You are on page 1of 23

int. j. geographical information science, 1997 , vol. 11 , no.

5 , 475 497

Research Article
M odelling topographic variation in solar radiation in a GIS environment
LALIT KUMAR, ANDREW K. SKIDMORE and EDMUND KNOWLES
School of Geography, University of New South Wales, Sydney, NSW 2052, Australia email: p2114659@geog.unsw.edu.au ( Received 9 May 1996; accepted 5 December 1996 ) Clear sky shortwave solar radiation varies in response to altitude and elevation, surface gradient (slope) and orientation (aspect), as well as position relative to neighbouring surfaces. While the measurement of radiation ux on a relatively at surface is straightforward, it requires a dense network of stations for mountainous terrain. The model presented here uses a digital elevation model to compute potential direct solar radiation and di use radiation over a large area, though the model may be modi ed to include parameters such as cloud cover and precipitable water content of the atmosphere. The purpose of this algorithm is for applied work in forestry, ecology, biology and agriculture where spatial variation of solar radiation is more important than calibrated values. The ability of the model to integrate radiation over long time periods in a computationally inexpensive manner enables it to be used for modelling radiation per se , or input into other hydrological, climatological or biological models. The model has been implemented for commercially available GIS ( viz. Arc Info and Genasys) and is available over the Internet.
Abstract.

1. Introduction

Solar radiation powers micrometeorological processes (such as soil heat ux and soil temperature), sensible heat ux, surface and air temperatures, wind and turbulent transport, evapotranspiration and growth and activity of plants and animals. In fact, 998 per cent of energy at the Earths surface comes from the Sun ( Dickinson and Cheremisino 1980 ). The total global radiation at the Earths surface consists of both short and longwave radiation. Shortwave radiation may be absorbed by terrestrial bodies and cloud cover and re-emitted as longwave radiation. The shortwave radiation reaching the surface of the Earth may be direct, di use or re ected ( gure 1). Direct radiation reaches the surface of the Earth from the solar beam without any interactions with particles in the atmosphere. Di use radiation is scattered out of the solar beam by gases ( Rayleigh scattering) and by aerosols (which include dust particles, as well as sulphate particles, soot, sea salt particles, pollen, etc.). Re ected radiation is mainly re ected from terrain and is therefore more important in mountainous areas. Direct shortwave radiation is the most important component of global radiation because it contributes the most to the energy balance and also the other components depend on it, either directly or indirectly ( Kondratyev 1965 ).
1365 8816/97 $1200

1997 Taylo r & Francis Ltd.

476

L . Kumar et al.

Figure 1.

Downward irradiance received in a mountainous region: (1 ) direct irradiance, ( 2) di use irradiance from the sky, and (3 ) terrain re ected irradiance.

Global radiation at a location is roughly proportional to direct solar radiation, and varies with the geometry of the receiving surface. The other components, such as di use radiation, vary only slightly from slope to slope within a small area and the variations can be linked to slope gradient ( Kondratyev 1965, Williams et al . 1972 ). In fact, di use radiation comprises less than 16 per cent of total irradiance at visible wavelengths in the green and red region ( Dubayah 1992), rising to 30 per cent for blue. The ux of clear-sky di use radiation varies with slope orientation much the same way as the ux of direct solar radiation, hence preserving the spatial variability in total radiation ( Dubayah et al . 1989 ). Parameters such as rainfall and temperature are frequently measured at one site and generalized for the surrounding region but extrapolating vertical atmospheric values, even using satellites, is di cult. Recording solar radiation at one site and extrapolating is seldom feasible as it is highly variable from place to place due to changing slope (surface gradient) and aspect (surface orientation). On at terrain and under clear-sky conditions, the downwelling shortwave radiation is nearly the same from point to point over relatively large areas and so one measurement can be taken to be representative of the entire area. However for mountainous terrain such point measurements do not adequately represent the shortwave radiation over large areas because mountain terrain sets up localized weather conditions ( Barry 1981 ) and so point samples are representative of only the locality from which they are collected. It is thus obvious that to get a reasonable accuracy in the measurement of incoming uxes in mountainous terrain one has to use either a very dense network of data collection stations or use approaches such as radiation modelling ( Duguay 1993 ). Solar radiation over large areas has often been estimated by measuring the number of hours of sunshine at a single site and then converting the hours into ngstrom equation radiation values by the use of empirical relationships, such as the A (A ngstrom 1924, Glover and McCulloch 1958), and generalizing this for the whole ngstrom equation is of the form area. The conventional A
Q / Q A = a + b n /N

( 1)

