Mapping diffuse photosynthetically active radiation from satellite data in Thailand

Available online at www.sciencedirect.com

ScienceDirect Advances in Space Research 60 (2017) 2345–2354 www.elsevier.com/locate/asr

Mapping diffuse photosynthetically active radiation from satellite data in Thailand P. Choosri a, S. Janjai a,⇑, M. Nunez b, S. Buntoung a, D. Charuchittipan a b

a Department of Physics, Faculty of Science, Silpakorn University, Nakhon Pathom 73000, Thailand School of Land and Food, Geography and Spatial Sciences Discipline, University of Tasmania, Hobart 7001, Australia

Received 19 February 2017; received in revised form 15 July 2017; accepted 1 September 2017 Available online 8 September 2017

Abstract In this paper, calculation of monthly average hourly diffuse photosynthetically active radiation (PAR) using satellite data is proposed. Diffuse PAR was analyzed at four stations in Thailand. A radiative transfer model was used for calculating the diffuse PAR for cloudless sky conditions. Differences between the diffuse PAR under all sky conditions obtained from the ground-based measurements and those from the model are representative of cloud effects. Two models are developed, one describing diffuse PAR only as a function of solar zenith angle, and the second one as a multiple linear regression with solar zenith angle and satellite reflectivity acting linearly and aerosol optical depth acting in logarithmic functions. When tested with an independent data set, the multiple regression model performed best with a higher coefficient of variance R2 (0.78 vs. 0.70), lower root mean square difference (RMSD) (12.92% vs. 13.05%) and the same mean bias difference (MBD) of
2.20%. Results from the multiple regression model are used to map diffuse PAR throughout the country as monthly averages of hourly data. Ó 2017 COSPAR. Published by Elsevier Ltd. All rights reserved.

Keywords: Mapping; Diffuse photosynthetically active radiation; Model; Satellite

1. Introduction Photosynthetically active radiation (PAR) forms one part of the solar spectrum with a wavelength range of 400–700 nm. This radiation is the main energy source of photosynthesis which is the start of the food chain. Plants use PAR for the photosynthesis and plant growth. Usually, the amount of PAR is measured as photosynthetic photon flux density, PPFD in mol s
1 m
2 unit (1 mol = 6.022  1023 photons). Normally, PAR incident at the earth’s surface consists of direct PAR and diffuse PAR and the summation of both components is called global PAR. The relative amount of direct and diffuse components of global PAR measured ⇑ Corresponding author.

E-mail address: [email protected] (S. Janjai). http://dx.doi.org/10.1016/j.asr.2017.09.001 0273-1177/Ó 2017 COSPAR. Published by Elsevier Ltd. All rights reserved.

at the earth’s surface varies depending on atmospheric parameters and geography. Atmospheric gases such as ozone and water vapor will absorb PAR, while clouds and aerosols are mainly scattering agents. Several researches have examined various depletion agents of PAR (Alados et al., 2000; Grant et al., 1996; Jacovides et al., 1997, 2007; Cho et al., 2003; Tripathy et al., 2015; Tanga et al., 2016), which are different for global and diffuse PAR. Diffuse PAR at the earth’s surface mainly depends on cloud, aerosol and solar zenith angle. These parameters are difficult to predict especially for clouds as they vary both spatially and temporally. Studies on global PAR are well covered in the literatures (Lo´pez et al., 2001; Hu et al., 2007; Janjai and Wattan, 2011; Wang et al., 2013; Janjai et al., 2013; Janjai et al., 2015a; Yu et al., 2015; Laliberte´ et al., 2016), but the research on diffuse PAR is very scarce as it is more difficult to measure

