Sachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Estimation of the vapour pressure deficit using noaa avhrr data
1. This article was downloaded by: [Siirt Universitesi]
On: 16 January 2013, At: 00:41
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
International Journal of Remote
Sensing
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tres20
Estimation of the vapour pressure
deficit using NOAA-AVHRR data
Mehmet Şahin
a
, Bekir Yiğit Yıldız
b
, Ozan Şenkal
c
& Vedat
Peştemalci
d
a
Engineering Faculty, Siirt University, Siirt, Turkey
b
Karaisalı Vocational School, Çukurova University, Adana, Turkey
c
Faculty of Education, Department of Computer Education and
Instructional Technology, Çukurova University, Adana, Turkey
d
Department of Physics, Çukurova University, Adana, Turkey
Version of record first published: 15 Jan 2013.
To cite this article: Mehmet Şahin , Bekir Yiğit Yıldız , Ozan Şenkal & Vedat Peştemalci (2013):
Estimation of the vapour pressure deficit using NOAA-AVHRR data, International Journal of Remote
Sensing, 34:8, 2714-2729
To link to this article: http://dx.doi.org/10.1080/01431161.2012.750021
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-
conditions
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
3. International Journal of Remote Sensing 2715
VPD affects plants. For example, the water stress index was recently developed using
VPD by employing a radiometric measurement of foliage temperature and a psychometric
measurement of air VPD. It is necessary to find the relationship between the foliage-air
temperature differential and the air VPD for plants (Idso 1982). Strawberry plants were
monitored for water stress by measuring the foliage temperature with a hand-held infrared
thermometer. In addition to the foliage temperature measurement, weather variables, the
difference between leaf and air temperature, derived crop water stress index, soil matric
potential, leaf water potential, photosynthetic gas exchange rates, transpiration rates, photo-
synthetic pigments, sugars, starch, canopy structure, and accumulated yield were measured.
A regression analysis showed that the air VPD contributed significantly to variations in leaf
water potential under very mild water stress conditions tested in a wet treatment. However,
the contribution was not significant under mild water stress of dry treatment (Peñuelas et al.
1992). Leonardi, Guichard, and Bertin (2000) investigated how a high VPD influences the
growth, transpiration, and quality of tomato fruits. In this study, plants were grown in two
glasshouse compartments under two VPD levels: a low VPD was obtained by increasing
air humidity with a fogging system, and a high VPD was obtained during sunny hours in
a greenhouse where the air humidity was not controlled. The mean value of the six driest
hours of the day during the considered growing period of the fruits was 16 mb under low
VPD and 22 mb under high VPD conditions. The study showed that as VPD increases from
16 to 22 mb during summer, effects can be observed on both tomato growth and quality.
During daylight hours, the relative fruit growth rate was significantly reduced for plants
grown under higher VPDs. The same trend was not observed at lower VPDs, where fruit
growth varied more regularly during daylight hours. Habermann et al. (2003) found that
healthy sweet orange plants measured at a VPD of 25 mb showed a 50% decrease in the
transpiration rate and an 80% decrease in stomatal conductance when compared to mea-
surements at 12 mb. The amount of proportion between photosynthesis and transpiration
rates and stomatal conductance of leaves from healthy plants was measured at both VPDs.
Reducing VPD has no long-term effect on the growth of temperate and tropical
rainforest trees. Therefore, large reductions in stomatal conductance and net photosyn-
thesis were measured with increasing VPDs in tropical rainforest trees. Earlier studies did
not show significant reductions in growth (Cunningham 2004, 2005). Possible explana-
tions for the lack of effect of VPD on growth include avoidance of water stress due to
adequate soil moisture, increased nutrient uptake under ambient conditions due to higher
transpiration rates, and other factors such as temperature and light, and limiting the growth
of tropical species. In temperate rainforests, unlike in this experiment, higher VPD val-
ues in summer were associated with a limited water supply. Therefore, a more important
difference between temperate and tropical rainforest trees is the ability of the temperate
species to tolerate a mild soil drought. Temperate rainforest trees may possess adaptations
such as deeper, more extensive root systems and increased resistance to xylem cavitation
and osmotic adjustments that allow them to maintain photosynthesis and growth for longer
periods under low-soil moisture than wet, tropical rainforest trees (Cunningham 2006).
Williams and Baeza (2007) studied Vitis vinifera grown at different locations to deter-
mine the relationship among temperature, VPD, and leaf water potential under clear skies
during midday. Temperature and VPD were determined at the time of measurement. The
highest and lowest leaf water potential values measured on well-watered grapevines were
51.102
and 115.102
mb, respectively. According to the results, the leaf and stem water
potentials were linearly related to VPD. The coefficient of determination was greater for
the relationship between the leaf water potential and VPD (r2
= 0.74) than temperature and
VPD (r2
= 0.58). The leaf water potential and stem water potential values as a function
Downloadedby[SiirtUniversitesi]at00:4116January2013
4. 2716 M. ¸Sahin et al.
of VPD or temperature could serve as baselines, indicating whether the grapevines are
fully irrigated or are not water stressed under the environmental conditions found in semi-
arid grape-growing regions. Lendzion and Leuschner (2008) studied the effects of elevated
atmospheric water VPD on the European beech. They tested the hypothesis that an increas-
ing VPD negatively affects the growth and development of European beech saplings. Their
results show that the beech sapling growth and development strongly depend not only
on soil moisture but also on the prevailing VPD level. Glenn et al. (2008) found that
transpiration highly depended on the leaf area and was controlled by VPD in the atmo-
sphere. Sarcobatus vermiculatus tended to have higher transpiration rates than Atriplex
canescens and had a steeper response to VPD.