Sola r radiation modelling

477

where Q = total solar radiation on a horizontal surface on the Earth Q A= radiation on a horizontal surface at the upper limit of the atmosphere n =hours of bright sunshine N =total possible hours of sunshine, and a and b are constants found using regression. Bu o et al . ( 1972), Frank and Lee ( 1966) and Kondratyev ( 1969 ) have developed empirical relations between the e ects of slope exposure (e.g., gradient and aspect (Skidmore 1990 )) and clear sky radiation, but most of the results are in graphical or tabular form. This may be good for engineers, who require only point speci c data, but in forestry and ecological studies the variation in solar radiation over a study area is required. Swift ( 1976) modelled total shortwave radiation over a large area but it required measurements of solar radiation on a nearby site in addition to slope and aspect information. The shading e ects by adjacent features had to be visually estimated and manually calculated. Nunez ( 1980) developed a model which used cloud temperature data to obtain radiation uxes. Recently, Dubayah and Rich ( 1995 ) review the issues in modelling solar radiation for GIS. Di use sky irradiance under cloud free conditions may be estimated by assuming an isotropic sky, and calculating the proportion of the sky seen from a point (that is using the equivalent of the viewshed operation in GIS) ( Dubayah and Rich 1995). Under cloudy or partly cloudy conditions, di use radiation is anisotropic which may be explicitly modelled, but in practice this is computationally expensive to achieve as the di use radiation from di erent portions of the sky must be calculated. More importantly, in order to calculate actual solar ux, eld data such as pyranometer data (which measures actual incoming solar ux at a station), atmospheric optical data, or atmospheric pro ling (sounding) must be used. The aim of this research was to compute potential solar radiation (the amount of shortwave radiation received under clear-sky conditions), over a large area using only digital elevation and latitude data and to study the variation in radiation at di erent aspects and slopes. The model must cope with at or mountainous terrain, as well as shading by adjacent features. While the model calculates only the potential beam solar radiation and a simpli ed di use radiation, it can be modi ed to include other parameters such as e ects of cloud cover and precipitable water content of the atmosphere, if such data are available. It should be noted that for many parts of the earth, these data are not readily available, so we wished to develop a pragmatic approach to radiation modelling which may be utilized by GIS analysts. Thus the emphasis is on describing relative spatial variation in solar radiation, with modelled values as close to actual values as possible, rather than in exact values for validation and calibration purposes; relative solar radiation may be linked to the distribution of ora and fauna in the landscape, as well as productivity. Any model which is established in a given site using as input data records of classical weather parameters produced either by individual site installation or by a nearby meteorological station will only be applicable to that site. Therefore, while generating the model, an attempt has been made to include widely accepted empirical relations at the expense of probably more rigorous computational methods which require a lot of site speci c data. As has been mentioned previously, site speci c data for mountainous terrain is rarely available, hence inclusion of such parameters in the model would probably render it unusable for a lot of areas where it is actually meant to be used. However, where general empirical relations have been used other parameterizations are given and the reader is directed to relevant literature so that, where site speci c data is available, these relations could be utilized.

478
2. M ethod

L . Kumar et al.

2.1. Computatio nal procedures The intensity of solar radiation is a function of the solar direction relative to the local plane of the Earths surface at that instance. Variables such as solar azimuth angle and solar altitude angle change continuously throughout the day, and so have to be calculated every time the intensity of solar radiation is computed. Solar declination may be assumed to be constant and is calculated only once per day. To simplify the explanations, the Ptolemaic view of the Suns motion around the Earth will be used; in other words the site is xed and is the centre of origin. The fundamental equations used below are available from standard textbooks on solar engineering, such as Kreith and Kreider ( 1978), Du e and Beckman ( 1991) and Sayigh ( 1977 ). The Suns position in the sky is described by the solar altitude and solar azimuth angles. The solar altitude angle (a ) is the angular elevation of the Sun above the horizon ( gure 2). It is measured from the local horizontal plane upward to the centre of the sun. The solar altitude angle changes continuously; that is daily and seasonally. It is zero at sunrise each day, increases as the Sun rises and reaches a maximum at solar noon and then decreases again until it reaches zero at sunset. The noon solar altitude angle varies seasonally; the seasonal change being due to the declination angle changing daily. The solar azimuth angle (a s ) is measured in a horizontal plane between a due south line and the direction from the site to the sun as projected onto a horizontal plane ( gure 2). The solar altitude angle (a ) and solar azimuth angle (a s ) are related to the fundamental angles of latitude, solar declination (d s ) and hour angle (hs ) by equations (2 ) and ( 3 ) where L is the latitude of the site ( Kreith and Kreider 1978). sin a = sin L sin d s+ cos L cos d s cos hs sin a s= cos d s sin h s /cos a ( 2) ( 3)

The hour angle describes how far east or west the Sun is from the local meridian. It is zero when the Sun is on the meridian and decreases at a rate of 15 per hour. By convention morning values are positive and afternoon values are negative. Solar declination is the angle between the direction to the Sun and the plane of the Earths equator and is given in equation ( 4) ( Du e and Beckman 1991).
d s= 2345 sin ( 360 ( 284 + N )/365)

( 4)

Figure 2.

Illustration of solar altitude and solar azimuth angles.

