Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237683640

TERRAIN PARAMETERS IN SOLAR RADIATION MODELS An evaluation of


algorithms used for the extraction of morphological parameters from digital
elevation models and their impact in solar...

Article

CITATIONS READS

0 1,005

2 authors:

Fernanda Nery João Matos


European Environment Agency (EEA) Technical University of Lisbon
22 PUBLICATIONS   135 CITATIONS    32 PUBLICATIONS   128 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

USuET project View project

Interpolação Espacial de Dados Climatológicos View project

All content following this page was uploaded by Fernanda Nery on 18 December 2013.

The user has requested enhancement of the downloaded file.


TERRAIN PARAMETERS
IN SOLAR RADIATION MODELS
An evaluation of algorithms used for the extraction of morphological parameters from
digital elevation models and their impact in solar radiation modelling at different scales.

Fernanda NÉRY (1),(2) and João MATOS(2)


(1) (2)
Instituto Geográfico Português Instituto Superior Técnico

This research focuses on the topographic component of solar radiation models at


different analysis scales.

Two existing models (Kumar et al. 1997; Fu and Rich 2000) were compared using
synthetic and real surfaces, at local scales.

Kumar’s implementation was modified and tested for regional scales. The time
integration procedure was modified to correct the observed bias in estimated
radiation as a function of terrain aspect. New algorithms were used to incorporate
latitudinal gradients and convergence of meridians due to cartographic projection.

The change of spatial scope required new methods for the evaluation of focal
topography effects, since the original hillshading algorithm presents directional
artefacts due to the regular matrix data structure of the digital elevation model
(DEM) and does not account for Earth curvature. Similarly, different algorithms can
be used to calculate terrain gradient and minimise the smoothing effect of lower
spatial resolution DEMs.

Monthly potential radiation surfaces were produced for approximately half of the
Iberian Peninsula, using a low resolution DEM (1x1km). These surfaces will
subsequently be used as co-variables for the interpolation of meteorological fields
such as temperature.

This paper summarizes the different algorithms used in the modified model.

KEYWORDS
Solar radiation, clear-sky models, digital elevation model, morphological parameters, terrain gradient,
local horizon angle.

MODELLING SOLAR RADIATION

Solar radiation is commonly modelled as the sum of three components (Perez et al. 1986): direct beam
shortwave radiation, sky diffuse and ground reflected radiation.

At global scale, the temporal and spatial gradient of solar energy intercepted by the Earth is due to the
geometry of the planet and its rotation and revolution motion around the Sun. The available
extraterrestrial radiation can be calculated using astronomical formulas and the empirical solar
-2
constant (1367 Wm ).

Solar radiation models vary considerably in the complexity of their treatment of the atmosphere.
Atmospheric attenuation of incoming radiation is due to scattering and absorption by air molecules,
aerosols and clouds (Hofierka and Suri 2002), a combination of processes that selectively affects
different wavelengths (Bird and Riordan 1986). The turbidity values of the atmosphere vary seasonally
and are influenced regionally by the degree of urbanisation and industrialisation (Wald 2000). Cloud
cover is perhaps the most variable factor both in time and space.

However, the high spatial and temporal heterogeneity in local energy input is determined by factors
that operate at larger scales (Gates 1980; Fu and Rich 2000). Topographic variation is responsible for
local variation in air and soil temperature and moisture, local variation of vegetation type,
photosynthetic productivity, evapotranspiration and water balance (Gates 1980; Allen et al. 1998).
Digital elevation models can be used to calculate local parameters, such as terrain gradient magnitude
(slope) and direction (aspect), which affect mainly the direct radiation component. The effect of
surrounding topography on beam radiation can be modelled by evaluating shadows cast by nearby
elevations or calculating the local horizon angle for a given solar azimuth (Dozier and Frew 1990).

Isotropic diffuse radiation models may use an average local horizon, or sky-view factor (Iqbal 1983).
This can be reversed into a terrain-view factor and used for the reflected radiation component, normally
using a constant albedo (Gates 1980; Allen et al. 1998).

Again, focal parameters are of greater importance for beam radiation, although anisotropic diffuse
radiation models (Perez et al. 1986) may also incorporate obstruction effects into the geometric
description of the sky hemisphere using circumsolar and horizon bands.

This paper reviews some of the spatial solar radiation models with regard to solar position algorithms,
and their treatment of the topographic component. The atmospheric component is considered only with
respect to the treatment of solar altitude related effects upon atmospheric transmittance.

Spatial Solar Models Evaluation over Synthetic Surfaces


Two simple broadband models applicable to local scales were chosen for comparison:

• Kumar et al. (1997) implement Liu and Jordan (1960) model, as described by Gates (1980), using
functions available common commercial GIS to calculate terrain gradient and local horizon
obstruction to direct radiation. The Arc/Info coding of Kumar et al. (1997) solar model is currently
available at <http://www.wsl.ch/staff/niklaus.zimmermann/programs/aml.html>. Direct, diffuse and
reflected components can be integrated over daily periods. Hereafter, the model will be referred to
as LJGK1997.

• Fu and Rich (2000) developed Solar Analyst, which uses a simple transmission model for direct
radiation, and two alternative models for diffuse radiation. The reflected radiation component is not
estimated. Again, a commercial GIS system is used to calculate terrain gradient and local horizon
obstruction to both direct and diffuse radiation.

The two models use formulae with similar precision to determine the basic geometrical relationships
between the Sun and the Earth.

Atmospheric transmittance is averaged over wavelengths and only clear-sky conditions are evaluated
(although cloud attenuation can be subsequently overlaid). Unlike the r.sun model (Hofierka and Suri
2002) which uses Linke turbidity maps for different seasons, atmospheric attenuation is considered
constant over space and time when corrected for relative optical air mass. Diffuse radiation is modelled
using isotropic models (though Solar Analyst can consider the effect of solar zenith angle).

Topographic influence on solar radiation models can best be evaluated using artificial surfaces that can
be described by a mathematical model. The synthetic surfaces are then sampled to produce the raster
data structure used in digital elevation models. In this type of surfaces, the ‘true’ terrain value or
parameter (e.g. gradient) can be determined, which avoids uncertainty caused by data errors and
allows for patterns to emerge that would otherwise be masked by the complexity of real landscapes.

