Evaluation of photographic imagery parameters relating to agricultural inventories in the Republic

This report describes a study conducted in the Dominican Republic to determine the effect of the interpreters" skills, type of image, and image scale on interpretation accuracy for several crops. The test site, in the east central portion of the Cibao Valley, is covered by savannah and subtropical broadleaf forest and is generally flat. Aerial photography and related ground truth were obtained in March 1979 and a follow-up ground survey was conducted a month later. Test site aerial photography consisted of 70mm black-and-white, color infrared, and color transparencies at scales of 1: 10,000, 1: 20,000, and 1: 30,000. Ground truth consisted of written descriptions of crop types, growth stage, and boundaries. Four interpreters with varying experience were given 1.5 hours of training to become familiar with the appearance of crops in aerial imagery and to enhance their ability to classify fields. They then proceeded to interpret all 74 fields on every type and scale of imagery tested. Statistical analysis of the results revealed a mean interpreter accuracy of 77%. Differences among the four were negligible (2.9%), indicating that experience was not a significant determinant of accuracy. It is thus recommended that training be conducted using photography from the same scale, film type, and date as those to be used in actual photointerpretation. Although color and color infrared photography allow more accurate land use classification, black-and-white film can produce a less costly yet reliable inventory. It was also found that the optimal combination of successful classification is achieved using 1:30,000 photographs. The authors therefore recommend the use of black-and-white film at a scale of 1:30,000 as the most cost-efficient and versatile combination. The report contains 14 tables, and 10 figures. Appended are a classification key, information on films and filters, a tabulation of each interpreter"s findings, and data on interpretation accuracy.