Sola r radiation modelling

479

where N is the day number, 1 January being day 1 and 31 December being day 365. The declination varies from 2345 S to 2345 N. Again by convention values north of the equator are taken to be positive and those to the south are negative. 2.2. T he algorithm The owchart in gure 3 shows in detail the calculation of potential solar radiation at a site and the program, written in AML (Arc-Info) and Genasys script, is available at ftp site fatboy.geog. unsw.edu.a u (under solarradn directory). The program prompts for the latitude, day numbers when calculations are to begin and end, as well as the time interval to be used. Obviously if calculations are for one day only, the day number for the start and end of calculation will be the same and the program will integrate the values from sunrise to sunset. Therefore by just changing the day number, solar radiation for any length of time can be calculated. The user must specify latitude as a northern or southern value; northern values being positive and southern values being negative. The initial stage is to calculate the declination using equation ( 4 ). The solar hour

Figure 3.

Flowchart to calculate potential solar radiation.

480

L . Kumar et al.

angles at sunrise (hsr ) and sunset (hss ) for the day are also calculated. Sunrise is when solar altitude angle is zero degrees, (equations ( 5) and ( 6 )). sin a = 0=sin L sin d s+ cos L cos d s cos h sr (a = 0) ( 5) ( 6)

[ hsr= cos

( tan L tan d s )
h sr

and h ss=

Sunrise and sunset are the time when the centre of the sun is at the horizon, i.e., solar altitude angle is taken from the centre of the Sun. Fleming et al . ( 1995) have suggested the use of 08333 as the altitude angle for calculating the sunrise time as this takes atmospheric refraction into account. At this angle the top of the solar disc is just visible above the horizon and so starts illuminating the site, albeit with very weak rays. Once the sunrise and sunset times are calculated, solar radiation is integrated from sunrise to sunset. At this point it is necessary to decide on the time interval between solar irradiance calculations. This decision depends on both the accuracy required as well as the terrain, and must be balanced against computational expense. A rugged terrain will cause increased shading e ects as the Sun moves across the sky and so a smaller time interval should be used. For the example below, average solar radiation was calculated at the midpoint of a 30 minute interval; the midpoint being taken to represent the average of the ux over the time interval. The solar altitude and azimuth angles at the above hour angle are then calculated using equations ( 2 ) and ( 3 ). However, in using equation ( 3) to nd the azimuth angle, it is important to distinguish the case where the Sun is in the northern half of the sky from the case where it is in the southern half. If cos h s is greater ( less) than tan d s /tan L then the Sun is in the northern (southern) half of the sky and its azimuth angle is in the range 90 to 90 ( 90 to 270 ). In the next stage, shaded and illuminated points are calculated. The HILLSHADE command (in both Arc/Info and Genasys) creates a shaded relief grid by considering the illumination angle and shadows from a grid of elevation, solar azimuth angle and solar altitude angle. Shadows are de ned by the local horizon at each cell. In Arc/Info shadows are assigned a value of zero and illuminated cells are given a value of one, while in Genasys the values obtained have to be reclassi ed to get the shaded and illuminated cells. Solar radiation received at a site will depend upon the solar ux outside the atmosphere, the optical air mass, water vapour and aerosol content of the atmosphere. The solar ux (Io ) outside the atmosphere is given by Kreith and Kreider ( 1978 ) and Du e and Beckman ( 1991 ) (equation ( 7)).
Io = S o ( 1+00344 cos( 360 N /365))

( W m

( 7)

Equation ( 7 ) accounts for variation in the solar irradiance at the top of the atmosphere throughout the year as a result of the elliptical nature of the Earths path around the Sun. S o , the solar constant, is the irradiance of an area perpendicular to the Suns rays just outside the atmosphere and at the mean Sun Earth distance. The exact value of the solar constant is subject to much ongoing debate. A value of 1353 W m 2 has been accepted by NASA as a standard ( Jansen 1985) and the Smithsonian Institute also uses this value; however recent publications have cited values of 1367 W m 2 ( Duncan et al . 1982, Wehrli 1985 ) and 1373 W m 2 (Monteith and Unsworth 1990 ). The World Radiation Centre ( WRC) has adopted a value of

Sola r radiation modelling

481

1367 W m 2 ( Du e and Beckman 1991 ) and so this is the value used for calculations in this model. 2.2.1. Direct radiation Solar radiation is attenuated as it passes through the atmosphere and, in a simpli ed case, may be estimated using Bougers Law ( Kreith and Kreider 1978; equation ( 8 ))
Ib = Io e
kM

