Table 1. Species identification confusion matrices for three photo 
interpreters (A, B and C) 
Interpreted species codes 
Actual PM LP WG SD CU Other 
Species ee E 
2e; No BODEN CHA CAD CAS 
  
  
  
*PM 33 33 26 29 6 1 1.3 
LP 6 6 6 5 1 
WG 14 14 14 14 
SD 6 6 6 6 
CU 2 2.2.2 
Total 61 33 26.29.06 12 5 14.14 14. 6.6 6 2.2.2 ] 4 
  
*The following correspondence of codes and scientific names is based on bits of infor- 
mation of different sources. It seems to be correct except perhaps for species CU 
which might be wrongly named: 
PM — Pentaclethra macroloba 
LP — Laetia procera 
CU - Dipteryx panamensis (?) 
WG — Welfia georgii 
SD — Socratea durissima 
Figure 9. Only species represented by more than one tree are included in the table. 
Interpreter A made no identification errors in classifying the 61 trees; interpreter B 
made 7 errors mainly calling the species coded PM as LP and; interpreter C made 5 
errors mainly being unable to classify some of the trees as one of the five species. A 
weakness in this experiment was that some unclassified trees were not included to make 
the test more realistic. The results, nevertheless, are quite encouraging if one considers 
the limited knowledge and experience that the interpreters had of tropical trees. Based 
on comments by the local scientist in charge of the field plots who used more of the 
black and white enlargements, a sound knowledge of tropical forests could improve the 
accuracy by a factor of 2 or more. 
The results concerning the best photography appear to contradict earlier work 
of Sayn-Wittgenstein (1978) who stressed the importance of colour. The discrepancy may 
be explained by the differences in photo scale. At 1:500 morphological characteristic of 
the tree crowns, branching habit and leaf texture or pattern seem to offer better clues 
than colour. At smaller scales colour may become the dominant factor in species 
recognition. 
Conclusions 
i. The trials demonstrated that the radar altimeter, initially developed to provide a 
means to compute photographic scale, is capable of producing an accurate profile 
of the ground, even if the terrain is covered by a dense tropical rain forest. This