Integer programming in forestry

After finishing my Ph.D. in Berkeley in 1971, I needed a job in the San Francisco Bay Area, while my wife was finishing her Ph.D. in statistics. It was a tough year jobwise and the only available position was a part-time job with the US Forest Service Station in Berkeley, which funded a position as a Research Engineer at the OR Center at UC Berkeley. My boss and friend up to today, Daniel Navon had just developed TIMBERAM (Navon, 1971), the first widely used LP model for forest planning. There was even a US Congress directive compelling the use of FORPLAN in the next 10 year planning cycle. One of our main projects was extending TIMBERAM to include decisions on road building. Road building is not a minor issue, as it accounts for about 40% of operational costs. The planning process in those days was done separately for road building and harvesting. Road engineers took as input the harvesting plans from foresters and built a road network to access areas as they needed to be harvested. Intuitively this seemed suboptimal, and was proved so later by Jones, Hyde III, and Meacham (1986). So, it seemed a good idea to try a model that integrated harvesting and road building, which was my first project. My relation with the US Forest Service lasted almost two decades, most of them part time as I went back to Chile in 1974 to teach at the University of Chile. It was a complex bureaucratic, but legal set up. I would leave time cards signed with the secretary for about 25% time, and she would hand them in every month, while I was in Chile. I would do the work there, and come to Berkeley about twice a year. Stuart Dreyfus was the academic formally responsible, and he was always very helpful. I had been his TA in Dynamic Programming and we met several times for dinner while I was a student, so it was a comfortable