( 8)

where it is assumed that the sky is clear, Ib and Io are the terrestrial and extraterrestrial intensities of beam radiation, k is an absorption constant and M is the air mass ratio. The air mass ratio is the relative mass of air through which solar radiation must pass to reach the surface of the Earth. It is the ratio of actual path length mass to the mass when the Sun is directly overhead. It varies from M=1 when the Sun is overhead to about 30 when the Sun is at the horizon. The two main factors a ecting the air mass ratio are the direction of the path and the local altitude. The paths direction is described in terms of its zenith angle, y , which is the angle between the path and the zenith position directly overhead. The ratio M is proportional to sec y , which is equal to 1/cos y (Gates 1980 ). The adjustment in air mass for local altitude is made in terms of the local atmospheric pressure, p, and is de ned in equation ( 9) where p is the local pressure and Mo and po are the corresponding air mass and pressure at sea level.
M = ( p /p o ) Mo

( 9)

This above formula is valid only for zenith angles less than 70 ( Kreith and Kreider 1978 ). When the zenith angle is greater than 70 , the secant approximation underestimates solar energy because atmospheric refraction and the curvature of the Earth have not been accounted for. Frouin et al . ( 1989 ) have suggested the use of equation ( 10).
M = [cos y +015 ( 93885
y )
1253

( 10)

Keith and Kreider ( 1978 ) and Cartwright ( 1993 ) have suggested using the relationship in equation ( 11 ). The model presented here used equation ( 11) to calculate the value of air mass ratio.
M = [1229 + ( 614 sin a )2 ]1/2

614 sin a

( 11)

As the solar radiation passes through the Earths atmosphere it is modi ed due to: absorption by di erent gases in the atmosphere, molecular (or Rayleigh) scattering by the permanent gases, aerosol (Mie) scattering due to particulates.

Absorption by atmospheric molecules is a selective process that converts incoming energy to heat, and is mainly due to water, oxygen, ozone and carbon dioxide. Equations describing the absorption e ects of the above are given by Turner and Spencer ( 1972 ). A number of other gases absorb radiation but their e ects are relatively minor and for most practical purposes can be neglected ( Forster 1984 ). Atmospheric scattering can be either due to molecules of atmospheric gases or due to smoke, haze and fumes ( Richards 1993). Molecular scattering is considered to have a dependence inversely proportional to the fourth power of the wavelength

482

L . Kumar et al.

of radiation, i.e., l 4 . Thus the molecular scattering at 05 m m (visible blue) will be 16 times greater than at 10 m m (near-infrared ). As the primary constituents of the atmosphere and the thickness of the atmosphere remain essentially constant under clear sky conditions, molecular scattering can be considered constant for a particular wavelength. Aerosol scattering, on the other hand, is not constant and depends on the size and vertical distribution of particulates. It has been suggested (Monteith and Unsworth 1990) that a l 13 dependence can be used for continental regions. In an ideal clear atmosphere Rayleigh scattering is the only mechanism present ( Richards 1993 ) and it accounts for the blueness of the sky. The e ects of the atmosphere in absorbing and scattering solar radiation are variable with time as atmospheric conditions and the air mass ratio change. Atmospheric transmittance (t ) values vary with location and elevation and range between 0 and 1. According to Gates ( 1980 ) at very high elevations with extremely clear air t may be as high as 08, while for a clear sky with high turbidity it may be as low as 04. Hottel ( 1976 ) has presented a set of empirical equations for estimating beam radiation transmitted through clear atmosphere. The equations take into account the zenith angle as well as the altitude, and are stated in equation ( 12);
t b = a 0 + a 1 e
k/cos

( 12)

where t b is the atmospheric transmittance for beam radiation. a 0 , a 1 and k are constants and for the standard atmosphere with 23 km visibility and altitudes less than 25 km are given in equation ( 13 ).
a0 = 04237

000821 ( 6

A )2 A )2 A )2

a1 = 05055+ 000595 (65 k =02711+ 001858 ( 25

( 13)

where A is the altitude of the site in kilometres. Equations for a standard atmosphere with 5 km visibility are also given by Hottel. Kreith and Kreider ( 1978 ) have described the atmospheric transmittance for beam radiation by the empirical relationship given in equation ( 14 ).
t b= 056 (e
065 M

+ e

0095 M

( 14)

The constants account for attenuation of radiation by the di erent factors discussed above. Because scattering is wavelength dependent, the coe cients represent an average scattering over all wavelengths. This relationship gives the atmospheric transmittance for clear skies to within 3 per cent accuracy ( Kreith and Kreider 1978) and the relationship has also been used by Cartwright ( 1993 ). Equation ( 14 ) is used in the model instead of the Bouger form (equation ( 8)) as the Bouger form applies to a simpli ed case and does not account for factors considered in developing equations (14 ) and ( 15 ). Therefore the shortwave solar radiation striking a surface normal to the Suns rays (Is ) is given by equation ( 15). The atmospheric transmittance in the above equation can be replaced by site speci c values if they are available.
Is = Io t b

( 15)

The last stage is to calculate the solar radiation on a tilted surface (Ip ). This is given in equation ( 16), where cos i =sin d s (sin L cos b cos L sin b cos a w ) +