2346

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354

and predict (Alados et al., 2002; Wang et al., 2005; Xiaoli et al., 2014). Diffuse PAR is very important for plant growth in some environments where plants do not receive high intensity of direct beam from the sun. Other environments, such as high latitude polar regions may receive very little direct radiation and practically all the PAR radiation is diffuse. In Thailand, there is a predominance of diffuse PAR during the Southwest monsoon season when cloud depletes most of the direct beam radiation (Janjai et al., 2015b). Productivity is high with ample precipitation, although photosynthesis occurs mostly by diffuse PAR. Given the importance of PAR in photosynthesis, plant growth, and ocean biological productivity (Gates, 1980), it is important to examine how this term behaves, its dependence on external factors, and how it may be modelled. Towards this objective, we develop two regression models for calculating diffuse PAR using the radiation and related ancillary data from four stations located at four different climate regions of the country. The models are constructed from related data available at the four sites and then tested using an independent PAR data set. The best model is then used for mapping this variable over the entire country on a monthly average hourly basis. 2. Data acquisition 2.1. Ground-based measurements This analysis has used monthly average hourly data of diffuse PAR and atmospheric parameters. To measure diffuse PAR, quantum photon sensors (EKO, model ML-020P) were used. The sensors are silicon photodiodes which can filter only radiation in the wavelength interval 400–700 nm with a spectral response as shown in Fig. 1. The instruments were installed at four stations in Thailand, namely Chiang Mai (CM; 18.78°N, 98.98°E), Ubon Ratchathani (UB; 15.25°N, 104.87°E), Nakhon Pathom (NP; 13.82°N, 100.04°E) and Songkhla (SK; 7.20°N, 100.60°E) as shown in Fig. 2. Each photon sensor measuring diffuse PAR was installed on a sun tracker (Kipp&Zonen, model 2AP) on 1.5 m steel pole deployed on a building roof at each station and obstructed from direct solar radiation by a shaded ball. The voltage signals from the instrument was recorded every second by a data logger (Yokogawa, model DX2000). Data conversion to

Response of photon sensor

Spectral response

Solar spectrum (103 W/m2/nm)

Solar spectrum

Wavelength (nm)

Fig. 1. Spectral response of a photon sensor compared to solar spectrum.

diffuse PAR irradiance used the sensitivity of each instrument. Then values of the irradiance in mmol_ss
1s_ m
2 unit were averaged to obtain monthly average hourly data. The data were separated into two groups. The first group (from January 2011 to December 2013) was used for model formulation, while the second group (January–December 2014) was employed for model validation. Aerosol optical depth at 550 nm is used in this work as it is an important parameter which can both scatter and absorb solar radiation. The measurement of aerosol optical depth is rare in Thailand and as an alternative we use visibility data obtained from 84 meteorological stations around the country to calculate Angstrom’s turbidity coefficient using an empirical model developed by Janjai et al. (2003): b ¼ 0:589
0:068ðVISÞ þ 0:0019ðVISÞ

2

ð1Þ

where b is Angstrom’s turbidity coefficient and VIS is visibility (km). Afterward, aerosol optical depth at 550 nm is calculated by using a formula of Angstrom (1929). s ¼ bk
a

ð2Þ

where s is aerosol optical depth, a is wavelength exponent and k is wavelength (mm). Data of a from 106 sunphotometers of Aerosol Robotic Network (AERONET) located in South, East and Southeast Asia were gathered. Then they were spatially interpolated to cover all areas of Thailand and the values of the interpolated a were used in this work. Water vapor also has an effect on solar radiation as it can absorb solar radiation above 400 nm. The atmospheric water vapor can be estimated using the method of Janjai et al. (2005) as shown in Eq. (3). Relative humidity, air temperature and saturated vapor pressure data are used in this equation. These data are obtained from the 84 meteorological stations.   rh ps w ¼ 0:8933exp 0:1715 ð3Þ T where w is precipitable water in cm, rh is relative humidity in decimal, T is air temperature in K and ps is saturated vapor pressure in mbar. Those meteorological parameters can be used to obtain precipitable water at each station, and then the data from 84 meteorological stations were spatially interpolated to cover the entire Thailand region. Solar radiation reaching the earth’s surface is absorbed by stratospheric ozone especially in the short wavelengths. Although stratospheric ozone does not influence radiation in the PAR wavelengths much, it will still be considered in this work. Daily total ozone column over the globe can be retrieved from OMI/AURA satellite with spatial resolution of 1°latitude  1°longitude. These data can be downloaded from http://disc.scu.gsfc.nasa.gov/giovanni during 2006 to 2014, the same period of MTSAT-1R data. For this study, the daily ozone data from OMI/AURA satellite were averaged over individual months to obtain the monthly mean total ozone column. Prior to utilization, the pixels of the

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354

2347

Fig. 2. Diffuse PAR instruments at the four stations. A, B, C and D indicate northern region, northeastern region, central region and southern region, respectively.