In this test, a very simple surface was build using Eqs.1, where z is the cell elevation, a is the distance
to the centre cell, b is the maximum distance to the centre cell and c is the maximum elevation at
centre cell. The following values were applied: cell size equal to 10m; b, 5000m; c, 500m.

c  π  
z =  cos a  + 1
2  b   Eq. 1

The resulting grid is a smooth low-gradient surface where horizon effects are likely to have little
influence over direct radiation (as compared with real landscapes).
With this approach, potential direct incident radiation was estimated using the two models (Kumar et al.
1997; Solar Analyst: Fu and Rich 2000), for different time-steps and periods of integration. For annual
periods, values were added over 60-min time steps and McKune and Keon (2002) empirical equations
were used for further comparison of geometric effects.

An east-west bias on the beam component was detected in both models, producing asymmetric
radiation totals in the symmetrical surface, due to the coding of the time integration procedure. As
would be expectable, the bias attenuates for shorter time steps, summer months, and longer
integration frames.

Extending LJGK1997 Model for Regional Scales


In this project, the LJGK1997 model was migrated to a different low-cost GIS platform and
programming language and modified for regional scales and simultaneous computation of the three
radiation components.

A more accurate solar position algorithm was selected for the determination of daily solar declination
and earth-sun distance. The time integration procedure was modified to correct the observed bias in
estimated radiation as a function of terrain aspect: variable time intervals centred on solar noon are
used. A varying latitude factor was incorporated into the determination of the sun’s position and an
additional factor allows the correction of the ‘convergence’ of meridians when using Transverse
Mercator projections (the angle between the direction of the meridian and the cartographic ‘North-
South’ direction at that point). This avoids the need for rotating or reprojecting the DEM data, which
usually involves a resampling procedure that smoothes topographic variation.

When working at regional scales, a lower resolution DEM is a computational necessity. However,
estimating terrain gradient using algorithms that use a 3x3 cells kernel may filter out local variation.
rd
By default, local gradient is calculated using 3 order finite differences weighted by the inverse
squared distances (Horn 1981).

However, if the DEM is produced by spatial aggregation of higher resolution information, then slope
magnitude and aspect can be similarly aggregated. Being a vector quantity, they are calculated from
the average gradient vector. The two alternatives were compared, using real surfaces, for daily
periods using the two extreme situations (Summer and Winter Solstice).

Finally, further analysis was carried on concerning the effect of terrain obstruction algorithms.
Synthetic data revealed the existence of a directional artefact in the commercial GIS hillshading
algorithm used by Kumar’s model. Additionally, the hillshading algorithm does not account for the
Earth curvature effect, and thus the influence of major mountain ranges propagates through unrealistic
distances.

Thus, in the modified model, the evaluation of horizon obstruction is done using an alternative
algorithm that estimates maximum obstruction systematically for angular sectors, attenuating the
influence of the regular grid cardinal directions and correcting for curvature effects.

The different algorithms and expression used in the modified model will be described in next three
sections:

• The Sun and Earth Geometry section describes solar position algorithms and the time integration
procedure in solar models.

• The Atmospheric Components section briefly describes path length influence on atmospheric
transmittance.

• The Sky and Ground Geometry section describes the effect of beam incidence angle on tilted
surfaces, common algorithms for estimating slope and aspect in gridded DEM, horizon effects and
sky-view and ground-view factors for diffuse and reflected radiation.

Finally, an annex lists the formulae used for the correction of cartographic azimuth in the Transverse
Mercator projection used for the Portuguese work area.
SUN AND EARTH GEOMETRY
This section briefly reviews the effect of solar position on incident radiation and presents some of the
available solar position algorithms. Solar position will be described using an equatorial system (Figure
1) and a local horizontal system (Figure 2):

• In the equatorial system, solar declination (δ) is measured on the star’s hour circle from the
equator to the solar vector - using positive values due north, and negative values due south - and
hour angle (ω) is measured on the equatorial plane from the observer's meridian to the projection
of the solar vector - using positive values before solar noon, and negative values in the afternoon.

• In the local horizontal system, solar zenith angle (θs) is the angle measured from the local zenith
to the sun, on the star’s hour circle. Solar altitude angle is the complement of solar zenith angle.
Solar azimuth angle (φs) is measured on the horizontal plane from south to the projection of the
solar vector. Solar azimuth values are positive or negative if oriented east or west from south,
respectively.

In spatial solar models, accuracy in the determination of solar position is traded for computational ease:

• Solar position is calculated for a "generalised" year, using a sequential day number from 1 to
365(366), instead of determining it for a particular Julian day. Likewise, correction parameters for
nutation or the precession of the equinoxes are ignored.

• Solar time can be used when irradiation is calculated for daily or larger time periods: hence the
conversion of civil to solar time using the equation of time is not required.

• Solar declination is computed only once a day.

However, high precision solar position algorithms are required and available for applications such as
pyranometer calibration (Reda and Andreas 2003) or solar concentrator tracking systems (Blanco-
Muriel et al. 2001). Reda and Andreas (2003) present a C implementation of Meeus (1998) algorithms
that allows the determination of solar azimuth and zenith angle with an uncertainty of ±0.0003º in the
period from the year –2000 to 6000. Other implementations are available in Java (Cornwall et al. 2004)
and VBA (Pelletier 2002). The latter will be used to evaluate algorithms used in spatial solar models
and select the most appropriate.

N
Zenith

Zenith Solar
vector
δ Solar
θs
vector
N

ϕ ϕ δ
Equator αs
ω
W E

φs

Figure 1: Solar position using an equatorial system: Figure 2: Solar position using a horizontal system:
geometrical relations between solar declination (δ), geometrical relations between solar zenith angle (θs), solar
solar hour angle (ω) and observer's latitude (ϕ). altitude angle (αS) and solar azimuth (φs).
Sun-Earth distance
The solar constant (Io) measures extraterrestrial radiation on a surface perpendicular to the sun's rays
located in the outer limit of the earth's atmosphere at mean Sun-Earth distance (Allen et al. 1998).
Currently accepted value is 1367 Wm-2 (∴0.08202 MJm-2min-1).

The density of solar radiation varies inversely with the square of the Sun-Earth distance (Gates 1980).
Given the eccentricity of earth's orbit, a correction factor (ε) must be applied to obtain the
extraterrestrial irradiance (Go) given a sequential day number N, from 1 to 365 or 366 (Chart 1).

Go = ε I o Eq. 2

Extraterrestrial irradiance varies a maximum of 0.3% (Chart 2) depending on the formulae that is
utilized (Table 1). The equation used by Hofierka and Suri (2002) presents the lower sum of squared
deviations to an average ε obtained from the different formulae. It will be used in the modified regional
version of the LJGK1997 model.

