It is obvious that for short yarding distances the prediction models can result 

 in unreasonable times. Forcing the models through the origin in order to correct 

 this deficiency was tried but this resulted in significantly lower R-square values and 

 higher variances. Logs were not generally yarded with the balloon system unless they 

 were several hundred feet from the yarder. Therefore, not much data were available 

 for the shorter distances during the model building. As with any regression equation, 

 care should be exercised when using the equation for predictive purposes. Use of 

 short distances should be avoided, especially since they are not normal balloon yarding 

 range . 



Of the TLO models built, four were approximately the same in reference to the six 

 criteria (fig. 6) . Since they were approximately the same, model number one chosen. 

 It was fairly easy to use and did not depend on the limited temperature (TEMP) and wind 

 velocity (WINVEL) data. The resulting equation: 



TLO = 0.10 + 0.0123LDI - 0.386LSLO + 0.541LSL0 2 + 0.000073LDI 2 



has an R-squared of 58.44 percent. Only the first two variables, LDI and LSLO, norm- 

 ally would be included in the model for predictive purposes since the last two variables 

 add little to the cumulative R-squared. Plots of the data and the physical situation, 

 however, showed a definite quadratic relationship with LDI and LSLO. The contribution 

 of the quadratics to the predictive equation is significant only at long lateral yard- 

 ing distances (LDI) and steep lateral slopes (LSLO) . Extension of the data on these 

 variables could be used to verify and refine the model. The variables used in model 

 number one had a physical relationship to the TLO element and were the only variables 

 left after the regression screening process. 



The HC variable was found to be best explained by its mean: 



HC = 1.01, with a variance of 0.606 as shown below: 



Mean 1.01 

 Median .70 

 Variance .606 



Skewness 2.3632. 



None of the independent variables were significant predictors of the time required 

 to hook the chokers. The number of logs per turn would normally be expected to affect 

 the time for HC, but in balloon logging, the chokers are preset so that the hooking 

 operation only requires attaching the choker rings to the tagline. 



Four satisfactory TL models were found using the six criteria. Model number one 

 (fig. 7) was selected as the "best" model, due to its relative simplicity and lack of 

 dependence on WINVEL: 



TL = 1.28 + 0.00138DI + 0.0000868WT - 1.151LSL0 - 1.626LSL0 2 + 0.00508LDI 



with an R-squared of 27.86 percent. 



Only the first three variables need to be included for predictive purposes; the 

 last two variables (LSLO 2 and LDI) add little to the R-squared value. Travel loaded 

 is a function of distance and weight as would be expected. Involvement of lateral 

 slope and lateral distance arise from the first operation in the TL component, wherein 

 the logs are yarded from their position to the side of the skyline before they can be 

 yarded along the skyline. All the independent variables, except distance, contribute 

 relatively small values to the R-squared; however, these additional variables improve 

 the residual plots at the extremes. Hopefully, the R-squared value could be increased 

 by trying the model on a larger range of data from another balloon logging site. 



in 