VPD is one of the critical variables that drives evapotranspiration (ET) and is funda-
mentally important to many models. The estimation of VPD can be used in several ET
estimation equations that are used to estimate regional ET patterns in which only tem-
perature, precipitation, and insolation measurements are available (Castellvi et al. 1996,
1997). Furthermore, VPD is one of the key controls in opening the stomata in plants and
is thus an important force for ET, plant respiration, biomass production, and the uptake of
harmful pollutants such as ozone through the stomata (Andersson-Sköld, Simpsonb, and
Ødegaard 2008). To survive in adverse environments subject to drought, high-salt con-
centration, or low temperature, some plants seem to be able to synthesize biochemical
compounds, including proteins, in response to changes in water activity or osmotic pres-
sure. Water activity or osmotic pressure measurements of simple aqueous solutions have
been based on freezing point depression or VPD. Osmotic pressure measurements of plants
under water stress have been mainly based on VPD (Kiyosawa 2003).
As observed in recent studies, VPD plays an important role in various fields.
Additionally, it was estimated using meteorological stations that are limited in the regions
studied. For these reasons, new methods to estimate VPD are necessary. One of these meth-
ods uses satellite data. Choudhury (1998) estimated VPD over land surfaces from satellite
observations. This study presented a method for estimating the monthly mean VPD from
satellite observations and then evaluated the accuracy of the estimated values by compar-
ing them with globally distributed ground measurements. The square of the correlation
coefficient and standard error of estimation were found to be 0.85 and 4 mb, respec-
tively. Prince et al. (1998) conducted a study to obtain VPD and then compared the values
with NOAA/AVHRR (National Oceanic and Atmospheric Administration/Advanced Very
High Resolution Radiometer) and field observations. The comparison resulted in a VPD
root mean square error (RMSE) of 10.9 mb over a range of 58 mb. Hay and Lennon
(1999) studied meteorological variables and control of vector-borne disease across Africa
and compared remote sensing and climate spatial interpolation using VPD obtained
from the NOAA/AVHRR data set. According to the result, the mean accuracy for the
year was an RMSE of 6 mb (range 4.91–6.43) with a mean adjusted r2
= 0.63 (range
0.40–0.71) and an RMSE of 5.3 mb (range 3.65–7.64) with a mean adjusted r2
= 0.78
(range 0.67–0.86). Hashimoto et al. (2008) developed simple linear models to predict VPD
using saturated vapour pressure calculated from MODIS (Moderate Resolution Imaging
Spectroradiometer) – LST (land surface temperature) at a number of different temporal and
spatial resolutions. They estimated model parameters for VPD estimation both regionally
and globally with RMSE values ranging from 3.2 to 3.8 mb. VPD was overestimated along
coastlines and underestimated in arid regions with low-vegetation cover. Additionally, the
residuals were larger with higher VPDs because of the non-linear relationship between sat-
uration vapour pressure and LST. Linear relationships were observed at multiple scales and
appeared useful for estimation within the range 0–25 mb.
Downloadedby[SiirtUniversitesi]at00:4116January2013
5. International Journal of Remote Sensing 2717
In this study, land surface temperature (Ts), total precipitable water in the atmospheric
column (U), and dew point temperature (Td) were calculated to estimate VPD using satel-
lite data. Meanwhile, VPD was estimated using meteorological data. Then, the results were
statistically compared.
2. Data
Two different data sets received from the Scientific and Technological Research Council
of Turkey and the Turkish State Meteorological Service were used to obtain VPD. First,
raw NOAA12-14-15/AVHRR data were translated into a Level 1b format using Quorum
Software, and in the second step, the brightness temperatures of channel 4 and channel
5 (range 10.3–11.3 µm and range 11.5–12.5 µm, respectively) were obtained from Level
1b data by employing the Envi 4.3 image-processing program and data received from the
Scientific and Technological Research Council of Turkey during 2002.
Land meteorological values were necessary to determine whether the VPD estimates
obtained from the satellites are indeed adequate. To achieve this, air temperature and vapour
pressure (VPair) values were received from the Turkish State Meteorological Service.
3. Estimation of land-surface temperature
Land-surface temperature is important because it is one of the key factors in determin-
ing the exchange of energy and matter between the Earth’s surface and atmosphere.
Simultaneously, it is an important measurement for energy-balance applications and can
be especially useful when determined by thermal infrared remote sensing (Seguin and
Itier 1983). An approach based on the differential absorption in two adjacent infrared
channels, called the ‘Split-Window’ technique, is used for determining the surface temper-
ature. The AVHRR channels 4 and 5 are widely used for deriving the surface temperature
(Kant and Badarinath 2000). Many algorithms have been proposed by Price (1984), Becker
(1987), Becker and Li (1990), Vidal (1991), Prata (1993, 1994), Sobrino et al. (1994,
1996), Coll et al. (1994), Becker and Li (1995), Coll and Caselles (1997), Ouaidrari et al.
(2002), Pinheiro et al. (2006), and Katsiabani, Adaktilou, and Cartalis (2009). These stud-
ies indicated that it is possible to retrieve LST at a reasonable accuracy (RMSE of 1–3 K)
from current operational and research satellite-borne visible/infrared radiometers. In this
study, the Price (1984), Becker and Li (1990), Vidal (1991), and Ulivieri et al. (1994)
split-window algorithm techniques were used to obtain land-surface temperatures.