It must be noted that some models use tabulated values for I0 (ASHRAE 1999 in Wong and Chow,
2001). McKelley et al. (1999) use a polynomial fit to tabulated values of monthly average I0 versus
latitude, to determine the value used in the SRAD model. Since algebraic expressions are available
and daily calculation of Sun-Earth distance does not impact on computational performance, this
alternative was not considered.

Table 1: Formulae for the inverse earth-sun relative distance correction factor (ε).
st st
N is a sequential day number from 1 (January, 1 ) to 365 (December, 31 ).
All angles are given in radians, unless otherwise stated.

Correction factor (ε) Source


ε = 1.00011 Spencer 1971 in Bird and Riordan 1986

+ 0.034221cos D + 0.00128 sin D


+ 0.000719 cos 2 D + 0.000077 sin 2 D

D= ( N − 1)
where day angle is 365
 2π  Kreith and Kreider 1978 in Kumar et al.
ε = 1 + 0.0344 cos N 1997
 365 
2 Young and Vidal 1990 in Pons 1996
  2π
 1 + E cos (N − 4) 
ε =  365.25 
 1− E2 
 
 
where orbit eccentricity is E=0.01673
 2π  Allen et al. 1998
ε = 1 + 0.033 cos N
 365 
 2π  Hofierka and Suri 2002,
ε = 1 + 0.0344 cos N − 0.048869  Remund and Page 2002,
 365.25  Meteotest 2003
1,05 4,5

Maximum Variation in Go (Wm-2)


4,0
Inverse Relative Distance

3,5

1 3,0

2,5

2,0

0,95 1,5
10 90 180 270 360 10 90 180 270 360

Day number Day number

Chart 1: Seasonal variation in the inverse relative distance Chart 2: Maximum variation in calculated extraterrestrial
correction factor (ε). irradiance Go due to the formula used for (ε).

Solar declination and hour angle


Solar declination (Chart 3) values for solar noon were calculated according to Yallop (1978) and using
the simplified formulae proposed in the literature and listed in Table 2. Values were compared against
reference solar declination values computed using Pelletier's (2002) implementation of Meeus (1998),
for different years between 1990 (Chart 5) and 2010.

A simplified ranking of the available algorithms was made using the sum of square differences to
reference values. From smaller to larger differences, the ranking is: Bourges (1985 in Corripio 2003);
Gruter 1984 in Hofierka and Suri 2002; Duffie and Beckman 1991 in Kumar et al. 1997; Allen et al.
1998. Bourges (1985 in Corripio 2003) reports a mean error of 0.008º and maximum error of 0.02º,
values that are consistent with the ones obtained between 1990 and 2010.

In the reviewed spatial solar models, solar declination is computed only once a day (Kumar et al. 1997;
Corripio 2003). Considering that maximum daily variation in declination is less than 0.5º at the
equinoxes and less than 1' at the solstices (Chart 4), the cosine of solar zenith angle is estimated with
maximum uncertainty of 0.2% for the test latitude of 39º40'N (Chart 5). Bourges expression will be
used in the modified regional version of the LJGK1997 model.

Table 2: Formulae for solar declination (δ).


st st
N is a sequential day number from 1 (January, 1 ) to 365 (December, 31 ).

Solar declination Source


Bourges 1985 in Corripio 2003
 0.3723 
 
 + 23.2567 sin D − 0.758 cos D  π
δ =
+ 0.1149 sin 2 D + 0.3656 cos 2 D  180
 
 − 0.1712 sin 3D + 0.0201cos 3D 
 

D= (N − 79.346)
where 365.25
δ = sin −1 (0.3978 sin (D − 1.4 + 0.0355 sin (D − 0.0489))) Gruter 1984
in Hofierka and Suri 2002

D= N
where 365.25
 2π
(284 + N ) π
Duffie and Beckman 1991
δ = 23.45 sin in Kumar et al. 1997
 365  180
 2π  Allen et al. 1998
δ = 0.409 sin  N − 1.39 
 365 
25 0,50

Daily declination change (º)


12,5 0,25
Solar declination (º)

0 0,00

-12,5 -0,25

-25 -0,50
11 91 181 271 361 01 90 180 270 360
Day number Day number

Chart 3: Seasonal variation solar declination (δ). Chart 4: Daily change in solar declination (δ N - δ N-1).
0.25 0.5%

Difference in estimated beam irradiance


0.00 0.0%
Declination difference (º)

-0.25 -0.5%

-0.50 -1.0%

-0.75 -1.5%

-1.00 -2.0%

-1.25 -2.5%
01 90 180 270 360 10 90 180 270 360

Day number (Year 1990) Day number (Year 2010)


Yallop 1978 Gruter 1984 Yallop 1978 Gruter 1984
Bourges 1985 Duffie and Beckman 1991 Bourges 1985 Duffie and Beckman 1991
Allen et al. 1998 Allen et al. 1998

Chart 5: Difference in solar declination angle calculated Chart 6: Difference in estimated beam irradiance at
using different algorithms (for year 1990). solar noon for horizontal surfaces due to solar
declination algorithm (for year 2010 and ϕ=39º40’N).

Hour angle
The hour angle is defined as the angular distance along the celestial equator from the observer's
meridian to the hour circle of the sun (Figure 1). By convention, positive values will be used before
solar noon and negative values in the afternoon.

Sunrise is defined as the hour angle for which the altitude of the centre of the solar disc is zero (Table
3). This definition does not incorporate the effects of refraction nor the partial obscuration of the solar
disc (∅ =0.5336º). Fleming et al. (1995 in Kumar et al. 1997) suggest the use of αs=-0.8333º to
account for refraction effects. Fu and Rich (2000) incorporate penumbral effects, i.e., reduction of
beam radiation at the edge of shadow or local horizon.

Table 3: Formulae for solar hour angle (ω) at sunrise and sunset.

Parameter Formula
Hour angle at sunrise
   sin ϕ sin δ    
ω sr = cos −1  max − 1, min  − ,1

    cos ϕ cos δ    
= cos −1
(max(− 1, min((− tan ϕ tan δ ),1)))
Hour angle at sunset ω ss = −ω sr

In the tested spatial solar models, the time interval between solar irradiance calculations is established
by the user. The hour angle is calculated for the midpoint of each sampling interval in order to
determine solar zenith angle and azimuth. Irradiation is calculated for each sampling interval using mid-
point irradiance and integrated over daylight hours (Chart 7).

