trees selected on each photo subplot for testing 

 individual tree species identification were located 

 on the ground and the species recorded. Similar 

 methods were applied to the 1 :2400 scale photog- 

 raphy. However, ground examination was limited 

 to a subset of those areas used to ground check 

 the 1:1 600 scale photography. 



As a further test of accuracy of identifying 1 -year 

 mortality trees, in summer 1 972, the flight lines on 

 the two subcompartments were reflown by Robert 

 Heller, then with the Pacific Southwest Forest and 

 Range Experiment Station. The photographs were 

 70 mm, 1:3200 scale true color transparencies. 

 This scale was selected to provide some assur- 

 ance that a majority of the 1971 photographs 

 would be included in the 1972 coverage. 



RESULTS OF PHOTO 

 INTERPRETATION 



The first study objective deals with the accuracy 

 of identifying 1 - and 2-year mortality trees. Tables 

 1-4 summarize the accuracy with which date of 

 mortality was assigned to mortality trees. 



The results indicate that 1-year mortality trees 

 can be accurately identified on 1 :1 600 scale color 

 aerial photography. There were errors in identify- 

 ing and dating white pine mortality on one of the 

 flight lines on the 1 :1 600 scale photography. This 

 was due to inadequate training of interpreters. 

 Careful examination has led me to conclude that, 

 with proper training, this problem can be avoided. 



If white pine are omitted from the analysis, the 

 number of mortality trees correctly dated, ex- 

 pressed as a percent of ground truth, increases 

 from 74 percent to 85 percent (tables 3 and 4). 

 Thus, I anticipate that 1-year mortality can be 

 correctly dated at least 85 percent of the time on 

 1:1600 scale photography. Interpretation accu- 

 racy is reduced to 70 percent when photography 

 scale is reduced to 1:2400. 



The number of 2-year mortality trees correctly 

 dated, expressed as a percent of ground truth, is 

 only 50 percent on 1:1600 scale photography. 

 When the photo scale is reduced to 1 :2400, inter- 

 pretation accuracy decreases to 22 percent. 



This reduction in interpretation accuracy is pri- 

 marily due to an increase in the number of missed 

 trees on the 1:2400 scale photography. If white 

 pine are deleted from the analysis virtually no 

 difference exists in the interpreter's ability to 

 identify 1-year mortality on the two photo scales, 

 given the dead tree is detected on the photograph 

 (tables 3 and 4). On 1 :1 600 scale photography, 88 

 percent of the trees rated as 1 -year mortality trees 

 were correctly rated. On 1:2400 scale photog- 

 raphy, 90 percent of the trees rated as 1-year 

 mortality trees were correctly rated. 



Tables 1 and 2 may also be used to compare the 

 distribution of mortality trees identified on the 

 aerial photographs with the actual distribution on 

 the ground. On 1:1600 scale photography, the 

 number of 1-year mortality trees is within 2 per- 

 cent of the number found on the ground. On 

 1 :2400 scale photography, the number of 1 -year 

 mortality trees is within 7 percent of the number 

 found on the ground. Thus, the number of errors of 

 omission (1-year mortality trees either missed or 

 misclassified on the photos) must be almost equal 

 to the number of errors of commission (trees 

 identified on the photos as 1-year mortality that 

 actually had either been dead more than 1 year or 

 were not dead). 



Interpretation accuracy of 2-year mortality is 

 much poorer. On 1:1600 scale photography the 

 number of 2-year mortality trees identified is only 

 within 25 percent of the number found on the 

 ground. The number of 2-year mortality trees iden- 

 tified on 1 :2400 scale photography is only within 

 50 percent of the number found on the ground. 

 The reduction in interpretation accuracy is due 

 primarily tothe many 2-year mortality trees missed 

 on the photographs. Most of the missed 2-year 

 mortality trees were identified as being dead for 

 more than 2 years and thus were not recorded. 



3 



