
at
constant
temperature,
is
the
psychro-
metric
constant,
T0is
the
soil
surface
temperature,
and
Tz0is
the
soil
temperature
at
depth
z0.
The
organ
or
plant
temperature
balances
the
energy
budget
equation
(Eq.
(1)).
For
certain
sets
of
environmental
conditions
(air
temperature,
solar
radiation,
vapor
pressure
and
wind
speed),
only
one
surface
temperature
that
balances
the
energy
budget
equation
exists.
Net
radiation
is
influenced
by
the
PVPs
through
(1)
the
reduc-
tion
of
the
solar
incident
radiation
during
day-time
only
(Rs),
(2)
the
modification
of
downwards
long
wave
radiation
(R)
coming
from
the
sky
and
from
the
PVPs
in
the
case
of
agrivoltaic
systems.
Latent
heat
flux
H
could
change
under
PVPs
if
air
temperature
or
wind
speed
were
modified
by
the
PVPs
(leading
to
variations
in
the
aero-
dynamic
resistance
of
the
crop).
Heat
storage
is
negligible
for
leaves
or
small
canopies.
For
the
wheat
crop,
there
is
no
heat
conduction
between
leaves
and
ground
as
the
crop
stands
at
a
sufficient
height
(more
than
50
cm)
above-ground
in
the
second
part
of
the
crop
cycle.
In
the
case
of
short
or
creeping
plants
(lettuce
and
cucum-
ber),
conduction
occurs
and
the
intensity
of
this
flux
depends
on
the
vertical
gradient
of
soil
temperature.
Any
change
in
soil
tempera-
ture
could
result
in
a
modified
G
flux
under
PVPs.
Finally,
energy
can
be
released
as
latent
heat
(E)
through
crop
transpiration
or
dew
evaporation
in
the
morning,
which
are
monitored
by
aerody-
namic
and
stomatal
(for
transpiration
only)
resistances.
PVPs
could
modify
the
latent
heat
exchanges
between
the
plant
and
the
sur-
rounding
air
by
a
change
in
leaf
stomatal
resistance
or
air
vapor
pressure.
2.6.
Statistical
analysis
All
statistical
analyses
were
performed
with
the
R
software.
Analyses
of
variance
(ANOVA)
and
covariance
(ANCOVA)
were
realized
with
the
‘lm’
procedure
and
boxed
linear
models
were
Author's personal copy
122 H.
Marrou
et
al.
/
Agricultural
and
Forest
Meteorology
177 (2013) 117–
132
Table
1
Sum
of
TAcalculated
for
the
totality
or
a
part
of
each
cropping
season.
Differences
in
crop
cycle
length
between
the
shaded
situations
and
FS
are
expressed
in
thermal
time
(C
d)
and
in
equivalent
number
of
leaves,
according
to
the
literature.
Sums
of
temperature
are
calculated
with
a
base
temperature
of
3C
for
lettuces,
0C
for
wheat,
and
15 C
for
cucumbers.
From
planting
to
harvest
FS
FD
HD
Spring
lettuces
(2011)
Crop
cycle
length
(C
d) Sum
of
TA(C
d)
797
819
812
Difference
(C
d)
0
22.8
15.6
Equivalent
NB
of
leaves Min
development
rate
0.052
leaves
(C
d1)
0
1.18
0.81
Max
development
rate
leaves
(C
d1)
0
2.96
2.03
From
mid-cycle
to
harvest FS
FD
HD
Summer
lettuces
(2010)
Crop
cycle
length
(C
d) Sum
of
TA(C
d) 465
465
471
Difference
(C
d)
0
0.00
5.5
Equivalent
NB
of
leaves Min
development
rate
0.052
leaves
(C
d1)
0
0.00
0.29
Max
development
rate
0.130
leaves
(C
d1)
0
0.00
0.72
From
planting
to
harvest
FS
FD
HD
Cucumbers
Crop
cycle
length
(C
d) Sum
of
TA(C
d)
467
474
469
Difference
(C
d)
0
7.4
1.5
Equivalent
NB
of
leaves Min
development
rate
0.069
leaves
(C
d1)
0
0.51
0.10
Max
development
rate
0.083
leaves
(C
d1)0
0.61
0.12
From
tillering
to
harvest
FS
FD
HD
Durum
wheat
Crop
cycle
length
(C
d) Sum
of
TA1820
1872
1845
Difference
(C
d)
0
52.98
24.54
Equivalent
NB
of
leaves Min
development
rate
0.0014
leaves
(C
d1)
0
0.07
0.03
Max
development
rate
0.0147
leaves
(C
d1)0
0.78 0.36
compared
according
to
the
maximum
of
likelihood
ratio
(‘anova’
procedure).
Mean
comparison
between
treatments
was
performed
using
Student
tests
(t-test
procedure)
and
least
significant
differ-
ences
test
(LSD-test
procedure).
Sigmoid
adjustments
were
fitted
for
the
number
of
cucumber
leaves
using
the
nls
procedure.
3.
Results
3.1.