ozone from this satellite were subdivided to match the pixel of MTSAT-1R for use in this work. Aerosol optical depth, precipitable water and total ozone column were used as input data in a radiative transfer model for calculating diffuse PAR under cloudless sky conditions for the entire area of the country and then used in mapping process. 2.2. Satellite data Cloud strongly scatters solar radiation, very often increasing the amount of diffuse solar radiation at the surface. Clouds vary markedly in time and space, and thus their effects are difficult to predict. In this analysis, earthatmospheric reflectivity from satellite data was used to represent cloud effects. In order to obtain cloud data covering the country, digital data from visible channel of Multifunctional Transport Satellite (MTSAT-1R) is used in this

work. The data were collected at 8:30, 9:30, 10:30, 11:30, 12:30, 13:30, 14:30, 15:30 and 16:30 h which is a total of 9 datasets per day. These digital data which cover entire area of Thailand were transformed into a cylindrical projection and then navigated by using points on the coastline as references (Fig. 3). Information in raw satellite data is in the form of gray level (0–255). The gray level of each satellite pixel was converted to an earth-atmospheric reflectivity using a conversion table supplied by the satellite agency (JMA, 2009). Finally, the reflectivity of each hour within each month were averaged so as to obtain a monthly average hourly reflectivity data. 3. Analysis In general terms, it is possible to argue that cloud cover and solar zenith angle are important factors influencing diffuse PAR at the surface (Iqbal, 1983). However, mod-

2348

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354

20.80o N, 96.0oE

850 pixels

20.80o N, 106.0oE

at the each hour within each month as shown in Fig. 4. The solar zenith angle at 8:30, 9:30, 10:30, 11:30, 12:30, 13:30, 14:30, 15:30 and 16:30 h were calculated for each month and then the data of each hour within a month were averaged. These monthly average hourly solar zenith angles were then used in the analysis. Fig. 4 shows that both diffuse PAR for cloudless (D0) and all sky (DC) conditions are sensitive to solar zenith angle, although the patterns are different. At low zenith angles, the difference between D0 and DC is large. Thus, we can say that variation of diffuse PAR is mainly from clouds and solar zenith angle, but with likely smaller contributions from aerosol optical depth. The all-sky measurement includes a specific cloud climatology for the particular month and hour and therefore represents contributions from both cloudless and cloudy periods. 3.2. Solar zenith angle dependence

4.90o N, 96.0oE

4.90o N, 106.0oE

550 pixels Fig. 3. An example of a rectified image.

The influence of solar zenith angle on the first data group containing DC from ground-based measurements was investigated. Best results were obtained with an exponential relationship containing the cosine of solar zenith angle as shown in Fig. 5. They show a relation with high R2 = 0.74. Therefore, it may be used as a radiative model for estimating diffuse PAR. DC ¼ 188:17 expð1:460 cos ZÞ ; R2 ¼ 0:74

elling of these parameters requires a more detailed knowledge of depletion agents and their relative importance. This is best accomplished using regression modelling provided the functional form that is known – whether the independent variables act in linear, logarithmic, exponential functions in affecting diffuse PAR. Therefore, much of the subsequent analysis is devoted to obtain these functional forms. 3.1. Cloud cover dependence The effects of cloud cover may be examined by comparing diffuse PAR under all sky and clear sky conditions. As no measurement of diffuse PAR for continuously cloudless conditions was available, the radiative transfer model libradtran-1.7 (Mayer et al., 1997) was used to calculate cloudless PAR, which was then compared against surface measurements at the four stations. A 2-streams (twostr) solver was selected for the PAR (400–700 nm) range with six streams (nstr). A tropical standard atmosphere was used in the calculation. Input parameters used for calculating diffuse PAR consisted of Angstrom parameters for aerosol optical depth (Eq. (2)), total ozone column, water vapor, day of year and solar zenith angle. Irradiance from the radiative transfer model was converted into a photon flux density by using factor, 1 W m
2 equal to 4.57 mmol s
1 m
2 (McCree, 1972). Resulting cloudless (from the model) and all sky (from the measurements) PAR was then plotted against solar zenith angle averaged