Sola r radiation modelling

483

cos d s cos h s (cos L cos b +sin L sin b cos a w )+cos d s sin b sin a w sin h s . i is the angle between the normal to the surface and the direction to the Sun and b is the tilt angle of the surface (slope) and aw is the azimuth angle of the surface (aspect). Slope and aspect of each grid cell are stored in separate grids, and accessed as required to calculate the solar radiation.
Ip = Is cos i

( 16)

2.2.2. Di use and re ected insolatio n Di use solar radiation (Id ) was calculated using the method suggested by Gates ( 1980 ), and is given by the equation ( 17).
Id = Io t d cos2 b /2 sin a

( 17)

where t d is the radiation di usion coe cient. t d can be related to t b by equation ( 18) ( Liu and Jordan 1960) which applies to clear sky conditions, and shows that the greater the direct solar beam transmittance, the smaller the transmittance to scattered skylight. Typical values of direct beam transmittance for a dust free, clear sky range from 0400 to 0800, and the corresponding di use skylight transmission varies from 0153 to 0037 (Gates 1980).
t d = 0271

0294 t b

( 18)

The magnitude of re ected radiation depends on the slope of the surface and the ground re ectance coe cient. The re ected radiation here is the ground-re ected radiation, both direct sunlight and di use skylight, impinging on the slope after being re ected from other surfaces visible above the slopes local horizon. The re ecting surfaces are considered to be Lambertian. Here re ected radiation (Ir ) was calculated based on equation ( 19) (Gates 1980)
I r = rI0 t r sin 2 b /2 sin a

( 19)

where r is the ground re ectance coe cient and t r is the re ectance transmittivity. The ground re ectance coe cient is the mean re ectivity of the surface over a speci c spectral band normalized by the full solar spectrum (Monteith and Unsworth 1990 ). t r can be related to t b by the relationship in equation ( 20) (Gates 1980).
t r= 0271+0706 t b

( 20)

A value of 020 was used for the re ectance coe cient of vegetation (Gates 1980 ). 2.3. Study site The algorithm described was used to compute the potential shortwave radiation at Nullica State Forest near Eden, New South Wales, at a latitude of 365 S. The terrain at the site is fairly rugged with elevation ranging from 9 to 880 metres. Figure 4 shows the location of the site while gure 5 gives a summary of the slope, aspect and elevation distributions. The total area of the study site was 16406 ha, made up of 182288 30 m by 30 m grid cells.
3. Results and discussion

3.1. Computatio n time The program was run on a Sun Sparc5 workstation. For each time interval it took about 235 seconds of CPU time to compute the solar radiation for the 182288 grid cells. Using a time interval of 30 minutes it takes 57 hours of CPU time to

484

L . Kumar et al.

Figure 4.

Location of study site.

calculate the solar radiation over the whole area for one year. However, because of the time taken for ancillary processes such as le transfers, the real time taken is about 10 times the CPU time. 3.2. Choice of time interval While it would be ideal to have a very short time interval to obtain accurate results, this is not always possible because of constraints such as time and the availability of a fast computer. The time interval chosen can be larger for terrain which is reasonably at but has to be small for rugged terrain where shadowing e ects by adjacent features will lead to signi cantly di erent results. The calculation of the radiation ux at the midpoint of the time interval somewhat reduces this e ect. For the study site, solar radiation was calculated using seven di erent time intervals for 23 September (equinox). The values for all the cells were then added and the mean and standard deviation calculated. The results are shown in table 1. While the mean values are similar, the standard deviation, minimum and maximum values show the wider range of values obtained at larger time intervals. To illustrate the impact of large time intervals, the mean deviation in solar

Sola r radiation modelling

485

Figure 5.

Cumulative frequency distributions for slope, aspect and elevation for study site.

486
Table 1. Time interval 60 45 30 15 10 5 1

L . Kumar et al.
E ect of decreasing time interval on insolation. Minimum 9478 9806 10087 9801 10334 10265 10262 Maximum 33689 33573 33460 33421 33398 33415 33410 Mean 25582 25554 25530 25534 25534 25532 25533 Standard deviation 2935 2868 2857 2828 2831 2825 2825

radiation was calculated for individual grid cells. For each interval and for each grid cell the actual solar radiation obtained was subtracted from that obtained using a time interval of one minute. All positive and negative deviations were separated and the mean deviation was calculated. The results ( gure 6) show that smaller time intervals give much more accurate results than larger time intervals. It also shows that overall there is no particular bias in the algorithm in overestimating or underestimating the radiation received, as both the positive and negative deviations are almost equal (assuming that the 1 minute interval readings are not biased). It should also be noted that results obtained using a 30 minute interval di er by only about one per cent when compared to those obtained using a one minute interval. 3.3. Radiatio n modelling Solar radiation was calculated every 30 minutes and integrated to give monthly values. Distribution of shortwave radiation over the study site for the whole year is shown in gure 7 and seasonal variations are shown in gure 8. Figure 7 shows a considerable range in shortwave radiation received over the area, the range being

Figure 6.