3.1. Price (1984) algorithm
By reducing the effect of atmosphere and using radioactive transfer theory, Price (1984)
developed a split-window algorithm technique that has been used extensively. The basic
split-window algorithm can be written as
Ts = T4 + a(T4 − T5) + b, (1)
where coefficients a and b account for atmospheric conditions (related to spectral radi-
ance and transmission) and surface emissivity, respectively. However, linear empirical
formulations do not always hold. Hence, the water vapour dependence was subsequently
incorporated into a non-linear quadratic equation (Coll et al. 1994; François and Ottle
1996). Coefficient a in Equation (1) was given by a = ((a5/a4) – 1)−1
, where a5/a4 was
determined from T5 ( T4)−1
(the brightness temperature spatial variations in AVHRR
Downloadedby[SiirtUniversitesi]at00:4116January2013
6. 2718 M. ¸Sahin et al.
channels 4 and 5) for the small study area. The a5/a4 value was calculated to be 1.30,
a = 3.33, and b was linked to the emissivity difference. The emissivity difference, ε =
ε4 − ε5 = –0.005, and ε depend on ε4 and ε5 as in the relation, ε =
ε4 + ε5
2
= 0.975
(Caselles, Coll, and Valor 1997; Chrysoulakis and Cartalis 2002), where ε4 and ε5 are
the emissivities of channels 4 and 5; and T4 and T5 are the brightness temperatures of
NOAA/AVHRR channels 4 and 5, respectively (Dash et al. 2002). The final form of the
equation is
Ts = [T4 + 3.33 (T4 − T5)]
5.5 − ε4
4.5
− 0.75T5 ε. (2)
3.2. Becker and Li (1990) algorithm
Based on radioactive transfer theory and numerical simulations, Becker and Li (1990)
proposed a local split-window algorithm for AVHRR channels 4 and 5:
Ts = 1.274 + P
T4 + T5
2
+ M
T4 − T5
2
, (3)
where temperature is in K, and the coefficients P and M are given by
P = 1 + 0.15616
1 − ε
ε
− 0.482
ε
ε2
, (4)
M = 6.26 + 3.98
1 − ε
ε
+ 38.33
ε
ε2
, (5)
where P and M are local coefficients that depend on the surface emissivity, but are indepen-
dent of atmospheric effects. ε = ε4 − ε5 = –0.005, and ε = 0.975 (Caselles, Coll, and
Valor 1997; Chrysoulakis and Cartalis 2002). The coefficient 1.274 in Equation (3) was
calculated from numerical simulations and local atmospheric effects (Becker and Li 1990).
3.3. Vidal (1991) algorithm
Ts = T4 + 2.78 (T4 − T5) + 50
1 − ε
ε
− 300
ε
ε
. (6)
The coefficients related to the emissivity in this algorithm were obtained from a study by
Becker (1987). ε = ε4 − ε5 = –0.005, and ε = 0.975 (Caselles, Coll, and Valor 1997;
Chrysoulakis and Cartalis 2002). This algorithm was generated from a large number of
satellite data and land-surface temperature calculations (Vidal 1991).
3.4. Ulivieri et al. (1994) algorithm
This algorithm was developed by Ulivieri et al. (1994) for its simplicity, robustness, and
superior performance in independent tests. Becker and Li (1995) and Vazquez, Reyes, and
Arboledas (1997) tested the algorithm with different data sets and different split-widow
algorithms. In all cases, the Ulivieri et al. (1994) algorithm performed well. The Ulivieri
et al. algorithm can be written as
Downloadedby[SiirtUniversitesi]at00:4116January2013
7. International Journal of Remote Sensing 2719
Ts = T4 + 1.8 (T4 − T5) + 48(1 − ε) − 75 ε, (7)
where ε = ε4 – ε5 = –0.005 and ε = 0.975 (Caselles, Coll, and Valor 1997; Chrysoulakis
and Cartalis 2002). This equation was developed for cases of column atmospheric water
vapour less than 3 g cm−2
, a reasonable condition for many of the semi-arid areas of
continental Africa (Pinheiro et al. 2006).
4. Estimation of VPD
Generally, researchers use two sources to find VPD: meteorological station data and satellite
data. Therefore, these two sources were used in this study to perform comparison.
4.1. Estimation of VPD using meteorological station data
The vapour pressure (VPair) is a measure of how much water vapour is in the air, i.e. how
much water in the gas phase is present in the air. The presence of more water vapour in the
air leads to a greater water vapour pressure. When the air reaches its maximum water con-
tent, the vapour pressure is called the saturation vapour pressure (VPsat), which is directly
related to temperature. Thus, the difference between the saturation vapour pressure and
the real air vapour pressure is called VPD. The magnitude of VPD gives an indication
of how close the air is to condensation (Choudhury 1998; Prenger and Ling 2000). Very
simply, VPD is a measure of the lack of moisture equilibrium between an object and the
surrounding atmosphere (Hay and Lennon 1999). VPD is given by Unwin (1980) as
VPD = VPsat − VPair, (8)
where the saturation vapour pressure, VPsat (mb), is given by
log10 VPsat = 9.24349 −
2305
Tair
−
500
T2
air
−
100000
T3
air
, (9)
where Tair is the air temperature in Kelvin.
4.2. Estimation of VPD using satellite data
Determining VPD, and the difference between saturated and actual atmospheric vapour
pressures, involves estimating the precipitable water in the atmospheric column using
the thermal infrared channels 4 and 5 of AVHRR, from which the surface humidity is
derived. The total precipitable water in the atmospheric column, U (kg m−2
), is estimated
as Equation (10) (Eck and Holben 1994):
U = A + B (T4 − T5) , (10)
where A and B are constants equal to 1.337 and 0.837, respectively. The total precipitable
water is expressed in units of pressure and is converted to the amount of water in centime-
tres that would be precipitated from the atmospheric column by dividing by 10, because the
density of water is 1 g cm−3
. The estimated precipitable water content U is used to obtain
the surface dew temperature Td (◦
F); surface dew temperature can be calculated using the
following equation (Smith 1966):
Downloadedby[SiirtUniversitesi]at00:4116January2013
8. 2720 M. ¸Sahin et al.
Td(◦
F) =
ln U − (0.113 − ln(λ + 1))
0.0393
. (11)
The λ values given by Smith for different latitudinal zones were used. In this analysis, a
mean value of λ = 2.99 was calculated from the annual mean λ presented by Smith for
locations between 0 and ±40 degrees of latitude. Then, the Td values should be converted
into kelvin (Smith 1966).
