? was justifi- 
FAP to an 
ty of several 
30 sq.km), NHA 
the ITC 
Settlement 
Reconnaissance 
ÏRIMENT 
эг to reduce 
aker, normally 
naissance 
st of fuel is 
as two engined 
ly US $ 675 per 
si SLX 55 x 55 
tervalometer 
c mounting 
nt. The first 
sd from ITC. 
y a navigation 
: a pilot, a 
undertook a 
Pathom (60 km 
The test was 
f the staff 
nn and the 
sed airphoto 
so satisfac- 
for the whole 
to 4 zones, 
tm (film 
ates. The 
a which was 
ilometer grids 
chart. This 
rate land marks 
follows that 
<is or E-W axis 
n lens on the 
8,500 ft, 
rcrews need 
nle ground 
Intervals were 
30% side lap on 
n 1/500 second, 
st at 17.12 
f 120 knots 
to automatic. 
sxposures. It 
/eloper, such 
clearly marked 
for the flight 
in an A4 size 
hours flying 
February 1985. 
a scale of 
ne interpreta- 
L985. 
{le mission, 
"inting to 
area of 5,000 
Bht. 32 (US$ 
set of conven- 
scale is Bht. 
5) per sq.km. 
trated an 
icies can now 
for develop 
ment planning purpose at low cost. Apart from the 
low cost factor there is the added convenience of 
handling the negative on any good quality 55 mm 
photo enlarger, and be able to provide photos to any 
scale required by different specialists involved in 
development planning. 
The SFAP airphotos have been used in the BMR study 
to collect data on various aspects: 
- the location of vacant land, potentially 
suitable for new residential development 
- a general land use inventory 
- a quantification of the BMR housing stock ac 
cording to various residential types. 
This paper focusses on the methodology and results 
of the housing stock inventory. 
5 A REVIEW OF AIR PHOTO INTERPRETATION METHODS FOR 
ESTIMATING POPULATION OR DWELLING UNITS. 
The study found that spatial and population 
characteristics in unplanned residential area are 
less homogeneous than in the planned residential 
area. Larger scale air photographs, with a large 
number of samples based on random selection of 
residential buildings, will be required in the 
unplanned residential areas, in order to obtain a 
fairly accurate population estimate. A stepwise 
regression analysis further showed that the density 
of buildings, and the average population per build 
ing, account for the greatest variation in 
population distribution. 
The influence of multi-family buildings has been 
analyzed by Watkin in 1984. He used airphotos at a 
scale of 20,000 and 1:6,000 to define the amount of 
error caused by multiple unit structure types. The 
result of Watkin*s study supports past findings that 
multiple unit structures were a primary cause of 
enumeration error, even with an increase in the 
scale of the photographs from 1:20,000 to 1:6,000. 
In the literature, quite a number of articles deal 
with the problem of counting houses from airphotos. 
Some relevant studies are shortly discussed below. 
Krause et al. (1974), measured the area of three 
dominant residential land use types from enlarged 70 
mm. B&W photos at the scale of 1:40,000, supple 
mented by 9 x 9 in colour infrared photos, scale of 
12,000, and 6,000, for four Californian cities, and 
determined the characteristic ’spatial population 
density’ associated with each residential land use. 
An estimated population for any area under inves 
tigation was calculated from the following 
function:- 
P A R1 ,D R1 + A RM‘ D RM + A RTP* D TP 
Where P = total estimated population 
ArI’ Arm’ ArTP = area devoted to each land use type 
and D, 
characteristic spatial den- 
Rl’ RM’ RTP 
sities associated with each land use types. 
The result showed that the land use map was 
generally accurate. The percent errors for the four 
cities ranged from an underestimate of 9.17% to an 
overestimate of 7.00%. The explanation was that the 
enlarged airphoto scale 1:40.000 made accurate 
measurement difficult and the wide variation of lot 
sizes deviated the characteristic spatial density. 
Other recent studies can be found by BOAPEH & 
TIPPLE (1983) and OLORUNFEMI (1984). 
Results of these studies can be summarized as 
follows: 
- The most common method to estimate the number of 
dwelling units or population was through land 
use in combination with housing unit count in 
selected sample area. Others have achieved only 
marginal success. 
- The degree of accuracy of the housing unit count 
depends on scale of photographs, and good inter 
pretation of residential buildings, while the 
land use method depends more on good land use 
interpretation, accurate measurement of area, 
and sound estimation of appropriate 
housing/population density. In addition, major 
limitations are the omission of small parcels 
due to the minimum area adopted for mapping 
purposes, and the inability to account for 
residential land uses above ground, e.g. high- 
rise buildings. 
6 METHODOLOGY APPLIED IN THE BMR HOUSING STOCK SURVEY 
6.1 Available materials. 
Adeniyi (1983) adopted the same technique as Kraus 
et al. to derive the total estimate of urban popula 
tion in Nigeria in combination with a stepwise 
regression analysis to evaluate the variable which 
account for variation in the characteristic popula 
tion density. The classification of residential 
areas was based on some local socio-cultural vari 
able including dwelling type, building usage, 
building type, number of storeys, density of build 
ing and other environmental variables, such as plot 
size, landscaping, lay out of building and the 
presence of garden and kiosks. Using these vari 
ables, the residential areas of Lagos were 
classified in to nine classes with the following 
example of class description. 
Example code 10: One-two storey (flat) building with 
vegetation, open space, average 
plot size (APS) 2000 m2, average 
building density (ABD) 5/ha 
These detailed classifications were delineated on 
airphoto scale 1:20,000. To derive population den 
sity, sample area representing 3.14 percent of the 
total residential land in 1974 was selected, based 
on stratified random sampling. Residential blocks 
were selected at random for each residential class. 
In each block, data about the number of persons per 
family and number of families per different type of 
building was collected. 
The following material was available for the Study: 
- B&W AP scale 1:15,000, year 1974, format 23x23 
cm. 
- SFAP enlarged to 1:15,000, year 1984, format 
18x18 cm. 
- Large scale police map of Bangkok for inner area 
covered the area of 346 sq.km at the scale of 
1:1,000 of year 1983 update by ground survey. 
- Enumeration district map. 
- Landuse map (ground survey) scale 1:50,000/1982. 
- Topographic map scale 1:50,000 year 1976. 
- Map of BMR showing district boundary scale 
1:50,000 year 1983. 
- Instruments: mirror stereoscope, 
pocket stereoscope, 
8x magnifier, 
dot grid for area measurement 
6.2 Primary sources of information 
In this study, there were two primary sources of 
information. In the inner city area (346 sqkm), data 
for the housing count were obtained from police map 
scale 1:1,000. It was found that newly built houses 
were not yet mapped, while in the informal settle 
ments mapping was in general incomplete, so 
airphotos were used to complement the work on the 
maps. Outside the large scale map area, data were 
obtained from air photographs.