Mean deviations in insolation at each time interval from values obtained using one minute intervals.

Sola r radiation modelling

487

Figure 7.

Annual solar radiation distribution.

488

L . Kumar et al.

Figure 8.

Seasonal solar radiation distribution.

Sola r radiation modelling

489

from 3647 MJ m 2 to 11253 MJ m 2 per year. There are also substantial areas which receive no direct radiation at all during the winter season, as shown in gure 8. The months of November, December and January receive the most radiation and the radiation is fairly evenly spread over all aspects, especially for atter grid cells. The variation in solar radiation for di erent aspect and slope gradients are shown as gures 9 and 10. On an annual basis, northern facing slopes generally receive far more radiation than the south facing slopes. In many places the south facing slopes are fairly poorly irradiated, receiving only about half the shortwave radiation of the northern slopes. Such variations would surely have a signi cant e ect on the heat budget of di erent sites, thus in uencing latent and sensible heat uxes. At slopes greater than about 20 , radiation data have a higher variance with the northern slope receiving the highest radiation. It is also interesting to note that southern aspects receive a lot of radiation during the summer months. The reason for this is that during these months the hours of sunshine are almost equal between the northern and southern halves of the sky. The Sun rises in the southern half, and during the course of the day, traverses through the southern half, crossing over to the northern side around 820 am and then back to the southern side around 340 pm, before setting. In summer the Sun is also much higher in the sky, thus reducing the shadow e ects on southern aspects. Figure 11 shows the sunrise, sunset, daylength and the division of hours between the northern and southern halves of the sky. During the months of April to August the di erences in the amount of radiation received are pronounced ( gure 9). The radiation received on the south, south-east and south-west aspects decrease rapidly with increasing slope, with the di erences being larger on steep slopes. On slope gradients greater than 30 , southerly aspect sites receive only about 25 per cent of the radiation received on the north, north-east and north-west aspects. Figure 10 shows the radiation for each aspect by month. The northerly aspects (viz. north, north-west and north-east) receive relatively constant radiation regardless of slope gradient, while for south, south-east and south-west aspects less radiation is received on steep slopes except that in summer those aspects very near south receive slightly more radiation on steeper slopes. This e ect is more pronounced during the winter months when the sun is always in the northern half of the sky and is closer to the horizon. In other words, steep southerly slopes can expect to be in shadow more of the time. Figure 12 shows the annual radiation for the di erent aspects. As previously noted, north facing slopes receive the highest radiation throughout the year, though the radiation during the winter months of June and July is only about half of that received during the summer months of December and January. The variance in radiation for the southern aspect is the most. During December and January, the radiation for the southern aspect is very close to that of north aspect, at about 33 MJ m 2 day 1, but during winter falls to as low as 11 MJ m 2 day 1 , which is only a third of the summer value. Radiation received by the other aspects lies between the north and south aspects for most of the year, except in summer when they are generally less than the amount received by the southerly aspect. North east and north west aspects receive almost the same amount of radiation, as do east and west aspects, and south-east and south-west aspects. 3.3.1. Di use and re ected radiation Under clear skies, direct shortwave radiation (as discussed above) is the most important component of the total global radiation as it contributes the most to the

490

L . Kumar et al.

Figure 9.

Distribution on incoming solar radiation by month.

Sola r radiation modelling

491

Figure 10.

Variation of incident solar radiation by aspect.

energy balance. Apart from direct shortwave radiation, other components which contribute to the total radiation are the di use and re ected radiation. As the re ected insolation was fairly small it was added to the di use component and henceforth the term total di use radiation will be used to denote the sum of the two. Table 2 shows the magnitudes of the direct and di use radiation for the solstices and equinox and gure 13 shows the variation of both direct and di use radiation over the year. The total di use radiation is between 8 and 11 per cent of the direct insolation and lies within the range given in the literature ( Kondratyev 1969). Di use radiation

492

L . Kumar et al.

Figure 11.

Sunrise, sunset and daylength at study site.

Figure 12. Table 2.

Annual insolation on various aspects for study site.

Direct and di use radiation for the solstices and equinox. Direct Mean S.D. 1642 2925 4257
2

Di use Mean 2860 2340 1624 day


1.

S.D. 92 48 5

Summer solstice Equinox Winter solstice

31776 25545 15618

* Values for insolation are in kJ m S.D. = Standard Deviation.

Sola r radiation modelling

493

Figure 13.

