Global illuminance on the horizontal plane was calculated with the Perez correlation (Perez et al., 1990) for the effective illuminance of the global radiation on the horizontal plane.

Global illuminance on the on the four vertical planes was calculated using two methods, both developed by Perez.

The first method is based on the reconstruction of global incident illuminance on the tilted surface however orientated (Perez et al., 1990), (Cucumo et al., 1994), obtained from the sum of the direct illuminance, diffuse illuminance and incident reflected illuminance on the

surface, taking into account the anisotropy of the diffuse light, interpreted as the sum of three parts, the circumsolar part, coming from a region around the sun and striking the tilted surface with an incidence angle the same as that of the direct light, one part isotropic received evenly from the rest of the sky, and a third part called horizon illuminance, coming from a thin strip of sky adjacent to the horizon.

The starting datum for this calculation method is the global hourly incident solar radiation on the horizontal plane, a datum commonly measured in many places. By means of a resolution correlation (Erbs. et a/.,1982) the direct and diffuse hourly radiation datum is obtained from the global radiation. Applying the Perez correlations to the calculation of the effective illuminance of the direct and diffuse radiation on the horizontal plane, the values of direct and diffuse illuminance Eb0 and Ed0 are obtained on the horizontal plane.

The illuminance on the surface however orientated is calculable by means the equation:

The second method consists in calculating the illuminance of a surface however orientated as the sum of the direct illuminance from the sun, the diffuse illuminance from the sky calculated by integrating the luminance of the sky (Perez et a/., 1993) and the reflected illuminance from part of the surrounding buildings and land:

E = Eb + Ed + Er

All the calculation correlations used are reported in the appendix.

The data of calculated hourly illuminance using the two methods were compared with the data of the experimental hourly illuminance on the horizontal and vertical surfaces.

Using the former calculation method, the mean percentage deviations (і) were obtained as well as the mean square deviations (RMS) between the experimental and calculated values.

The percentage error for each datum was evaluated using the relation:

exp er

Tab/e 1 — Mean deviations and mean square deviations between the experimental and ca/cu/ated hourly illuminance using the first method (Perez 1).

The deviations, shown in table 1, show optimum agreement between the experimental and calculated values for the horizontal surface, whereas for the other surfaces and specifically for the South, West, North and East-facing surfaces there are notable differences.

Using the second method instead (Perez 2), the values shown in table 2 were obtained

Examination of the values in table 2 shows a slight improvement using the second method: the mean deviations relative to the horizontal surface and to those facing south and east almost coincide, while there are 8% and 5% improvements for the west and north-facing surfaces respectively; the coefficient of variation improves by 7% for the west­facing surface, by 14% for the north-facing surface and by 19% for the east-facing surface. For the west, north and east-facing surfaces, both the mean deviations and mean square deviations are still great.

The illuminance measurement instruments were then recalibrated, to exclude errors caused by measurement. Recalibration confirmed the previously recorded measurements.