Finally, VPD in kilopascals (kPa) can be calculated from Td and Ts, as given by
Equation (12), following Prince and Goward (1995):
VPD = 0.611 exp 17.27
Ts − 273
Ts − 36
− exp 17.27
Td − 273
Td − 36
. (12)
5. Evaluation of the estimation results
The choice of the relevant criteria allowing the estimation methods’ performance evalua-
tion is an important issue. Various statistical parameters can be used to measure the strength
of the statistical relationship between the estimated and reference values. We assume that vi
(i = 1, n) is the set of n reference values and ei (i = 1, n) is the set of estimates; ¯v and ¯e are
mean reference and estimate value, respectively. The bias, linear correlation coefficient (r),
and RMSE can be calculated using the standard deviations of the reference (σv) and esti-
mate (σe) values, means of the reference and estimate values, and estimated and reference
values. The bias is the difference between the mean estimate ¯e and the mean reference value
¯v. The statistical criterion formula of the linear correlation coefficient r is the following:
r =
n
i=1 (vi − ¯v) (ei − ¯e)
nσvσe
, (13)
where r measures the proximity between estimate and reference and is not sensitive to a
bias (Kendall and Stuart 1963). The formula of RMSE is
RMSE =
1
n
n
i=1
(ei − vi)2
1
2
. (14)
In statistics, RMSE is a frequently used measure of the differences between values pre-
dicted by a model or an estimator and the values actually observed from the subject being
modelled or estimated (Laurent, Jobard, and Toma 1998).
6. Results
6.1. Land-surface temperature
After obtaining brightness temperatures of NOAA/AVHRR channels 4 and 5, split-window
algorithms were used to obtain the land-surface temperature. The Price (1984) algorithm
was calculated first using Equation (2) (see Figure 1). Then, the Becker and Li (1990), Vidal
(1991), and Ulivieri et al. (1994) algorithms were calculated using Equations (3), (6), and
(7), respectively (see Figures 2–4). When the maps in Figures 1–4 were examined through
image-processing programs, the land-surface temperature from the Price (1984) algorithm
was measured at a minimum of 299 K, maximum of 310 K, and an average of 302.75 K.
Downloadedby[SiirtUniversitesi]at00:4116January2013
9. International Journal of Remote Sensing 2721
293.88
<273.57
276.47
279.37
282.27
285.18
288.08
290.98
296.78
299.69
302.59
305.49
308.39
311.29
314.20
317.10
320.00
N
Figure 1. Map of land-surface temperature obtained as based on the Price (1984) algorithm at
06.51 local time on 6 July 2002 (K).
N
<273.00
275.94
281.81
284.75
290.63
293.56
296.50
299.44
302.38
305.31
308.25
311.19
314.13
317.06
320.00
287.69
278.88
Figure 2. Map of land-surface temperature obtained as based on the Becker and Li (1990) algorithm
at 06.51 local time on 6 July 2002 (K).
In the Becker and Li (1990) algorithm, the minimum value was determined to be 296.91 K,
the maximum was 307.78 K, and the average was 301.16 K; in the Vidal (1991) algorithm,
the minimum value was found to be 299.28 K, the maximum was 309.30 K, and the average
was 302.84 K; in the Ulivieri et al. (1994) algorithm, the minimum value was calculated as
297.56 K, the maximum was 307.47 K, and the average was 300.83 K.
Accordingly, the land-surface temperature calculation was completed using
24 NOAA/AVHRR satellite images for each split-window algorithm. The values obtained
from the split-window algorithms had to be evaluated with the meteorological data from
chosen control point cities on Turkey’s map. Therefore, it was important to choose the
cities on the map by taking into consideration Turkey’s different geographical regions and
at least one city from each region had to be included. The cities were chosen according to
the geographical regions of Turkey and the city locations on the map (see Figure 5).
As observed on the map, Adana, Ankara, Antalya, Balıkesir, Denizli, Erzurum, ˙Izmir,
Kayseri, Malatya, Samsun, Sivas, Rize, and Van were used as the control points to
Downloadedby[SiirtUniversitesi]at00:4116January2013
10. 2722 M. ¸Sahin et al.
N
<273.00
320.00
275.94
281.81
284.75
290.63
293.56
296.50
299.44
302.38
305.31
308.25
311.19
314.13
317.06
287.69
278.88
Figure 3. Map of land-surface temperature obtained as based on the Vidal (1991) algorithm at
06.51 local time on 6 July 2002 (K).
N
<273.00
320.00
275.94
281.81
284.75
290.63
293.56
296.50
299.44
302.38
305.31
308.25
311.19
314.13
317.06
287.69
278.88
Figure 4. Map of land-surface temperature obtained as based on the Ulivieri et al. (1994) algorithm
at 06.51 local time on 6 July 2002 (K).
determine land-surface temperature accuracy. The minimum, maximum, and average
values of land-surface temperature obtained from control points were compared to
meteorological values on a monthly basis (see Table 1). Although the averages of the
meteorological and algorithmic values in January were the same (281.05 K), the land-
surface temperature values ranged from 268.70 to 293.21 K. In February, the average of
the meteorological values was 285.78 K, and the nearest estimate was 288.94 K, which
was from Ulivieri et al. (1994). The four algorithms gave results that were somewhat
similar to the meteorological values in March, October, and December. The average of
the meteorological values in April was 283.95 K, and the closest estimates belonged to
the Becker and Li (1990) and Ulivieri et al. (1994) algorithms. In May, the meteoro-
logical values were 270.20–282.30 K, and the average was 291.60 K. The best results
in terms of the algorithm average values belonged to Ulivieri et al. (1994) and Becker
and Li (1990) with 283.34 and 283.63 K, respectively. In June and July, the Ulivieri
Downloadedby[SiirtUniversitesi]at00:4116January2013
11. International Journal of Remote Sensing 2723
Kayseri
Samsun
Ankara
Antalya
Denizli
Afyonkarahisar
izmir
Bahkesir
Istanbul
Konya
Adana
Malatya
Sivas
Rize
Artvin
Van
Kars
Erzurum
Figure 5. Locations of the cities used to estimate land-surface temperature and VPD.
et al. (1994) algorithm gave the results most consistent with the meteorological val-
ues. In August, the average error values of the Becker and Li (1990) and Ulivieri
et al. (1994) algorithms were approximately 2 K compared to the meteorological val-
ues. In September, the average error temperature values of all algorithms had a deviation
ratio with the meteorological values in the range of 0.07–1.9 K. In November, the clos-
est value to the meteorological value was from the Vidal (1991) and Ulivieri et al. (1994)
algorithms.