Monthly variation of direct and di use radiation.

forms a higher percent of the total radiation during the winter months (approximately 11 per cent) while during the summer months it is about 7 per cent. The other point of interest is the variance, or the lack of it, in di use radiation, as seen from table 2 and gure 13. The small values of the standard deviation indicate that di use radiation is fairly uniform across the terrain, and that slope and aspect have very little e ect on it, contrary to the suggestion of Dubayah et al . ( 1989 ). This being the case, one wonders whether there is any justi cation in implementing highly complex models to model di use radiation, particularly when only the approximate solar radiation values are considered; unless the model is to be used for alpine environments, where snow covers (with a re ectance coe cient of around 08) will result in considerably higher amounts of re ected radiation impinging on surfaces. 3.4. Performance of model at other latitudes The model was tested at latitudes ranging from 1 to 60 south of the equator, by assuming the study site was located at these di erent latitudes. Figure 14 shows the annual variation of solar radiation at di erent latitudes, while gure 15 shows the solar radiation received at di erent aspects. The general trend is as expected, with high radiation in summer and low radiation in winter. Latitudes of 50 and 60 have reduced radiation during winter compared with equatorial latitudes. During the summer months, the possible radiation at higher latitudes outstrips that at equatorial latitudes because of longer daylengths at higher latitudes. While the ux of radiation at the equator is higher than that at other latitudes, the receipt of sunshine for much longer hours during summer months leads to the very high possible total radiation values. This result is similar to that reported elsewhere, for example Morse and Czarnecki ( 1958 ), Kondratyev ( 1969 ), Robinson ( 1966) and Du e and Beckman ( 1974). However it should be noted that these graphs are for clear sky conditions; the incoming radiation will decrease with decreasing

494

L . Kumar et al.

Figure 14.

Clear day mean solar radiation at various southern latitudes for the 21st of each month.

atmospheric transparency, and the higher the latitude the more pronounced is the decrease in incoming solar radiation ( Kondratyev 1969). The results obtained from this simulation exercise are consistent with published work (e.g. Morse and Czarnecki 1958, Kondratyev 1965). However, because the model integrates solar radiation from sunrise to sunset, care has to be taken in its use to calculate radiation for polar latitudes where daylengths run into days. The results obtained here will not be for single days but for the period from sunrise to sunset, for example at the south pole this would be around 186 days during summer.
4. Conclusion

A method to compute the potential solar radiation on any slope and aspect and at any latitude is described. Parameters such as the optical air mass, which has been taken to be constant, can be changed to account for the changing elevation as suggested by Garnier and Ohmura ( 1968 ) and Hottel ( 1976). The model could be enhanced to include information such as cloud type, percentage cover, thickness and altitude (Satterlund and Means 1978). However, as detailed data of this type are generally not available for forest areas, they have not been used here. The purpose of this algorithm is for applied work in forestry, ecology, biology and agriculture where the spatial variation of solar radiation is more important than calibrated values. The accuracy and the reliability with which a meteorological variable should be measured and operated vary with the purpose for which the measurement is required. In general, for most research purposes, the statement of accuracy is intended to ensure data compatibility in both space and time ( Dogniaux 1994 ); requirements which are met by this model. The potential to integrate solar radiation over long periods of time for large areas means that this model can be used in a variety of areas such as plant growth,

Sola r radiation modelling

495

Figure 15.

Solar radiation received by di erent aspects at various latitudes (south of the equator).

496

L . Kumar et al.

species location, water balance studies, biodiversity and identi cation of possible ora and fauna sites, although it is used here to map topographic variations in direct shortwave radiation. It must be added that, while the model successfully computes the solar radiation, the output will only be as good as the raw data supplied. In this case the models accuracy will greatly depend on the accuracy of the DEM used. Errors in the DEM will lead to errors in the computed values of aspect and slope and these have a direct e ect on the calculation of solar radiation. The other assumption of the atmosphere being uniform will have an insigni cant e ect on calculated radiation values as only clear sky radiation is being modelled.
Acknowledgement