ð4Þ

where DC is diffuse PAR in mmol s
1 m
2 and Z is solar zenith angle in degrees. 3.3. Solar zenith angle, aerosol optical depth and cloud dependence In order to observe the effect of the geometrical factor and atmospheric parameters on diffuse PAR, an intermediate parameter d was created. It is defined to be the difference between diffuse PAR under all sky condition (DC) and diffuse PAR under clear sky condition (D0) as: d ¼ DC
D0

ð5Þ

For the case of geometrical effect, d was plotted against solar zenith angle (Z) and the result is shown in Fig. 6. Fig. 6 shows that d is sensitive to solar zenith angle (Z). It may be represented by a linear regression in Z with a statistical level of confidence being better than 95% and R2 = 0.47. d ¼ D C
D 0 ¼ a0 þ a1 Z

ð6Þ

where a0 and a1 are regression coefficients with their related statistics shown in Table 1. From Table 1, all regression coefficients are significant at 99% level of confidence. Therefore, Eq. (6) is useful to determine the difference between diffuse PAR under all sky and clear sky conditions. However, the data in Fig. 6 exhibit considerable scatter, indicating other processes influencing d mainly cloud cover and aerosols. Ozone and

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354

2349

1200

all sky condition Cloudless sky condition

Diffuse PAR (μmol·m-2·s-1)

1000

N = 2264 800

600

400

200

0 0

10

20

30

40

50

60

70

80

90

Solar zenith angle (degrees) Fig. 4. Relations between diffuse PAR under all sky () from the measurement data and clear sky (D) from the radiative transfer model and solar zenith angle. N is total number of data.

1000

1200

800

600

400

200

600

400

200

0

-200

0

0.00

y = -7.112x + 450.5 R² = 0.47 N = 1057

800

DC -D0 (μmol·m-2·s-1)

DC (μmol·m-2·s-1)

1000

y = 188.17e1.4606x R² = 0.7384 N=1057

-400

0.20

0.40

0.60

0.80

1.00

1.20

0

10

20

30

40

50

60

70

80

Z (degrees)

CosZ Fig. 5. A relation between diffuse PAR under all sky conditions (DC) from the ground-based measurement and cosine of solar zenith angle (CosZ). N is the total number of data.

Fig. 6. Differences between diffuse PAR under all sky and clear sky conditions (DC–D0 or d) were plotted with solar zenith angle (Z). N is the total number of data.

water vapor absorption have minimal effects and are not considered in this analysis. To observe the functional dependence of cloud depletion and aerosol optical depth on d, the following procedure is developed. We denote d0 as the deviation between any point in Fig. 6 and the regression line, and plot this difference d0 as a function of aerosol optical depth (AOD) and satellitederived earth-atmospheric reflectivity (qEA ) as shown in Fig. 7a and b, respectively From Fig. 7, d0 is related to AOD and qEA as logarithmic and linear functions, respectively. Therefore, from Figs. 6 and 7, the final form of the regression may be written as a multiple linear regression containing linear terms in Z and qEA and a logarithmic term to describe aerosol optical depth:

Table 1 Regression coefficients of parameters in Eq. (6). These were obtained using the statistical package of Excel.

DC
D0 ¼ a1 Z þ a2 qEA þ a3 lnðAODÞ

ð7Þ

Coefficient

Value

t-statistic

p-value

a0 a1

450.5035
7.1120

45.8274
30.5444

<0.001 <0.001

The first groups of data consisting in 1057 h were used in a multiple linear regression containing the solar zenith angle (Z), the reflectivity (qEA ) and the aerosol optical depth (AOD) as independent variables, and DC
D0 as dependent variables (Eq. (7)). Table 2 shows all the coefficients and their statistical significance. These were obtained using the statistical package of Microsoft Excel version 2010. From Table 2, all regression coefficients are significant at 99% level of confidence. According to this approach,

2350

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354 400.00

400.00

b)

300.00

300.00

200.00

200.00

100.00 0.00

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

-100.00 -200.00 -300.00 -400.00

100.00 0.00

0

0.1

0.2

0.3

0.4

0.5

0.6

-100.00 -200.00 -300.00

y = -557.93x + 149.96 R² = 0.145 N = 1057

-400.00

y = 214.4ln(x) + 352.57 R² = 0.5281 N = 1057

-500.00

δ' (μmol·m-2·s-1)

δ' (μmol·m-2·s-1)

a)

-500.00

-600.00

-600.00

AOD

ρEA

Fig. 7. The relation of d0 ðZÞ as a function of (a) aerosol optical depth (AOD) and (b) satellite-derived earth-atmospheric reflectivity (qEA ).