Kumar et al. (1997) use a sampling procedure that is not symmetric on solar noon. As a consequence,
the last sampling period for each day may include a variable period of time after sunset or else leave
unsampled the period just before sunset (Chart 8).
This effect can be corrected by centring sampling intervals on solar noon. A variable length interval
can be included at sunrise and sunset, while other intervals are kept at the user-required duration. An
alternative solution is to use a flexible time interval for each day and latitude, so that the user-required
duration is never exceeded (Chart 9). This option is used in the modified regional version of the
LJGK1997 model.

16 30

15

14 15

Time after sunset (min)


Daylight hours

13

12 0

11

10
-15

9
1
8
-30
0 90 180 270 360
01 90 180 270 360
Day number
Day number
36ºN 44ºN 36ºN 44ºN

Chart 7: Seasonal variation in daylight hours, for ϕ=36ºN Chart 8: Time after sunset included in (positive values),
and ϕ=40ºN (Iberian Peninsula work area). or time before sunset excluded of (negative values) the
last 60-min sampling interval, for ϕ=36ºN and ϕ=40ºN
(Iberian Peninsula work area).

60

58
Time interval (min)

56

54
01 90 180 270 360

Day number
36ºN 44ºN

Chart 9: Flexible time interval used when a maximum


60-min interval is specified in the modified algorithm, for
ϕ=36ºN and ϕ=40ºN (Iberian Peninsula work area).

Solar Zenith Angle and Solar Azimuth Angle

The specification of solar position in a local horizontal system (Figure 2) requires solar azimuth φs and
solar zenith angle (θs) or its complementary angle, solar altitude (αs) (Table 4).

Table 4: Formulae for solar zenith angle (θs), solar altitude (αs) and solar azimuth (φs).
Solar azimuth is measured from south (positive due east, negative due west).

Parameter Formula
Solar zenith angle
θ s = cos −1 (cos ϕ cos δ cos ω + sin ϕ sin δ )
Solar altitude π
αs = −θs
2
Solar azimuth
ω   sin ϕ cos θ s − sin δ 
φ s =   cos −1  
ω   cos ϕ sin θ s 
ATMOSPHERIC COMPONENTS
Atmospheric transmittance
The transmittance response of the different constituents of the atmosphere is not similar (Chart 10) and
is a function of radiation wavelength. Global and clear-sky direct radiation models vary in the number of
atmospheric attenuators that are modelled separately and in the use of broadband (Bird and Hulstrom
1981, Wong and Chow, 2001) or wavelength-specific transmittance functions (Bird and Riordan 1986).

Recent models (Rigollier et al. 2000; Hofierka & Suri, 2002; Remund and Page, 2002) use the Linke
turbidity factor TL which takes account of scattering and absorption by both the atmospheric aerosol
and the atmospheric gases (Kasten, 1996). In Europe, the average Linke turbidity factor is around 3.5
(Rigollier et al, 2000), but there is strong spatial and temporal variation, as can be seen in the 5’-of-arc
resolution TL maps available at http://www.helioclim.net/linke/linke_helioserve.html.

Since this work was focused on the topographic component, the two tested models use a single
broadband atmospheric transmittance value for beam irradiance in the zenith direction. Suggested or
default values are 0.8 (Kumar et al. 1997), 0.6 to 0.7 for very clear sky conditions (Fu and Rich 2000).
All runs were performed with a value of 0.8, which is an extremely clear and unrealistic atmosphere.

Relative optical air mass


The attenuation of beam solar radiation depends on the thickness of atmosphere it has to cross, i.e.,
on optical air mass. For small solar zenith angles, optical air mass can be estimated using the secant
of the angle (Figure 3) and correcting for atmospheric pressure, which is a function of site altitude and
temperature. For higher angles, the secant overestimates path length (Figure 4) due to Earth
curvature.

Zenith
1.0 CO2+O2 Sun
Transmittance

0.9
O3
0.8
H2O
0.7 θs
mzenith=1 m
0.6 Rayleigh
scattering
0.5

0.4 Total
molecular
Figure 3: Radiation path length estimated using the
0.3 secant of zenith angle (modified from Gates 1980).

0.2

Aerosol Zenith
Sun

mzenith=1
θs
m

0.1
Total
rEarth
60 76 80 83 84 85º θs

2 4 6 8 10 12 sec(θs)

Chart 10: Transmittance as a function of solar zenith Figure 4: Earth curvature’s influence on radiation path
angle (modified from Bird and Hulstrom 1981). length (modified from Gates 1980).
Table 5: Formulae for optical air mass (m) or equivalent path length modifier (m’).

Formula Source
p 1 p 1
m= =
p0 cosθ s p0 sin α s Gates 1980
This is a geometric interpretation without the refraction correction and
the Earth curvature correction.

( )
LJGK1997 model
m = 1229 + (614 sin α s )
2 0.5
− 614 sin α s Keith and Kreider 1978
in Kumar et al. 1997
1 Monteith and Unworth
m'= 0.00206 p
sin α s 1990 in Adaixo 1999

p 1
m'=
1000 sin α s
Atmospheric pressure [kPa] is estimated by (Burgman et al. 1985 in
Allen et al. 1998):
g
 T0,k − α 1 z  α1R
p = p 0  

 T0 , k  Allen et al. 1980
-1
where T0,k is sea-level temperature [K], g=9.807m.s is the acceleration
-1 -1 -1
of gravity, R=287 J.kg .K is the gas constant, and α1=0.0065 K.m is
the constant lapse rate. A simplification can be employed, assuming
T0,k= 20ºC= 297K and p0=101.3kPa:
5.26
 297 − 0.0065 z 
p = 101.3 
 297 
1
e (−0.00118 z −1.638⋅10 z )
−9 2
Solar Analyst
m=
cosθ s Fu and Rich 2000

p 1
m= ⋅
p0 sin α s ,c + 0.50572(57.29578α s ,c + 6.079995)−1.6364
with atmospheric pressure:
 −z  r.sun model
 
Kasten and Young 1989
p = p0 e  8435.2 
in Rigollier et al. 2000
and refraction corrected altitude:
0.1594 + 1.1230α s + 0.065656α s2
α s ,c = α s + 0.061359
1 + 28.9344α s + 277.3971α s2
SKY AND GROUND GEOMETRY
Solar Incidence Angle

For a horizontal surface, the beam or direct irradiance (Bh) is a function of solar zenith angle (θs) and
atmospheric attenuation (TB) of extraterrestrial radiation (Go).