Incident
radiation
The
average
proportion
of
daily
radiation
transmitted
below
the
PVPs
(FD
and
HD
treatments)
compared
to
the
FS
treatment
ranged
around
32%
in
FD,
and
48%
in
HD
during
the
lettuce
crop
cycle
(DOY
117–143),
52%
in
FD
and
68%
in
HD
during
the
wheat
crop
cycle
(DOY
35–168)
and
37%
in
FD
and
62%
in
HD
during
the
cucum-
ber
crop
cycle
(DOY
181–240)
(Fig.
2).
The
average
proportion
of
radiation
transmitted
daily
below
the
PVPs
varied
from
one
day
to
another
one
with
a
coefficient
of
variation
equal
to
37%
in
FD
and
46%
in
HD
over
the
whole
measurement
period
(DOY
35–240).
The
fraction
of
transmitted
variation
also
varied
within
day-time
(Cv=
57%
in
FD
and
60%
in
HD,
between
09:00
and
18:00,
in
average
over
all
the
days
with
means).
For
a
given
day
of
the
year,
the
proportion
of
transmitted
radia-
tion
over
24
h
varies
depending
on
the
position
of
the
plants
on
the
North–South
axis.
The
coefficient
of
spatial
variation
of
the
trans-
mitted
radiation
equals
29%
in
FD
and
38%
in
HD,
in
average
over
the
measurement
period
(DOY
35–244).
3.2.
Aerial
microclimate
A
wide
range
of
climatic
conditions
was
explored
over
the
four
crop
cycles:
air
temperature
(TA)
varied
from
3
to
28 C
and
daily
incident
solar
radiation
varied
from
1
to
31
MJ
m2d1in
FS.
How-
ever,
TAin
the
shaded
treatments
(both
FD
and
HD)
remained
nearly
equal
to
that
in
the
FS
during
all
cropping
cycles,
from
July
2010
to
August
2011
(Fig.
3),
according
to
ANOVAs.
ANOVAs
were
repeated
for
every
day,
using
shading
treatments
as
an
explicative
factor
of
mean
daily
TA.
Only
12
days
were
found
to
have
a
risk
p-value
below
5%
(i.e.
with
a
significant
effect
of
shading
on
air
temperature).
Dur-
ing
days
with
low
wind
speed
(u
<
1.2
m
s1and
umax <
6
m
s1)
or
high
global
radiation
(Rs>
24
MJ
m2d1),
air
temperature
below
the
solar
panels
tended
to
be
higher
than
in
FS.
Variations
in
daily
TAbetween
shade
treatments
resulted
in
vari-
ations
of
the
crop
cycle
length
in
thermal
time
that
never
exceeded
23 C
d
for
lettuces,
8C
d
for
cucumber
and
53 C
d
for
wheat.
According
to
literature
references,
these
differences
were
too
small
to
allow
the
production
of
even
one
more
leaf
in
the
case
of
cucum-
ber
(Horie
et
al.,
1979)
and
wheat
(Porter
and
Gawith,
1999).
For
lettuces,
in
2011
only,
two
to
three
extra
leaves
(on
a
total
of
80
leaves
in
average,
Marrou
et
al.,
2013)
could
have
been
emitted
in
FD
as
mean
air
temperatures
tended
to
increase
in
average
over
the
second
part
of
the
cropping
cycle.
However,
as
the
thermal
time
required
for
lettuce
leaves
to
reach
the
length
of
1
cm
is
over
200 C
d
at
20 C
in
FS
(Bensink,
1971),
this
increase
in
the
num-
ber
of
leaves
may
have
little
impact
on
lettuce
size
or
dry
weight.
Results
are
summed
up
in
Table
1.
When
carrying
the
same
type
of
analysis
at
an
hourly
time
step
over
the
measurement
period,
differences
in
hourly
records
of
air
temperature
were
never
found
between
the
shaded
(FD
or
HD)
and
unshaded
treatments
at
the
same
time
of
the
day.
Similarly,
no
significant
effect
of
shading
was
found
on
air
rela-
tive
humidity
or
VPD,
neither
at
a
daily
nor
at
an
hourly
time
step.
The
maximal
increase
in
VPD
in
shaded
treatments
compared
to
FS
was
0.11
kPa,
while
the
mean
VPD
over
the
measurement
period
is
0.91
kPa,
in
FS.
Besides,
the
horizontal
wind
speeds
measured
at
an
hourly
time
step
in
FS
and
in
each
of
the
shaded
treatments
(FD
and
HD)
were
found
to
be
similar,
regarding
the
precision
of
measurements
Author's personal copy
H.
Marrou
et
al.
/
Agricultural
and
Forest
Meteorology
177 (2013) 117–
132 123
Fig.
4.
Mean
daily
ground
temperature
measured
in
non-irrigated
winter
wheat
at
(a)
0.05
m
and
(b)
0.25
m
depth;
and
in
sprinkler
irrigated
summer
lettuce
at
(c)
25
cm
depth.
(±0.3
m
s1).
However,
the
significance
of
differences
between
treatments
could
not
be
tested
due
to
the
lack
of
replications.
3.3.
Soil
temperature
During
the
wheat
cycle,
soil
temperature
at
0.05
m
depth
(TS5)
decreased
by
1.9 C
in
FD
and
1.8 C
in
HD
compared
to
FS
(Fig.
4).
A
covariance
analysis
using
shade
treatment
as
a
factor
and
LSD
comparison
tests
on
mean
values
showed
that
these
differences
were
significant
(p-value