4. Model validation

Table 2 Regression coefficients of parameters in Eq. (7). Coefficient

Value

t-statistic

p-value

a1 a2 a3


6.099 249.979
209.347


42.9794 7.9869
49.7065

<0.001 <0.001 <0.001

the final form of the model for calculating diffuse PAR can be written as DC ¼ D0 þ a1 Z þ a2 qEA þ a3 lnðAODÞ

ð8Þ

The calculation of DC needs the value of D0 which is obtained from the radiative transfer model (Libradtran1.7) using the model input for the time that DC is intended to calculate.

The models described in Eq. (4) and (8) are then tested against the independent data for 2014. Fig. 8 shows the validation for Eq. (4) using only solar zenith angle. The accuracy of the models when compare with groundbased measurement data were assessed by root mean square difference (RMSD) and mean bias difference (MBD) (Iqbal, 1983). Results show that the diffuse PAR from simple model (Eq. (4)) agrees with diffuse PAR from the measurement with RMSD and MBD of 13.05% and
2.20%, respectively. Results for the model (Eq. (8)) which include solar zenith angle, aerosol optical depth and cloud effects are shown Fig. 9. From Fig. 9, the results show that diffuse PAR from the model agrees well with diffuse PAR from the measurement with an R2 of 0.78, an RMSD of 12.92% and a MBD of

1200

1200

1000

800

DC, model (μmol·m-2·s-1)

DC, model (μmol·m-2·s-1)

1000

600

400

RMSD = 13.05% MBD = -2.20% R2 = 0.70 y = 0.9621x N = 367

200

800

600

400

RMSD = 12.92% MBD = -2.2% R2 = 0.78 N= 367

200

0

0

200

400

600

800

1000

1200

DC,meas (μmol·m-2·s-1)

0 0

200

400

600

800

1000

1200

DC,meas (μmol·m-2 ·s-1)

Fig. 8. Comparisons between diffuse PAR from the simple model (Eq. (4)) (DC,model) and the measurements at the four stations in Thailand (DC,meas).

Fig. 9. Comparisons between diffuse PAR from model (Eq. (7)) (DC,model) and the measurement data at the 4 stations in Thailand (DC,meas).

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354

8:30

9:30

10:30

JAN 12:30

11:30

13:30

2351

14:30

15:30

16:30

20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

98°

100°

102°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

FEB 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

MAR 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

APR 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

98°

100°

102°

104°

MAY 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

JUN 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

320

160

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

640

480

800

960

1120

μmol·m-2·s-1 <80

240

400

560

720

880

1040

>1200

Fig. 10. Geographical distribution maps of long-term monthly average hourly diffuse PAR over Thailand during January to June (2006–2014).

2352

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354

8:30

9:30

10:30

JUL 12:30

11:30

13:30

14:30

15:30

16:30

20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

98°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

104°

AUG 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

SEP 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

OCT 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

98°

100°

102°

104°

NOV 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

DEC 20° 18°

16° 14°

12° 10°

8°

6°

96°

98°

100°

102°

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

320

160

104°

106°

98°

100°

102°

104°

106°

98°

100°

102°

640

480

800

960

1120

μmol·m-2·s-1 <80

240

400

560

720

880

1040

>1200

Fig. 11. Geographical distribution maps of long-term monthly average hourly diffuse PAR over Thailand during July to December (2006–2014).