Additionally, statistical evaluation was performed by considering satellite and meteoro-
logical data with Equations (13) and (14). According to the evaluation result, the correlation
coefficients (r) were found to be 0.958, 0.961, 0.967, and 0.970 according to Price (1984),
Becker and Li (1990), Ulivieri et al. (1994), and Vidal (1991), respectively (see Figure 6).
The correlation coefficient results show a strong relationship between the satellite and
meteorological data.
The other statistical result was the RMSE values of the algorithms. The RMSE values
ranged from 2.7 K, which was calculated using the Ulivieri et al. (1994) algorithm, to nearly
4 K, which was calculated from the Price (1984) algorithm. The algorithm with the smallest
RMSE value was that of Ulivieri et al. (1994); thus, this algorithm is suggested to estimate
the land-surface temperature among the Price (1984), Becker and Li (1990), Vidal (1991),
and Ulivieri et al. (1994) algorithms.
6.2. Vapour pressure deficit
Two approaches (the first for the meteorological data, and the second for the satellite data)
were followed to estimate VPD. The VPD values for the meteorological data were cal-
culated using Equations (8) and (9), whereas the VPD values for the satellite data were
calculated using Equations (7), (10), (11), and (12) over the satellite images (see Figure 7).
When Figure 7 was examined through an image-processing program, it was observed
that VPD values were in the ranges 0–10 mb and 10–20 mb, which were considerably
low. Moreover, the VPD values in the range 20–30 mb were rather frequent over Turkey.
It was found that the VPD values were between 30 and 40 mb over the following regions
of Turkey: Central Anatolia, South-Eastern, Mediterranean, and the coastal lines of the
Downloadedby[SiirtUniversitesi]at00:4116January2013
12. 2724 M. ¸Sahin et al.
Table 1. Min, max, and average temperature values of meteorological and algorithm data (K).
Month
Minimum/
maximum/
average (K)
Meteorological
value
Ulivieri et al.
(1994)
Becker and
Li (1990)
Vidal
(1991) Price (1984)
Min 271.40 268.70 269.69 271.10 270.00
January Max 289.40 291.19 291.39 293.21 293.00
Ave 281.05 281.05 281.05 281.05 281.05
Min 277.90 282.40 283.49 284.49 284.00
February Max 291.60 293.32 293.79 295.65 296.00
Ave 285.78 288.94 289.15 290.85 290.69
Min 270.20 267.50 267.63 268.85 268.00
March Max 289.30 289.77 289.58 291.14 290.00
Ave 283.78 283.46 283.37 284.82 283.85
Min 276.20 277.07 277.05 278.43 277.00
April Max 287.00 290.11 290.53 293.28 292.00
Ave 283.95 286.59 286.85 287.93 288.21
Min 270.20 267.50 267.63 268.85 268.00
May Max 291.60 293.32 293.79 295.65 296.00
Ave 282.30 283.34 283.63 285.07 284.48
Min 290.40 293.61 295.00 295.00 296.00
June Max 302.60 305.00 305.19 307.00 306.00
Ave 298.26 300.51 301.17 302.30 301.38
Min 293.60 297.56 296.91 299.06 299.00
July Max 303.50 307.47 308.92 309.84 310.00
Ave 298.04 300.73 301.62 302.48 303.01
Min 294.30 296.96 296.67 297.82 298.00
August Max 315.00 315.35 311.89 316.56 315.00
Ave 300.49 302.68 302.19 304.34 304.04
Min 283.40 284.26 284.22 285.03 285.00
September Max 308.00 304.38 304.22 307.47 306.00
Ave 293.19 293.26 293.83 295.08 295.09
Min 280.50 277.99 277.83 279.31 279.00
October Max 292.00 292.26 291.41 292.37 293.00
Ave 285.66 285.14 284.84 286.10 286.48
Min 273.40 269.70 260.92 271.00 260.00
November Max 284.00 285.81 282.67 284.70 285.00
Ave 276.57 274.35 272.37 275.24 273.10
Min 270.40 269.00 269.86 271.54 271.00
December Max 282.00 280.00 280.41 281.94 282.00
Ave 275.93 274.20 274.90 276.52 276.17
Aegean Sea. The VPD values in Central Anatolia and South-Eastern being between 30 and
40 mb was attributed to the heating weather and, because of that heat, deficit humidity in
the atmosphere. Although enough water was present in the regions of the Mediterranean
and Aegean Sea coastal line, the cause of the VPD values being between 30 and 40 mb
was the heating weather in the early times of the day and, in spite of the humidity holding
capacity increase, there was not enough evaporation due to the lack of warming of the sea-
water. Even if the VPD values between 50 and 80 mb were not frequently observed over
Turkey, these rates were observed frequently over Iraq and Syria. In a similar way, VPD
values were calculated over all 24 satellite images.
