Albedos of terrain, mainly of asphalt concrete surfaces

I've measured the albedos, or strictly speaking, relative luminances, in Sunday June 5, 2003.

The technique used has been Canon EOS D60 camera with a 28mm to 80mm focal length objective and a sheet of white paper (a usual kind for laser printers). For comparison, the paper had grayscale steps printed on it using 2pt black squares (see gr_scale.eps).

The raw images have been converted by ``dcraw'' both to 8-bit deep bitmaps using gamma=1 and to raw uncompressed data in a easy-to-read format (as *.pgm).

From thumbnail jpegs (*.thm), text files containing the exif info on the images (the important one is the exposure) have been made using ``jhead''.

From the uncompressed data the radiometric tables have been computed by raw2lum, including the luminances corresponding to the median pixel illuminations. A grid with such data for each tile with a 64-pixel side has been written as an *.eps file and overlaid onto the bitmap using common programme ``composite''.

These images with an overlay have been inspected, the luminances of the white paper and of the various surfaces noted. Calibration of luminances is rather poor, but luminance ratios are reliable.

From the luminance ratios, the albedos have been computed assuming a very conservative value of white paper albedo as 0.87. Quite probably, the paper can have 0.93 in fact, and so the all the albedos could be multiplied by 1.07 (even 1.10 is not excluded). The paper appears very white, so its albedo is definitely not less than 0.87.

True albedos would be obtained, if the illumination of the surfaces would be isotropic, from a sky with uniform luminance. It has been not the case, even if the sky has been very cloudy. If the photographed surfaces have been near to me, their illumination has been dependent on the distance from me.

Moreover, for surfaces which are not perfectly diffusing, the luminance is influenced by a presence (or absence) of the semi-specular component of the dispersed light. Even it would not be a problem for a paper laid on another surface, if the angular distribution of the dispersed light would be the same for both surfaces. In some cases it is evidently not the case, as e.g. a asphalt concrete road can be more glossy then the paper (i.e., with a larger proportion of semi-specular dispersion).

(You can notice the luminance variations as you look on a surface from a homogeneous material, which is however not polished in the same degree everywhere. The most polished spots can be bright when viewed from some directions and dark from another ones. Tombstones with gravings in the otherwise specular surfaces are an example. Most usual tyre paths on heavily used roads are apparent this way, even if not expressed in topography of the road yet.)

The results I've obtained are as follows:

image Nr. description albedo*100

4 pavement at the front of our observatory 20

5 asphalt parking place by the observatory 15

6 pavement on the footpath to the square 22 asphalt by that footpath 15

7 another asphalt part of the footpath 13 cement concrete repaired spot there 32

8 asphalts (coarse and smooth, newer) 14.5

9 by a ``white'' zebra strip, held by a shoe asphalt concrete surface 15 cast pure asphalt strip 7 zebra strip 29 shoe-black rubber shoe 5 (depends on the chosen part)

10 most of the road surface 12 thin dark sew 9.6 shoe 4

11 two different asphalts and a sew upper asphalt surface 13.2 lower one 12.6 thin dark sew 8 to 9

12 recently abandoned zebra rests of zebra white 28 surface hidden below zebra before 9.6 11 exposed surface 12.6

13 (white paper oversaturated, wide-angle image of the scene from image 12)

14 just a year old surface 13 in distance (with more specular comp.?) 16 white middle strip 43 to 51

15 an old asphalt against dark sky 8.3 to 9.4 a new asphalt, not polished yet 11 sew between them 6.5 to 7 black shoe 4

16 the same old asphalt, more tangentially 12 a new asphalt, not polished yet 13 sew between them 11 black shoe 5

17 (the same view, same results)

18 the same place viewed in opposite direction new asphalt, near to paper 13.5 new asphalt, nearer to me 12 old asphalt, near to the paper 15 cement concrete edgestone 28 low grass, still green 11 dry low grass 15 sand soil 20

19 wide-angle image of the same scene new asphalt 14 old asphalt 15

So, apart from sews made by pure asphalt, the asphalt roads and paths have an albedo over ten per cent, seldom below 12 per cent, with 15 per cent being quite common. These values may be still one per cent point higher, if the calibration paper has over 90 per cent and not just 87.

Using a proper albedo standard of greater size (then just a strip left white on a paper) and taking pictures under uniform sky along the street, the numbers could be precised. However, one result is sure already: if the needed illumination of the streets (solving for sufficient luminances as given in the technical standards) has been computed under assumption of their albedo being just 7 per cent, this assumption has been proven as wrong. The actual albedo is at least 1.7 times larger, and the illumination could have been something like that amount fainter, with many benefits.

Of course, the ultimate proof of the existing overlighting will be to take images at night, with proper luminance calibration.

Another possible task can be solved taking wide-field images with a very clear sky and the Sun at varying heights. Complete BRDF (a function describing dispersion of light by the surface) can be inferred from such images.

jenik hollan

PS. the printed grayscale agrees quite well with the measured values. The actual black albedo is a function of the density chosen, 7 per cent may be a good guess for density 3 I've chosen for printing this sheet.

PPS. relative luminances can be measured (and hence, albedos inferred) using 8-bit deep images as well and some image-processing software. The only absolute necessity is to make them from raw images with gamma=1, so that the values are really proportional to the true pixel illuminations. Could anybody give me an advice, how to get a median (or average) value for a selected part of the image, by gimp or any other image software?