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354


2.20%. These results show a better performance than Eq. (4) as they have a higher R2 (0.78 vs 0.70), a slightly lower RMSD (12.92% vs 13.05%) and the same MBD (
2.20%). Therefore, Eq. (8) was used to calculate DC over Thailand with results shown as PAR maps. 5. Mapping of diffuse PAR over Thailand The regressions developed in the previous sections performed well at all four stations. The methodology can be extended throughout the entire country as these stations represent different environments and climate regions in Thailand. In particular, satellite data on cloud cover is available throughout the region, as is aerosol optical depth, precipitable water vapor and total ozone column. These data were used as inputs in the Libradtran radiative transfer model to estimate diffuse PAR under clear sky conditions (D0) over the country. Therefore, Eq. (8) can calculate diffuse PAR over the country as monthly averages of hourly maps of diffuse PAR from 8:30 to 16:30 local time during 2006–2014 and then the diffuse PAR were averaged to long-term monthly average hourly. In addition, in order to verify the diffuse PAR retrieved from the mapping process, the value of diffuse PAR from the maps at the four ground-based stations were compared with those obtained from the measurements. The results showed a reasonable agreement with the RMSD and MBD of 14.9% and 1.1%, respectively. A total of 108 maps (9 h  12 months) of monthly average hourly diffuse PAR are presented in Figs. 10 and 11. Figs. 10 and 11 display both diurnal and seasonal variations in average hourly diffuse PAR. Diurnal variations feature a gradual increase from sunrise to noontime and then decreasing towards sunset. There is also a seasonal increase in diffuse PAR from January to April as a result of decreasing optical path length. During January the northeast monsoon begins, bringing cool dry air to the north, low cloud amounts and low levels of diffuse PAR. The summer season is prevalent from February to April, bringing high levels of aerosols (Janjai et al., 2012) and raising levels of diffuse PAR. The rainy season features southwest monsoon winds and is characterized by heavy cloud cover and intense rainy periods in the region from May to October (Janjai et al., 2015b). Diffuse PAR increases during this season while global PAR decreases because of cloud depletion. There are decreases in DPAR in the later months of November to December resulting from increasing solar zenith angles and optical path lengths. 6. Discussion and conclusion Our study is based on four years of diffuse PAR measurements using shaded balls to block direct beam radiation. In addition, a range of climate variables have been continuously recorded so as to examine their potential impact on the levels of diffuse PAR. These data have been

2353

used to build two regression models estimating monthly average hourly diffuse PAR
one relying only on solar zenith angle; the second one using solar zenith angle, earth-atmospheric reflectivity and aerosol optical depth. A total of three years of data were used to build up the regression models and the fourth year provided a validation data set. The best performance occurs when all three independent parameters (zenith angle, earth-atmospheric reflectivity, aerosol optical depth) are used as inputs into the regression model. The multiple linear regression model containing Z, qEA and AOD, were used to map month average hourly diffuse PAR throughout the country. These revealed significant diurnal and seasonal variabilities. Diffuse irradiance is highest at midday and decreases towards morning and evening. Southwest monsoonal clouds bring high levels of diffuse PAR. By contrast, the dry winter season feature low values due to the longer solar path lengths and absence of clouds. Acknowledgements The authors would like to thank the Thailand Research Fund (TRF) for providing financial support to this work under its International Research Network (IRN). Thai department of Meteorology is gratefully acknowledged for supporting the PAR measurements. References Alados, I., Foyo-Moreno, I., Olmob, F.J., Alados-Arboledas, L.Grupo de Fı´sica de la Atmo´sfera, 2002. Improved estimation of diffuse photosynthetically active radiation using two spectral models. Agric. For. Meteorol. 111, 1–12. Alados, I., Olmo, F.J., Foyo-Moreno, I., Alados-Arboledas, L., 2000. Estimation of photosynthetically active radiation under cloudy conditions. Agric. For. Meteorol. 102, 39–50. Angstrom, A., 1929. On the atmospheric transmission of sun radiation and on dust in the air. Geografis. Annal. 2, 156–166. Cho, H.K., Jeong, M.J., Kim, J., 2003. Dependence of diffuse photosynthetically active solar irradiance on total optical depth. J. Geophys. Res. 108. Gates, D.M., 1980. Biophysical Ecology. Springer-Verlag, New York, p. 611. Grant, H.R., Gordon, M.H., Gao, W., 1996. Photosynthetically-active radiation: sky radiance distributions under clear and overcast conditions. Agric. For. Meteorol. 82, 267–292. Hu, B., Wang, Y., Liu, G., 2007. Measurements and estimations of photosynthetically active radiation in Beijing. Atmos. Res. 85, 361– 371. Iqbal, M., 1983. An Introduction to Solar Radiation. Academic Press, New York. Jacovides, C.P., Tymvios, F.S., Assimakopoulos, V.D., Kaltsounides, N. A., 2007. The dependence of global and diffuse PAR radiation components on sky conditions at Athens, Greece. Agric. For. Meteorol. 143, 277–287. Jacovides, C.P., Timbios, F., Asimakopoulos, D.N., Steven, M.D., 1997. Urban aerosol and clear skies spectra for global and diffuse photosynthetically active radiation. Agric. For. Meteorol. 87, 91–104. Janjai, S., Kumharn, W., Laksanaboonsong, J., 2003. Determination of Angstrom’s turbidity coefficient over Thailand. Renew. Energy 28, 1685–1700.