This work was partly supported by an Australian Research Council Collaborative Grant with Genasys II Pty Ltd and the New South Wales Land Information Centre.
References
A ngstrom, A . K ., 1924, Solar and terrestrial radiation. Quarterly Journal of Royal Meteorological Society , 50, 121 125. B arry, R . G ., 1981, Mountai n Weather and Climate ( New York: Methuen). B uffo, J ., F ritschen, L . J ., and M urphy, J . L . , 1972, Direct solar radiation on various slopes from 0 to 60 North Latitude. U.S. Forest Service Paci c Northwest Forest Range Experimental Station Research Paper PN W-142 . C artwright, T . J ., 1993, Modelling the World in a Spreadsheet: Environmental Simulation on a Microcomputer (Baltimore: Johns Hopkins University Press). D ickinson, W . C . , and C heremisinoff, P . N ., 1980, Solar Energy T echnology Handbook D ogniaux, R ., 1994, Prediction of Solar Radiation in Areas with a Speci c Microclimate , D ubayah, R ., 1992, Estimating net solar radiation using Landsat Thematic Mapper and digital elevation data. Water Resources Research, 28, 2469 2484. D ubayah, R ., D ozier, J ., and D avis, F . , 1989, The distribution of clear-sky radiation over varying terrain, in Proceedings of International Geographic and Remote Sensing Symposium (Neuilly, France: European Space Agency) 2, pp. 885 888. D ubayah, R ., and R ich, P . M . , 1995, Topographic solar radiation models for GIS. International Journal of Geographical Information Systems, 9, 405 419. D uffie, J . A ., and B eckman, W . A ., 1974, Solar Energy T hermal Processes (New York: John D uffie, J . A ., and B eckman, W . A ., 1991, Solar Engineering of T hermal Processes (New York: D uguay, C . R ., 1993, Radiation modelling in mountainous terrain: review and status. Mountain Research and Development , 13, 339 357. D uncan, C . H ., W illson, R . C ., K endall, J . M ., H arrison, R . G ., and H ickey, J . R ., 1982, Latest rocket measurements of the solar constant. Solar Energy , 28, 385 390. F leming, P . M ., A ustin, M . P ., and N icholls, A . O ., 1995, Notes on a Radiation Index for use in studies of aspect e ects on radiation climate. CSIRO T echnical Bulletin (In Press) Division of Water Resources, Canberra. F orster, B . C . , 1984, Derivation of atmospheric correction procedures for LANDSAT MSS with particular reference to urban data. International Journal of Remote Sensing , 5, 799 817. F rank, E . C . , and L ee, R ., 1966, Potential solar beam irradiation on slopes: Tables for 30 to 50 latitude. U.S. Forest Services Rocky Mountai n Forest Range Experimental Station Paper R M-18 . F rouin, R ., L ingner, D . W ., G autier, C ., B aker, K . S ., and S mith, R . C . , 1989, A simple

( London: Butterworths) .

( Dordrecht: Kluwer Academic Publishers) .

Wiley & Sons).

John Wiley & Sons).

analytical formula to compute clear sky total and photosynthetically available solar irradiance at the ocean surface. Journal of Geophysical Research, 94, 9731 9742.

Sola r radiation modelling

497

G arnier, B . J ., and O hmura, A ., 1968, A method of calculating the direct shortwave radiation income of slopes. Journal of Applied Meteorology, 7, 796 800. G ates, D . M . , 1980, Biophysical Ecology ( New York: Springer-Verlag) . G lover, J . G ., and M c C ulloch, J . S . G ., 1958, The empirical relation between solar radiation and hours of bright sunshine in the high altitude tropics. Quarterly Journal of Royal Meteorological Society , 84, 56 60. H ottel, H . C . , 1976, A simple model for estimating the transmittance of direct solar radiation through clear atmospheres. Solar Energy , 18, 129 134. J ansen, T . J ., 1985, Solar Engineering T echnology (New Jersey: Prentice Hall). K ondratyev, K . Y a., 1965, Radiative Heat Exchange in the Atmosphere ( New York: K ondratyev, K . Y a., 1969, Radiation in the Atmosphere (New York: Academic Press). K reith, F . , and K reider, J . F . , 1978, Principles of Solar Engineering (New York: McGrawL iu, B . Y ., and J ordan, R . C . , 1960, The interrelationship and characteristic distribution of direct, di use and total solar radiation. Solar Energy , 4, 1 19. M onteith, J . L . , and U nsworth, M . H ., 1990, Principles of Environmental Physics ( London: M orse, R . N ., and C zarnecki, J . T . , 1958, Flat plate solar absorbers: T he e ect on incident radiation of inclination and orientation. CSIRO Report E.D. 6, Melbourne, Australia. N unez, M . , 1980, The calculation of solar and net radiation in mountainous terrain. Journal of Biogeography , 7, 173 186. R ichards, J . A ., 1993, Remote Sensing Digital Image Analysis: An Introduction ( Berlin: R obinson, N ., 1966, Solar Radiation (Amsterdam: Elsevier Publishing Company). S atterlund, D . R ., and M eans, J . E . , 1978, Estimating solar radiation under variable cloud conditions. Forest Science , 24, 363 373. S ayigh, A . A . M . , 1977, Solar Energy Engineering (New York: Academic Press). S kidmore, A . K ., 1990, Terrain position as mapped from a gridded digital elevation model. International Journal of Geographical Information Systems, 4, 33 49. S wift, L . W . , 1976, Algorithm for solar radiation on mountain slopes. Water Resources Research, 12, 108 112. T urner, R . E . , and S pencer, M . M . , 1972, Atmospheric model for correction of spacecraft data. In Proceedings of 8th International Symposium on Remote Sensing of the Environment , (Ann Arbor: E.R.I.M.), pp. 895 934. W ehrli, C . , 1985, Extra-terrestrial Solar Spectrum . Publication No. 615 ( Davos Dorf: World W illiams, L . D ., B arry, R . G ., and A ndrews, J . T . , 1972, Application of computed global radiation for areas of high relief. Journal of Applied Meteorology, 11, 526 533.

Pergamon Press). Hill ).

Edward Arnold ).

Springer Verlag).

Radiation Centre).