Then, the cities of Adana, Ankara, Afyonkarahisar, Artvin, Antalya, Balıkesir, Denizli,
Erzurum, Eski¸sehir, ˙Istanbul, ˙Izmir, Kars, Kayseri, Konya, Malatya, Samsun, Sivas,
Downloadedby[SiirtUniversitesi]at00:4116January2013
13. International Journal of Remote Sensing 2725
320
(a) (b)
(c) (d)
320
310
310
300
300
290
290
Meteorologic temperature (K)
280
280
270
270
Temperature(K)
r = 0.958
Meteorologic temperature (K)
320
310
300
290
280
270
320310300290280270
Temperature(K)
r = 0.961
Meteorologic temperature (K)
320
310
300
290
280
270
320310300290280270
Temperature(K)
r = 0.967
Meteorologic temperature (K)
320
310
300
290
280
270
320310300290280270
Temperature(K)
r = 0.970
Figure 6. Correlation coefficients of the algorithms. (a) Price (1984), (b) Becker and Li (1990),
(c) Vidal (1991), and (d) Ulivieri et al. (1994) algorithms.
0 10 20 30 40 50 60 70 80 (mb)
Figure 7. Map of VPD obtained at 06.51 local time on 6 July 2002 (mb).
¸Sanlıurfa, Rize, and Van were chosen as control points for the satellite prediction accuracy
(see Figure 5).
Upon examining the monthly average VPD values at the control points, the following
were found.
Downloadedby[SiirtUniversitesi]at00:4116January2013
14. 2726 M. ¸Sahin et al.
100
90
90
80
80
70
70
60
60
50
50
40
40
30
30
Meteorologic VPD (mb)
r = 0.957
SatelliteVPD(mb)
20
20
10
10
0
0
Figure 8. Correlation coefficient of VPD.
• In January, the meteorological and satellite values were 1.9 and 4.2 mb, respectively.
• In February, the meteorological and satellite values were 6.35 and 8.72 mb,
respectively.
• In March, the meteorological and satellite values were 3.16 and 8.23 mb, respectively.
• In April, the meteorological and satellite values were 17 and 20.01 mb, respectively.
• In May, the meteorological and satellite values were 28.33 and 30.91 mb, respectively.
• In June, the meteorological and satellite values were 52.73 and 53.62 mb, respectively.
• In July, the meteorological and satellite values were 50.64 and 50.16 mb, respectively.
• In August, the meteorological and satellite values were 43.96 and 46.89 mb,
respectively.
• In September, the meteorological and satellite values were 26.19 and 28.29 mb,
respectively.
• In October, the meteorological and satellite values were 11.95 and 15.21 mb,
respectively.
• In November, the meteorological and satellite values were 2.49 and 5.45 mb,
respectively.
• In December, the meteorological and satellite values were 3.22 and 3.38 mb,
respectively.
Equations (13) and (14) tested the satellite prediction accuracy by calculating the cor-
relation coefficient (r) and RMSE. While the correlation coefficient on a monthly average
basis was 0.991, the RMSE value was 2.67 mb. When the meteorological and satellite VPD
values were not measured on a monthly average basis but were directly compared, the cor-
relation coefficient amongst the VPD values was found to be 0.957, and the value of RMSE
was 5.665 mb (see Figure 8). According to statistical rules, the VPD RMSE value can be
written as 6 mb instead of 5.665 mb.
Downloadedby[SiirtUniversitesi]at00:4116January2013
15. International Journal of Remote Sensing 2727
7. Discussion and conclusion
In the literature, Price (1984), Becker (1987), Becker and Li (1990, 1995), Vidal (1991),
Prata (1993, 1994), Sobrino et al. (1994, 1996), Coll et al. (1994), and Caselles, Coll, and
Valor (1997) attempted to obtain LST at reasonable accuracies (RMSE of 1–3 K) from
current operational and research NOAA/AVHRR satellite-borne visible/infrared radiome-
ters. In our study, the accuracies of split-window algorithms resulted in an average RMSE
value of 3 K (range 2.733–3.731 K). The Ulivieri et al. (1994) algorithm was found to be
very successful compared to studies from the literature. Because of this result, the Ulivieri
et al. (1994) algorithm was used to estimate the VPD formula. The VPD accuracy was
determined by the RMSE value and the correlation coefficient, which were calculated to
be 6 mb and 0.957, respectively. Furthermore, on a monthly average basis, while the VPD
correlation coefficient was found to be 0.991, RMSE was found to be 2.67 mb. These val-
ues are rather compatible with studies from the literature, which range between 3.2 mb and
10.9 mb.
As a result, we conclude that the VPD values obtained using satellite data can be used
in studies related to plants (germination, growth, and harvest), outbreak control of illness,
drought determination, and ET over wide areas in which the meteorological station network
density is normally not sufficient.
References
Andersson-Sköld, Y., D. Simpsonb, and V. Ødegaard. 2008. “Humidity Parameters From
Temperature: Test of a Simple Methodology for European Conditions.” International Journal
of Climatology 28: 961–72.
Becker, F. 1987. “The Impact of Spectral Emissivity on the Measurement of Land Surface
Temperature from a Satellite.” International Journal of Remote Sensing 11: 369–94.
Becker, F., and Z. L. Li. 1990. “Toward a Local Split Window Method Over Land Surface.”
International Journal of Remote Sensing 11: 369–93.
Becker, F., and Z. L. Li. 1995. “Surface Temperature and Emissivity at Various Scales: Definition,
Measurement and Related Problems.” Remote Sensing Review 12: 225–53.
Bouma, M. J., and H. J. Van Der Kaay. 1994. “Epidemic Malaria in India and the El Nino Southern
Oscillation.” The Lancet 344: 1638–9.
Bouma, M. J., and H. J. Van Der Kaay. 1996. “The El Nino Southern Oscillation and the Historic
Malaria Epidemics on the Indian Subcontinent and Sri Lanka: An Early Warming System for
Future Epidemics?” Tropical Medicine and International Health 1: 86–96.
Caselles, V., C. Coll, and E. Valor. 1997. “Land Surface Emissivity and Temperature Determination
in the Whole HAPEX-Sahel Area From AVHRR Data.” International Journal of Remote Sensing
18: 1009–27.
