5 MANUAL INTERPRETATION 
Remote sensing technology is sometimes not 
considered in developing countries because of the 
lack of computer facilities for data processing. 
Computer, or digital, processing techniques are an 
important analysis tool in remote sensing, but 
they are not always necessary. In some 
situations, visual or manual interpretation of 
Landsat imagery is as accurate, and frequently 
more accurate than numerical analysis (Philipson, 
1980; Haack, 1983). In addition, visual 
techniques have the advantages of being less 
expensive and requiring less equipment and 
expertise than digital processing. It is 
important for scientists involved with remote 
sensing to be familiar with, and be willing to 
utilize, either digital or manual analysis 
techniques depending on their informational needs 
and available resources. 
The double sampling technique for assessing the 
extent of boro rice was also utilized by visual 
interpretation of an enlarged Landsat false color 
composite (FCC). The image study area was 
physically cut from the FCC and mounted in double 
glass slide binders for 35mm projection. This 
subscene was then projected onto a rearview screen 
and registered to a 1:63,360 scale topographic map 
of Baidya Bazar Thana. 
Four image interpreters with no familiarity of 
the study site mapped the boro for the ten Unions 
and determined the acreage of boro using a dot 
grid. The results from these interpretations are 
contained in Table 2. While there was 
considerable variability in the amount of boro 
mapped by the different interpreters, the 
correlation (Pearson's) of each interpreters 
results to the governmental supplied boro 
statistics were generally very good. These 
correlations were .88, .94, .82 and .91 with 
respect to columns 1 to 4 in Table 2. Certainly 
with images of improved quality, interpreters with 
greater familiarity of the site conditions, a 
later March date, and perhaps by using the 
improved spatial resolution data from the Landsat 
Thematic Mapper sensor or SPOT, the amount of boro 
identified by visual analysis would be closer to 
the reported amount; but the double-sample 
correlation may not be significantly improved. 
In any resource inventory using Landsat or other 
remotely sensing-' data, it is important to 
carefully evaluate the possible analysis 
techniques within the context of available data, 
informational needs and existing funds, equipment 
and expertise. This analysis indicates an ability 
to determine the amount of boro rice using a 
double sampling technique to be virtually as 
effective with a manual analysis as digital 
classification. The visual interpretation was 
inexpensive, required minimal equipment and 
expertise and provided good results. 
6 SUMMARY 
These initial results are promising but must be 
further investigated in other areas, and in 
regions with good ground truth. The greatest 
potential utility of Landsat data is probably the 
capability to analyze data over large regions to 
make regional acreage estimates, as well as to 
make maps showing the major boro growing regions 
and changes in these conditions over time. The 
logistics of collecting accurate field data and 
aggregating it to large areas is sufficiently 
difficult that present field-based techniques may 
not be adequate. Large area Landsat boro maps and 
adjusted Landsat estimates of boro acreage may 
provide useful, timely information to Bangladeshi 
planners. 
Table 2. Boro rice manual analysis double sampling 
results. 
Union Reported Manually identified 
acreage 
1 
2 
3 
4 
Aminpur 
Baradi and 
182 
240 
71 
120 
10 
Baidya Bazar 
610 
562 
231 
240 
94 
Jampur 
570 
541 
71 
220 
70 
Kachpur 
1140 
507 
587 
630 
191 
Mograpara 
330 
168 
36 
200 
10 
Noagaon 
580 
395 
302 
320 
10 
Pirijpur 
1190 
661 
507 
350 
160 
Sadipur 
420 
254 
36 
90 
0 
Sambhupura 
1410 
848 
951 
680 
241 
Sanmandi 
855 
523 
444 
680 
211 
REFERENCES 
Chaudhury, M. U. et al. 1978. A Landsat inventory 
of the agricultural and forest resources in 
Bangladesh. Proceedings: Twelfth International 
Symposium on Remote Sensing of Environment. 
1391-1400. 
Colwell, J. et al. 1978. Use of Landsat data to 
assess waterfowl habitat quality. Ann Arbor, 
Michigan: Environmental Research Institute of 
Michigan. 
Colwell, J. et al. 1977. Wheat yield forecasts 
using Landsat data. Proceedings: Eleventh 
International Symposium on Remote Sensing of 
Environment. 1245-1254. 
Cummings, R: W. Jr. 1977. Minimal Information 
System for Agricultural Development in Low- 
income Countries. New York: Agricultural 
Development Council, Inc. 
Gilmer, D. S. et al. 1980. Enumeration of prairie 
wetlands with Landsat and aircraft data. 
Photogrammetric Engineering and Remote Sensing. 
46: 631-634. 
Haack, B. N. 1983. A comparison of visual and 
numerical analysis of Landsat data for grassland 
and forest inventories in Swaziland. ITC 
Journal. 6-12. 
Horwitz, H. M. et al. 1971.Estimating the portions 
of objects within a single resolution element of 
a multispectral scanner. Proceedings: Seventh 
International Symposium on Remote Sensing of 
Environment. 1307-1320. 
National Academy of Sciences (NAS). 1977. Remote 
Sensing from Space: Prospects for Developing 
Countries. Washington, DC. 
Philipson, W. R. and W. R. Hafker, 1980. Manual 
versus digital Landsat analysis for modeling 
river flooding. Proceedings: American Society of 
Photogrammetry Fall Technical Meeting. RS/3/D/1- 
10.