Bh =GoTB cos(θ s )
Eq. 3

In the LJGK1997 model, atmospheric attenuation is modelled as

(
TB = 0.56 e (−0.65 m ) + e (−0.095 m ) ) Eq. 4

For tilted or inclined surfaces (Figure 5), beam or direct irradiance (Bi) is a function of solar incidence
angle (θi) (Table 6 ). Hence the surface’s slope (θt) and aspect (φt) must be determined (Figure 6). This
is trivial for solar panels but not for gridded terrain models.

Bi =GoTB cos(θ i )
Eq. 5

Table 6: Formulae for solar incidence angle (θi) for horizontal and tilted surfaces.

Parameter Formula
Incidence angle for cos θ i = sin δ (sin ϕ cos θ t - cos ϕ sin θ t cos φ t )
tilted surfaces
+ cos δ cos ω (cos ϕ cos θ t + sin ϕ sin θ t cos φ t )
+ cos δ sin θ t sinφ t sinω
Incidence angle for cos θ i = cos θs
horizontal surfaces
= cos δ cos ω cos ϕ + sin δ sin ϕ

Zenith

Normal
Normal
vector vector

θi Solar
vector

E
θt

φt
S

Figure 5: Solar incidence angle (θi) for tilted surfaces. Figure 6: Terrain slope (θt) and aspect (φt).
The origin of the coordinate system is actually the
observer’s point.
Terrain slope and aspect

Given any point in a surface z= f(x,y), slope (θt) and aspect (φt) are defined as a function of gradient in
the x-direction (fx) and in the y-direction (fy). Numerous algorithms have been proposed to estimate fx
and fy in gridded digital elevation models, most of which using a 3x3 roving window (Table 7 and Table
8).

The original LJGK1997 implementation uses the 3FDWRSD algorithm (Horn 1981), if the user does not
supply specific slope and aspect grids (Kumar et al. 1997). This is the default algorithm in both ESRI
GIS products and GRASS (Dunn and Hickey 1998), which support Solar Analyst (Fu and Rich 2000)
and the r.sun (Hofierka and Suri 2002) solar models, respectively.

Table 7: Formulae for terrain slope angle (θt) and aspect angle (φt).

Parameter Formula
Slope
θ t = arctan f x 2 + f y 2 
 
Aspect
  fy  π fx
arctan  − ,if f x ≠ 0
  fx  2 fx
φt = 0 ,if f x = 0 ∧ f y > 0
π , if f x = 0 ∧ f y < 0

not defined , if f x = 0 ∧ f y = 0

Table 8: Estimating gradient in grid models.

Acronym Description (Source) Formulae


2FD
a b c
Second-order finite differences
(Fleming and Hoffer 1979) fx =
(f − d )
d e f
(Equivalent to Zevenbergen and Thorne 1987 in 2rx
(b − h )
g h i
Jones 1998)
fy =
2ry
3FDWRSD
a b c
Third-order finite differences weighted by
reciprocal of squared distances (Horn 1981) fx =
(c - a ) + 2(f - d ) + (i - g )
d e f 8rx
(a - g ) + 2(b - h ) + (c - i )
g h i

fy =
8ry

For the 1km-grid spacing DEM used over the Portuguese conterminous territory, elevation was
calculated as the average of 1600 elevation values obtained from 25-m grid spacing DEM. Similarly
slope and aspect were calculated as the average gradient vector from 25-m grip spacing slope and
aspect grids.
Horizon effects
When the solar altitude angle is lower than the horizon angle for a given azimuth, there is no direct
incident radiation (Figure 7). This simplified approach neglects the partial obstruction of the solar disc
(∅s≅0.5º) and the effect of refraction or earth curvature.

In the LJGK1997 Arc/Info implementation, for each time period, solar altitude and azimuth is
determined, and hillshading effect is evaluated. Thus, when integrating radiation over a year period,
using 60-minutes time intervals, a total of 365x12=4380 evaluations are required, which is a significant
computational burden. If maximum upslope angle (max (θt↑)) is calculated beforehand, the number of
hillshade evaluations can be reduced, since there is no need to evaluate topographic obstruction when
solar altitude exceeds that value.

An alternative procedure is to evaluate focal obstruction for a given set of directions and interpolate
obstruction for the other azimuths (Figure 8). This procedure can be done in standard GIS hillshading
functions. Obtaining horizon angle for n directions, with a precision equivalent to the apparent solar
diameter, requires a total of (min( (max(θt↑)), (90-ϕ)+23.45) )⋅n/∅s runs. For instance, for ϕ=36º and
n=16, if max(θt↑)>77.45º then a maximum of 2480 runs are necessary. If the terrain has max(θt↑) = 45º,
the number reduces to 1440 runs. Horizon angles for intermediate azimuth direction can then be
interpolated and stored hence avoiding repeating the procedure during the daily integration. Note that
maximum daily change in solar azimuth at sunrise/sunset, for mid-latitudes, is ≅0.57º.

While this option requires little programming beyond what is already available as standard GIS
functions, it is extremely inefficient when compared with an adaptation of the R3 algorithm (Figure 10)
or especially with the simplified algorithm proposed by Dozier et al. (1981) (Figure 11).

W E

Figure 7: Horizon obstruction angle for three points along a W-E profile.

20 20
15 21 15 21

7 17 7 17

20 15 20 15

19 13 19 13

23 12 23 12

27 32 27
32
37 39 37
39 40
40
Figure 8: Horizon obstruction angle evaluated for 16 Figure 9: Horizon obstruction angle evaluated for 16
azimuths and interpolated for others (modified from Fu sectors (36 sectors were used in the modified model).
and Rich 2000). The dashed line represents average The dashed line represents average horizon
horizon obstruction angle. obstruction angle.
φs=60º

φs=80º
Figure 10: Interpolation of elevation at “grid crossings” Figure 11: Mapping or projection of grid nodes elevation
along the line of sight for which horizon angle is being to a nearest profile direction (for solar azimuths of 80º
determined, using the R3 algorithm. and -100º), using Dozier et al. (1981) algorithm.

For regional scales, the earth curvature effect must be accounted for when focal obstruction is
evaluated. Similarly, when latitudinal variation is incorporated, solar azimuth and altitude varies over
the territory for a given hour angle. Thus it is not possible to use traditional hillshading functions.

When implementing the modified regional version of the LJGK1997 model, a different approach was
chosen: horizon angles are evaluated as the maximum obstruction angle for a given angular sector
(Figure 9), corrected by earth curvature. Maximum search distance is heuristically chosen by
evaluating omni-directional obstruction, and finding a threshold distance such that increasing the
search radius beyond that distance affects maximum obstruction by less than 0.25º.