Castellvi, F., P. J. Perez, C. O. Stockle, and M. Ibañez. 1997. “Methods for Estimating Vapor Pressure
Deficit at a Regional Scale Depending on Data Availability.” Agricultural and Forest Meteorology
87: 243–52.
Castellvi, F., P. J. Perez, J. M. Villar, and J. I. Rosell. 1996. “Analysis of Methods for Estimating Vapor
Pressure Deficits and Relative Humidity.” Agricultural and Forest Meteorology 82: 29–45.
Choudhury, B. J. 1998. “Estimation of Vapor Pressure Deficit Over Land Surfaces From Satellite
Observations.” Advances in Space Research 22: 669–72.
Chrysoulakis, N., and C. Cartalis. 2002. “Improving the Estimation of Land Surface Temperature for
the Region of Greece: Adjustment of a Split Window Algorithm to Account for the Distribution
of Precipitable Water.” International Journal of Remote Sensing 23: 871–80.
Coll, C., and V. Caselles. 1997. “A Split-Window Algorithm for Land Surfaces Temperature from
Advanced Very High-Resolution Radiometer Data: Validation and Algorithm Comparison.”
Journal of Geophysical Research 102: 16697–713.
Coll, C., J. A. Sobrino, and E. Valor. 1994. “On the Atmospheric Dependence of the Split-Window
Equation for Land Surface Temperature.” International Journal of Remote Sensing 15: 105–22.
Cunningham, S. C. 2004. “Stomatal Sensitivity to Vapour Pressure Deficit of Temperate and Tropical
Evergreen Rainforest Trees of Australia.” Trees 18: 399–407.
Downloadedby[SiirtUniversitesi]at00:4116January2013
16. 2728 M. ¸Sahin et al.
Cunningham, S. C. 2005. “Photosynthetic Responses to Vapour Pressure Deficit in Temperate and
Tropical Evergreen Rainforest Trees of Australia.” Oecologia 142: 521–8.
Cunningham, S. C. 2006. “Effects of Vapour Pressure Deficit on Growth of Temperate and Tropical
Evergreen Rainforest Trees of Australia.” Acta Oecologica 30: 391–406.
Dash, P., F. M. Göttsche, F. S. Olesen, and H. Fischer. 2002. “Land Surface Temperature and
Emissivity Estimation from Passive Sensor Data: Theory and Practice-Current Trends I.”
International Journal of Remote Sensing 23: 2563–94.
Eck, T. F., and B. N. Holben. 1994. “AVHRR Split Window Temperature Differences and Total
Precipitable Water Ever Land Surfaces.” International Journal of Remote Sensing 15: 567–82.
François, C., and C. Ottle. 1996. “Atmospheric Correction in the Thermal Infrared: Global and Water
Vapor Dependent Split-Window Algorithms-Applications to ATSR and AVHRR Data.” IEEE
Transactions on Geoscience and Remote Sensing 34: 457–70.
Glenn, E. P., K. Morino, K. Didan, F. Jordan, K. C. Carroll, P. L. Nagler, K. Hultine, L. Sheader, and
J. Waugh. 2008. “Scaling Sap Flux Measurements of Grazed and Ungrazed Shrub Communities
with Fine and Coarse-Resolution Remote Sensing.” Ecohydrology 1: 316–29.
Green, R. M., and S. I. Hay. 2002. “The Potential of Pathfinder AVHRR Data for Providing Surrogate
Climatic Variables Across Africa and Europe for Epidemiological Applications.” Remote Sensing
of Environment 79: 166–75.
Habermann, G., E. C. Machado, J. D. Rodrigues, and C. L. Medina. 2003. “Gas Exchange Rates at
Different Vapor Pressure Deficits and Water Relations of Pera’ Sweet Orange Plants with Citrus
Variegated Chlorosis (CVC).” Scientia Horticulturae 98: 233–45.
Hashimoto, H., L. D. Jennifer, M. A. White, F. Yang, A. R. Michaelis, S. W. Running, and R. R.
Nemani. 2008. “Satellite-Based Estimation of Surface Vapor Pressure Deficits Using MODIS
Land Surface Temperature Data.” Remote Sensing of Environment 112: 142–55.
Hay, S. I., and J. J. Lennon. 1999. “Deriving Meteorological Variables Across Africa for the Study and
Control of Vector-Borne Disease: A Comparison of Remote Sensing and Spatial Interpolation of
Climate.” Tropical Medicine and International Health 4: 58–71.
Idso, S. B. 1982. “Non-Water-Stressed Baselines: A Key to Measuring and Interpreting Plant Water
Stress.” Agricultural Meteorology 27: 59–70.
Kant, Y., and K. V. S. Badarinath. 2000. “Studies on Land Surface Temperature Over Heterogeneous
Areas Using AVHRR Data.” International Journal of Remote Sensing 21: 1749–56.
Katsiabani, K., N. Adaktilou, and C. Cartalis. 2009. “A Generalised Methodology for Estimating
Land Surface Temperature for Non-Urban Areas of Greece Through the Combined Use
of NOAA–AVHRR Data and Ancillary Information.” Advances in Space Research 43:
930–40.
Kendall, M. A., and A. Stuart. 1963. “Distribution Theory.” In The Advanced Theory of Statistics,
edited by Griffin, Vol. 1, 1730. London: Charles Griffin & Company.
Kiyosawa, K. 2003. “Theoretical and Experimental Studies on Freezing Point Depression and Vapor
Pressure Deficit as Methods to Measure Osmotic Pressure of Aqueous Polyethylene Glycol and
Bovine Serum Albumin Solutions.” Biophysical Chemistry 104: 171–88.
Laurent, H., I. Jobard, and A. Toma. 1998. “Validation of Satellite and Ground-Based Estimates of
Precipitation Over the Sahel.” Atmospheric Research 47–48: 651–70.