2354

P. Choosri et al. / Advances in Space Research 60 (2017) 2345–2354

Janjai, S., Laksanaboonsong, J., Nunez, M., Thongsathitya, A., 2005. Development of a method for generating operational solar radiation maps from satellite data for a tropical environment. Sol. Energy 78, 739–751. Janjai, S., Nimnuan, P., Nunez, M., Buntoung, S., Cao, J., 2015b. An assessment of three satellite-based precipitation data sets as applied to the Thailand region. Phys. Geogr. 36 (4). Janjai, S., Nunez, M., Masiri, I., Wattan, R., Buntoung, S., Jantarach, T., Promsen, W., 2012. Aerosol optical properties at four sites in Thailand. Atmos. Climate Sci. 1, 441–453. Janjai, S., Sripradit, A., Wattan, R., Buntoung, S., Pattarapanitchai, S., Masiri, I., 2013. A simple semi-empirical model for the estimation of photosynthetically active radiation from satellite data in the tropics. Int. J. Photoenergy 2013 857072. Janjai, S., Wattan, R., 2011. Development of a model for the estimation of photosynthetically active radiation from geostationary satellite data in a tropical environment. Remote Sens. Environ. 115, 1680–1693. Janjai, S., Wattan, R., Sripradit, A., 2015a. Modeling the ratio of photosynthetically active radiation to broadband global solar radiation using ground and satellite-based data in the tropics. Adv. Space. Res. 56, 2356–2364. JMA, 2009. Conversion Table of Satellite Data. Japanese Meteorological Agency, Tokyo, Japan. Laliberte´, J., Be´langer, S., Frouin, R., 2016. Evaluation of satellite-based algorithms to estimate photosynthetically available radiation (PAR) reaching the ocean surface at high northern latitudes. Remote Sens. Environ. 184, 199–211. Lo´pez, G., Rubio, M.A., Martl´ınez, M., Batlles, F.J., 2001. Estimation of hourly global photosynthetically active radiation using artificial neural network models. Agric. For. Meteorol. 107, 279–291.

Mayer, B., Seckmeyer, G., Kylling, A., 1997. Systematic long-term comparison of spectral UV measurements and UVSPEC modeling results. J. Geophys. Res. 102 (D7), 8755–8767. http://dx.doi.org/ 10.1029/97JD00240. McCree, K.J., 1972. The action spectrum absorptance and quantum yield of photosynthesis in crop plants. Agric. Meteorol. 9, 191–216. Tanga, W., Qina, J., Yanga, K., Niua, X., Mind, M., Liang, S., 2016. An efficient algorithm for calculating photosynthetically active radiation with MODIS products. Remote Sens. Environ. 194, 146–154. Tripathy, M., Raman, M., Chauhan, P., 2015. Estimation of photosynthetically available radiation (PAR) from OCEANSAT-I OCM using a simple atmospheric radiative transfer model. Adv. Space. Res. 56, 1441–1452. Wang, L., Gong, W., Li, C., Lin, A., Hu, B., Ma, Y., 2013. Measurement and estimation of photosynthetically active radiation from 1961 to 2011 in Central China. Appl. Energy. 111, 1010–1017. Wang, Q., Tenhunen, J., Schmidt, M., Otieno, D., Kolcun, O., Droesler, M., 2005. Diffuse PAR irradiance under clear skies in complex alpine terrain. Agric. For. Meteorol. 128, 1–15. Xiaoli, R., Honglin, H., Li, Z., Guiriu, Y., 2014. Estimation of diffuse photosynthetically active radiation and the spatiotemporal variation analysis in China from 1981 to 2010. J. Geogr. Sci. 24, 579–592. Yu, X., Wub, Z., Jiang, W., Guo, X., 2015. Predicting daily photosynthetically active radiation from global solar radiation in the Contiguous United States. Energy Convers. Manage. 89, 71–82.