For each time period and latitude, solar azimuth and altitude are evaluated and compared against
horizon angle for the corresponding sector. Some other points should be noted:

• Refraction effects were not considered at this point, since that is not a purely geometrical effect,
and depends on altitude, atmospheric pressure and temperature.

• Earth curvature is modelled using a sphere of radius 6,370,997m.

• Given that maximum obstruction is assigned to the entire angular sector, a minimum of 16 sectors
is used to avoid overestimating horizon angle (36 10º sectors were used for the final model).

• Negative horizon angles are not considered (e.g. on mountain slopes).

• Directional artefacts are unavoidable given the DEM grid structure, but are less visible than in the
GIS built-in hillshade function.
Sky-view factor and ground-view factor
Diffuse irradiance can be described using an isotropic or anisotropic sky model.

For the LJGK1997 implementation, diffuse irradiance on inclined surfaces (Di) is modelled using
isotropic sky diffuse (Di,sky) and ground reflected components (Di,reflected) :

Di = Di , sky + Di ,reflected Eq. 6

Di , sky = (G0 cosθ i ) ⋅ TD ⋅ Fsky ,l Eq. 7

TD = 0.271 − 0.924TB Eq. 8

Di , reflected = (G0 cosθi ) ⋅ TR ⋅ Fground ,l ⋅ ρ Eq. 9

TR = 0.271 + 0.706TD Eq. 10

Where ρ is the ground albedo.

Note that Kumar et al. (1997, Eq. 18, pp.483) use Liu and Jordan’s (1960 in Gates 1980, Eq. 6.31
pp.115) formula for TD as function of transmittance to beam radiation in the zenith direction over the
entire range of solar altitude angles. Sky-view and ground view factors are evaluated using Gates
(1980) local formulas.

Table 9: Formulae for sky-view and ground-view factors.

Parameter Formula Source


Local sky-view factor
 θ  1 + cos θ t Gates 1980
Fsky ,l = cos2  t  = Perez et al 1986
2 2
Local ground-view
 θ  1 − cos θ t Gates 1980
factor Fground ,l = sin2  t  = Perez et al 1986
2 2
Focal sky-view factor Corripio 2003
Fsky ,f = sinθ h
where θ h is the average zenithal horizon angle
Focal ground-view Corripio 2003
factor
Fground ,f = 1 − sinθ h
where θ h is the average zenithal horizon angle

Ground albedo
The reflect component of diffuse radiation is neglected in Solar Analyst (Fu and Rich, 2000), thus
hindering its use in snow-covered areas.

In most small scale solar models, the ground albedo ρ is kept constant or at least it is not spatially
variable. Kumar et al (1980) use a value of 0.2, as proposed by Gates (1980). Allen et al. (1998)
indicate a 0.23 value from the grass reference cover in ETp calculation.

Remund and Page (2002) calculate monthly mean albedo as an empirical function of monthly mean
temperature (Eq. 11).

If used as a lumped parameter for the Portuguese territory, the predicted albedo varies between 0.2
and 0.23, as mean monthly temperature is approximately in the range between 8ºC and 22ºC.
However, if spatially disaggregated, the range of Tam increases to about 0ºC to 26ºC, hence an albedo
range of 0.41 to 0.2, which may lead to considerable differences in the winter months in higher regions.
 ρ + δ e (- 0.038 ⋅(Ta m + 3 ) ), if Ta > −3º C
2

ρ = 0 ρ m

0.5 Eq. 11

with ρ0 = 0 , δ ρ = 0.3
An alternative approach is to used land cover information (e.g. Corine Land Cover) and soil information
to obtain spatial disaggregated albedo values (Masson et al. 2003, Simpson et al. 2003) and further
use temporal information on growth period for annual cultures.

CONCLUSION
The modified model was tested in the Portuguese conterminous territory and part of the Spanish
territory (included because of topographic effects and other project requirements related to the location
of climatological stations).

Monthly potential radiation surfaces (direct, diffuse and reflected component) were produced using a
low resolution DEM (1x1km), but maintaining the local slope variability that determines radiation
variability (Dubayah et al. 1990).

Horizon obstruction angle grids for each sector are used in the modified LJGK1997 model, and can
similarly be used for different solar models. Sky-view and ground-view factors can be calculated using
a combined local and focal approach (Dozier and Frew, 1990).

Future work will involve the implementation of the ESRA model (Rigollier et al. 2000), is possible with
spatially and temporally variable Linke turbidity and albedo grids.
ANNEX – ‘CONVERGENCE’ OF THE MERIDIANS IN THE TRANSVERSE MERCATOR PROJECTION
The Portuguese military grid utilizes a Gauss-Krüger projection (the ellipsoidal version of the
Transverse Mercator projection) on the Hayford ellipsoid (a.k.a. International 1924). Required
parameters and applicable values are listed in Table 10.

Table 10: Parameters for the Portuguese military grid (Casaca et al. 2000).

Symbol Parameter Value


φ0 Latitude of true origin (reference latitude) 39º 40’ 0.0’’
λ0 Longitude of true origin (central meridian) -8º07’54.862’’
F0 Scale factor at central meridian 1
E0 Easting of true origin (false easting) 200000m
N0 Northing of false origin (false northing) 300000m
a Major semi-axis of ellipsoid (equatorial radius) 6378388m
b Minor semi-axis of ellipsoid (polar radius) 6356911.94612795m

Latitude and longitude were calculated for each DEM grid cell using the central point easting and
northing. Applicable formulae are listed in Table 11 (Ordnance Survey 1998; Casaca et al. 2000). The
formulae assume that linear measures have been scaled by F0 and that all angles are given in radians,
including φ0 and λ0.

The Gauss-Krüger or Transverse Mercator is a conformal projection that does not maintain true
directions. The ‘convergence’ of meridians (C) is the angle at any point in the projection between the
‘north-south’ grid line and the meridian at that point. C is equal to zero in the projection's Central
Meridian, where E = E0. C is negative to the West of the Central Meridian and is positive to the East
(Map 1).

Considering azimuth angles measured counterclockwise from North (Table 11), then:

TrueBearing = GridBearing - C

The correction is applied to terrain aspect:

TrueAspect = GridAspect - C

and also when converting solar azimuth (for horizon effects):

GridAzimuth = SolarAzimuth + C

Local angles are preserved in the Gauss-Krüger projection, but the straight line joining two points in
nature will normally be a curve when projected. The difference between the direction of the straight line
joining the two points on the projection and the initial direction of the curve is the (t-T) correction
(Ordnance Survey 1998). True azimuth is obtained by computing the cartographic bearing and
applying the convergence and the (t-T) correction. This correction is however negligible for the
majority of applications (see values in Chart 11 for the points in Map 2), and specially when compared
to the generalisations introduced by the digital terrain model grid structure and the algorithm for horizon
obstruction angle.
Map 1: 'Convergence' of the meridians for the tested area. Map 2: Points where the (t-T) correction was
calculated (see Chart 11).