Lendzion, J., and C. Leuschner. 2008. “Growth of European Beech (Fagus sylvatica L.) Saplings Is
Limited by Elevated Atmospheric Vapour Pressure Deficits.” Forest Ecology and Management
256: 648–55.
Leonardi, C., S. Guichard, and N. Bertin. 2000. “High Vapour Pressure Deficit Influences Growth,
Transpiration and Quality of Tomato Fruits.” Scientia Horticulturae 84: 285–96.
Linsay, S. W., L. Parson, and C. J. Thomas. 1998. “Mapping the Ranges and Relative Abundance
of the Two Principal African Malaria Vectors, Anopheles Gambiae Sensu Stricto and Anopheles
Arabiensis, Using Climate Data.” The Royal Society 265: 847–54.
Ouaidrari, H., S. N. Gowarda, K. P. Czajkowskib, J. A. Sobrinoc, and E. Vermotea. 2002. “Land
Surface Temperature Estimation from AVHRR Thermal Infrared Measurements an Assessment
for the AVHRR Land Pathfinder II Data Set.” Remote Sensing of Environment 81: 114–28.
Peñuelas, J., R. Savé, O. Marfà, and L. Serrano. 1992. “Remotely Measured Canopy Temperature
of Greenhouse Strawberries as Indicator of Water Status and Yield Under Mild and Very Mild
Water Stress Conditions.” Agricultural and Forest Meteorology 58: 63–77.
Pinheiro, A. C. T., R. Mahoney, J. L. Privette, and C. J. Tucker. 2006. “Development of a Daily Long
Term Record of NOAA-14 AVHRR Land Surface Temperature Over Africa.” Remote Sensing of
Environment 103: 153–64.
Downloadedby[SiirtUniversitesi]at00:4116January2013
17. International Journal of Remote Sensing 2729
Prata, A. J. 1993. “Land Surface Temperatures Derived from the Advanced Very High Resolution
Radiometer and the Along-Track Scanning Radiometer. I. Theory.” Journal of Geophysical
Research 98: 16689–702.
Prata, A. J. 1994. “Land Surface Temperatures Derived from the Advanced Very High Resolution
Radiometer and the Along-Track Scanning Radiometer. 2. Experimental Results and Validation
of AVHRR Algorithms.” Journal of Geophysical Research 99: 13025–58.
Prenger, J. J., and P. P. Ling. 2000. Greenhouse condensation control; understanding and using vapor
pressure deficit. Fact Sheet (Series) AEX–804, Ohio State University Extension, Columbus, 1–4.
Price, J. C. 1984. “Land Surface Temperature Measurements From the Split Window Channels of the
NOAA-7/AVHRR.” Journal of Geophysical Research 89: 7231–7.
Prince, S. D., S. J. Goetz, R. O. Dubayah, K. P. Czajkowski, and M. Thawley. 1998. “Inference
of Surface and Air Temperature, Atmospheric Precipitable Water and Vapor Pressure Deficit
Using Advanced Very High-Resolution Radiometer Satellite Observations: Comparison with
Field Observations.” Journal of Hydrology 212–213: 230–49.
Prince, S. D., and S. N. Goward. 1995. “Global Primary Production – A Remote Sensing Approach.”
Journal of Biogeography 22: 815–35.
Rogers, D. J., S. I. Hay, and M. J. Packer. 1996. “Predicting the Distribution of Tsetse Flies in West
Africa Using Temporal Fourier Processed Meteorological Satellite Data.” Annals of Tropical
Medicine and Parasitology 90: 225–41.
Seguin, B., and B. Itier. 1983. “Using Mid-Day Surface Temperature to Estimate Daily Evaporation
from Satellite Thermal Infrared Data.” International Journal of Remote Sensing 4: 371–83.
Shuttleworth, W. J. 1993. “Evaporation and Transpiration.” In Handbook of Hydrology, edited by
D. R. Maidment, 4.1–4.53. New York: McGraw-Hill.
Smith, W. L. 1966. “Note on the Relationship Between Total Precipitable Water and the Surface Dew
Point.” Journal of Applied Meteorology 5: 726–7.
Sobrino, J. A., Z. L. Li, M. P. Stoll, and F. Becker. 1994. “Improvements in the Split Window
Technique for Land Surface Temperature Determination.” IEEE Transactions on Geoscience and
Remote Sensing 32: 243–53.
Sobrino, J. A., Z. L. Li, M. P. Stoll, and F. Becker. 1996. “Multi-Channel and Multi-Angle Algorithms
for Estimating Sea and Land Surface Temperature with ATSR Data.” International Journal of
Remote Sensing 17: 2089–114.
Ulivieri, C., M. M. Castronuovo, R. Francioni, and A. Cardillo. 1994. “A Split-Window Algorithm
for Estimating Land Surface Temperature Satellites.” Advances in Space Research 14: 59–65.
Unwin, D. M. 1980. Humidity Calculations and Tables: In Microclimate Measurement for Ecologist,
71–82. London: London and Co.
Vazquez, D. P., F. J. O. Reyes, and L. A. Arboledas. 1997. “A Comparative Study of Algorithms
for Estimation of Land Surface Temperature From AVHRR.” Remote Sensing Environment 62:
215–22.
Vidal, A. 1991. “Atmospheric and Emissivity Correction of Land Surface Temperature Measured
From Satellite Using Ground Measurements or Satellite Data.” International Journal of Remote
Sensing 12: 2449–60.
Williams, L. E., and P. Baeza. 2007. “Relationships Among Ambient Temperature and Vapor
Pressure Deficit and Leaf and Stem Water Potentials of Fully Irrigated, Field-Grown Grapevines.”
American Journal of Enology and Viticulture 58: 173–81.
Downloadedby[SiirtUniversitesi]at00:4116January2013