States (Table 2). Because a compact city with smaller plots and closely spaced buildings represents
high spatial frequencies, the resolution requirements for urban studies of China are much greater than
those for the United States (Figure 5; Welch, 1982). This is further confirmed by comparing the
current and anticipated IFOVs for various satellite sensor systems with the average urban parcel sizes
in different countries (Figure 6).
Table 2. Relative Sizes for Urban Built-up Areas of 50,000 to 500,000 Population.
COUNTRY RELATIVE SIZE
United States 1.0
Canada 0.60
Sweden 0.15
Japan 0.11
China 0.03
A^ U.S. IFOV
80m
+
PARCEL
ox
z US
o Sy
> CANADA so
SPATIAL FREQUENCY
1 CHINA 39
SWEDEN
5 JAPAN 20
a 15
> CHINA 10
P1
—
SPATIAL FREQUENCY ii
Figure 5. Representative spatial frequencies Figure 6. Comparison of urban land parcel
of urban land parcels in the United size with the IFOV's of satellite
States and China (Welch, 1982). sensor systems (Welch, 1982).
It is generally accepted that a minimum of about four IFOV/pixel sized units are required to
adequately represent a parcel (Figure 7). The intersection of the horizontal line representing four
IFOV/pixel units with the curves in Figure 7 defines the threshold IFOV or resolution requirements for
a sensor system designed to permit urban land use mapping at the parcel level. From this figure it
appears that IFOVs of 30, 25, 15, 10 and 5 m, are required to map urban land use in the United States, -
Canada, Sweden, Japan and China. These values, -although approximations, do indicate the variable mm
spatial resolution requirements for different geographic areas. -
Ye p———— eT CT
IFOV'S OR PIXELS/PARCEL
u
le À 1
o 5 10 15 20 25 30 35 40 45 50 ss so 6 7 75
IFOV OR PIXEL DIMENSION (m)
Figure 7. Number of IFOVs contained within representative urban land
parcel sizes as a function of IFOV (or pixel) dimension
(Welch, 1982).
It is of interest to note that a land use classification system has been developed in the United
States which groups land use classes into four successively finer levels of detail based on the sensor
data available for compilation (Anderson et al., 1976). By calculation, the resolution levels (IFOV)
required to extract land use information for the classes in Level I, II, III and IV categories are
approximately 80, 2, 1 and 0.5 m, respectively (Figure 8). These values, which are based on long
empirical experience, indicate that spatial resolutions of better than 5 m IFOV are required to
construct land use maps depicting Level II (or finer) class information. Interestingly, although
53