E0 (km) N0 (km) E (km) N (km) C (DD) E 15


F 30

A 0 0 200 300 0
B 100 0 300 300 0.744 15

C 200 0 400 300 1.487


D 300 0 500 300 2.229
(t-T)
(t-T)

0 0
E 100 100 300 400 0.768
F 200 200 400 500 1.584 -15
G 300 300 500 600 2.451
-15 -30
0 90 180 270 360 0 90 180 270 360
t t

A 3
B 15
C 30

15
(t-T)

(t-T)

(t-T)

0 0 0

-15

-3 -15 -30
0 90 180 270 360 0 90 180 270 360 0 90 180 270 360
t t t

Chart 11: The (t-T) correction from each point in Map 2 to points 50km away. Point coordinates and (t-T) values in
arc seconds plotted against cartographic azimuth (t, in degrees, measured clockwise from North).
Table 11: Formulae utilized to calculate the convergence of meridians in a Transverse Mercator projection
(Ordnance Survey 1998).

Symbol Parameter Formulae


e Eccentricity
a2 − b2
e2 =
a2
M Arc of meridian from
 k1 (φ 2 − φ1 ) 
φ2 to φ1  
 + k 2 sin (φ2 − φ1 )cos(φ2 + φ1 ) 
M = M φ2 − M φ1 = b 
 + k 3 sin (φ2 − φ1 )cos (φ2 + φ1 )
2 2

 + k sin 3 (φ − φ )cos 3 (φ + φ ) 
 4 2 1 2 1 
where
5 5
k1 = 1 + n + n 2 + n 3
4 4
 21 
k 2 = − 3n + 3n 2 + n 3 
 8 
15 15
k3 = n 2 + n 3
8 8
35
k4 = − n3
24
a −b
n=
a+b
φ’ Latitude of the foot of
 N − N0 
the perpendicular φ' =   + φ0
drawn from a point on  a 
the projection to the
Do
central meridian
(
M = M φ ' − M φ0 )
 N − N0 − M 
φ'=   + φ'
 a 
Loop While ( N − N 0 − M < 0.00000001)
ν Radius of curvature at a
latitude φ ν=
perpendicular to a (1 − e 2
sin 2 φ )
12

meridian
ρ Radius of curvature of
a (1 − e 2 ) ν (1 − e 2 )
a meridian at latitude ρ= =
φ (1 − e 2
sin 2 φ ) 32
(1 − e2 sin 2 φ )
Et ‘True’ easting Et = E − E0
Nt ‘True’ northing Nt = N − N0
Table 11: (continued)

Symbol Parameter Formulae


φ Latitude
φ = φ '+ k 7 Et2 + k8 Et4 + k 9 Et6
where
tan φ '
k7 =
2 ρν
tan φ '
k8 = −
24 ρν 3
(
5 + 3 tan 2 φ '+η 2 − 9 tan 2 φ 'η 2 )
tan φ '
k9 =
720 ρν 5
(
61 + 90 tan 2 φ '+45 tan 4 φ ' )
ν
η2 = − 1 , with ν and ρ calculated at φ’
ρ
λ Longitude
λ = λ0 + k10 Et + k11 Et3 + k12 Et5 + k13 Et7 where
secφ '
k10 =
ν
sec φ '  ν 
k11 = − 3 
 + 2 tan 2 φ ' 
6ν  ρ 
secφ '
k12 =
120ν 5
( 5 + 28 tan 2 φ '+24 tan 4 φ ')
secφ '
k13 = −
5040ν 7
(61 + 662 tan 2 φ '+1340 tan 4 φ '+720 tan 6 φ ')
with ν calculated at φ’
C ‘Convergence’ of
the meridians on
C = k14 P + k15 P 3 + k16 P 5 where
the projection k14 = sin φ
sin φ cos 2 φ
k15 =
3
(
1 + 3η 2 + 2η 4 )
sin φ cos 4 φ
k16 =
3
(
2 − tan 2 φ )
P = λ − λ0
ν
η2 = − 1 , with ν and ρ calculated at φ
ρ
(t-T) Adjustment of N + N2
directions in the Calculateφ’ using N m = 1 in the initial estimate.
projection between 2
point (E1,N1) and
point (E2,N2)
(t1 − T1 ) = k17 (2 Et1 + Et 2 )(N1 − N 2 )
(t2 − T2 ) = k17 (2 Et 2 + Et1 )(N 2 − N1 )
1
k17 = , with ν and ρ calculated at φ’
6 ρν
REFERENCES
1. Adaixo, M., Evapotranspiração de referência. Validação de metodologias de cálculo para Portugal
Continental. Unpublished PhD Thesis. Universidade Técnica de Lisboa. Instituto Superior de
Agronomia. 283pp.

2. Allen, R.G.; Pereira, L.S.; Raes, D. and Smith, M., Crop evapotranspiration: Guidelines for
computing crop requirements. Irrigation and Drainage Paper No. 56, FAO, Rome, Italy, 300 pp,
1998

3. Bird, R. E., and Hulstrom, R. L., A Simplified Clear Sky Model for Direct and Diffuse Insolation on
Horizontal Surfaces. Solar Energy Research Institute. Colorado. SERI/TR-642-761, 46pp, 1981

4. Bird, R. E., and Riordan, C. J., Simple Solar Spectral Model for Direct and Diffuse Irradiance on
Horizontal and Tilted Planes at the Earth's Surface for Cloudless Atmospheres. Journal of Climate
and Applied Meteorology. 25 (1):87-97, 1986

5. Blanco-Muriel, M.; Alarcón-Padilla, D.; Lopez-Moratalla, T. and Lara-Coira, M., Computing the
solar vector. Solar Energy. 70 (5) :431-441, 2001

6. Burrough, P. A., Principles of geographical information systems for land resources assessment.
Clarendon Press, Oxford. 193 pp. 1986

7. Casaca, J.; Matos, J. and Baio, M., Topografia Geral, Lidel, Lisboa, 306pp, 2000

8. Cornwall, C.; Horiuchi, A. and Lehman, C. NOOA Solar Position Calculator.


http://www.srrb.noaa.gov/highlights/sunrise/azel.html, 2004

9. Corripio, J. G., Vectorial algebra algorithms for calculating terrain parameters from DEMs and the
position of the sun for solar radiation modelling in mountainous terrain. International Journal of
Geographical Information Science. 17(1) :1-23, 2003

10. Dozier, J.; Bruno,J. and Downey, P., A faster solution to the horizon problem. Computer and
Geosciences. 7 :141-151, 1981

11. Dozier, J. and Frew, J., Rapid calculation of terrain parameters for radiation modelling from digital
elevation data. IEEE Transactions on Geoscience and Remote Sensing. 28(5):963–969, 1990

12. Dubayah R., Dozier, J. and Davies, F. W., Topographic distribution of clear-sky radiation over
the Konza Prairie, Kansas. Water Resources Research 26 (4) :679-690, 1990

13. Dubayah R. and Rich P. M., Topographic solar radiation models for GIS. International Journal of
Geographical Information Systems. 9 :405-19, 1995

14. Dunn, M. and Hickey, R. The Effect of Slope Algorithms on Slope Estimates within a GIS.
Cartography. 27 (1):9-15, 1998

15. Florinsky, I., Accuracy of local topographic variables derived from digital elevation models.
International Journal of Geographical Information Science. 12 (1) :47-61, 1998

16. Fu P. and Rich P. M., The Solar Analyst 1.0 User Manual. Helios Environmental Modeling
Institute, http://www.hemisoft.com, 2000

17. Gates, D. M., Biophysical Ecology, Springer-Verlag, New York, 1980

18. Hofierka, J. and Suri, M. The clear-sky solar radiation model for Open source GIS:
implementation and applications. International GRASS users conference in Trento, Italy,
September, 2002

19. Horn, B., Hill-Shading and the Reflectance Map. Proceedings of the IEEE, 69(1) :14-47, 1981

20. Iqbal, M., An Introduction to Solar Radiation, Academic Press, Toronto, 1983
21. Jones, K., A comparison of algorithms used to compute hill slope as a property of the DEM.
Computer and Geosciences. 24 (4) :315-323, 1998

22. Jones, K., A Comparison of Two Approaches to Ranking Algorithms Used to Compute Hill
Slopes. GeoInformatica. 2(3) :235-256, 1998

23. Kasten F., The Linke turbidity factor based on improved values of the integral Rayleigh optical
thickness. Solar Energy. 56 (3) :239-244, 1996

24. Kumar, L.; Skidmore, A.K. and Knowles, E., Modelling topographic variation in solar radiation in
a GIS environment. International Journal of Geographical Information Science. 11(5): 475-497,
1997

25. Lee, S. and Clarke, K., An Assessment of Differences in Algorithms For Computing Fundamental
Topographic Parameters. Proceedings AutoCarto 2005. 20pp, 2005

26. Liu, B. and Jordan, R., The interrelationship and characteristic distribution of direct, diffuse, and
total solar radiation. Solar Energy. 4(3):1–19, 1960

27. Masson, V.; Champeaux, J.-L.; Chauvin, F.; Meriguet, C. and Lacaze, R. A Global Database of
Land Surface Parameters at 1-km Resolution in Meteorological and Climate Models. Journal of
Climate. 16 (9):1261–1282, 2003

28. McCune, B. and Keon, D., Equations for potential annual direct incident radiation and heat load.
Journal of Vegetation Science. 13 :603-606, 2002

29. McKenney, D. W.; Mackey, B. G. and Lavitz, B. L. Calibration and sensitivity analysis of a
spatially-distributed solar radiation model. International Journal of Geographical Information
Science. 13 (1) :49-65, 1999

30. Monteith, J.L. and Unsworth, Principles of environmental physics. Edward Arnold. New York,
1990

31. Ordnance Survey, The ellipsoid and the Transverse Mercator projection. Geodetic information
paper No.1.Version 2.2. Ordnance Survey. Southampton. 20 pp, 1998

32. Perez, R.; Steward, R.; Arbogast, C.; Seals, R. and Scott, J., An Anisotropic Hourly Diffuse
Radiation Model for Sloping Surfaces-Description, Performance Validation, and Site Dependency
Evaluation. Solar Energy 36 :481-98, 1986

33. Perez, R.; Seals, R.; Ineichen, P.; Stewart, R.; and Menicucci, D., A new simplified version of
the Perez diffuse irradiance model for tilted surfaces. Solar Energy 39(3):221-231, 1987

34. Perez, R.; Ineichen, P.; Seals, R.; Michalsky, J.; and Stewart, R. Modeling daylight availability
and irradiance components from direct and global irradiance. Solar Energy. 44 (5) :271-289, 1990

35. Reda, I. and Andreas, A., Solar Position Algorithm for Solar Radiation Applications, National
Renewable Energy Laboratory, NREL/TP-560-34302, 55pp, 2003

36. Remund, J. and Page, J.; Advanced parameters WP 5.2B: Chain of algorithms: short and
longwave radiation with associated temperature prediction resources. SoDa Project Deliverable
D5-2-2 and D5-2-3. June 2002. 72pp

37. Rigollier, C.; Bauer, o. and Wald, L., On the Clear Sky Model of the ESRA — European Solar
Radiation Atlas — with respect to the Heliosat method. Solar Energy 68 (1) :33–48, 2000

38. Simpson, D.; Fagerli, H.; Jonson, J.E; Tsyro, S.; Wind, P. and Tuovinen, J.-P., Transboundary
acidification and eutrophication and ground level ozone in Europe. Part I. Unified EMEP model
description. EMEP Status Report 2003. ISSN 0806-4520, 2003
39. Wald L., SODA: a project for the integration and exploitation of networked solar radiation
databases. European Geophysical Society Meeting, XXV General Assembly, Nice, France, 25-29
April, 2000

40. Wise, S., GIS data modelling—lessons from the analysis of DTMs. International Journal of
Geographical Information Science. 14 (4) :313-318, 2000

41. Wong, L.T. and Chow, W.K., Solar radiation model. Applied Energy. 69 :191–224, 2001

42. Yallop B.D., Formulae for computing astronomical data with hand-held calculators. National
Almanac Office, NAO Technical Note No. 46, 1978

43. Zhou, Q. and Liu, X., Analysis of errors of derived slope and aspect related to DEM data
properties. Computers and Geosciences. 30 :369–378, 2004

AUTHORS INFORMATION

Fernanda NERY João MATOS


fernanda.nery@igeo.pt jmatos@civil.ist.utl.pt
Instituto Geográfico Português Instituto Superior Técnico

View publication stats

You might also like