(R 2 = 0.795, F = 19.4, P = 0.008). Temperature variables did 

 not significantly (P > 0.25) add to the predictive power of 

 regressions on shrub yield. However, a combined index of forb 

 and shrub production was most significantly related to a 

 combination of precipitation from July through April and mean 

 temperature for May (R 2 = 0.87 8, F - 14.43, P = 0.017). 

 Stepwise procedures indicated that the precipitation factor 

 contributed most to the regression (R 2 = 0.643, F = 9.02, P = 

 0.03), but mean May temperature also contributed significantly 

 (R 2 = 0.235, F = 7.72, P = 0.05). 



Production measurements were limited to forbs and grasses 

 after 1982. Addition of those data to the regression model 

 did not significantly alter results. Precipitation during 

 April and May continued to be significantly positively 

 correlated with grass production (r = 0.863), explaining 74% 

 of the variation in yield between years. Production of forbs 

 also continued to be significantly correlated with mean 

 temperature for May and precipitation from July-April, 

 although the predictive capability of the resulting regression 

 was diminished (R 2 = 0.720, F = 10.28, P = 0.007) compared to 

 that derived from data for 1976-1982. 



An examination of the residuals of this regression 

 indicated that production during 1984 and 1986 deviated most. 

 A regression of mean temperature for May and precipitation 

 during July-April on forb production for the 9 years that 

 excluded 1984 and 1986 had very high predictive power (R 2 = 

 0.941, F = 47.62, P = 0.0006). Data from the residuals of the 

 full regression indicated it overestimated forb production 

 when precipitation increased in years following drought. This 

 apparently occurred because the regression could not account 

 for moisture necessary to recharge soil moisture above the 

 wilting point, when additional moisture becomes effective for 

 plant growth. 



As noted earlier, soil moisture measurements probably 

 would be best correlated with forb yield because they 

 encompass temperature, precipitation, and past conditions. In 

 the absence of soil moisture measurements, the regression of 

 mean temperature for May and precipitation during July-April 

 gave adequate predictive power, providing it was recognized 

 that yield was overestimated in the year of recovery following 

 drought. Despite those overestimates, the regression 

 accurately predicted trend in production (stable, increase, or 

 decrease) in production from previous years. 



Based on these findings, an index of relative summer 

 forage production (Fig. 3.9) was computed for each year from 

 1959 through 1987 as forage yield estimates using the 

 regression equation: [ Y ( forb and shrub yield ) = 1,704 - 

 30.04 X 1 (mean temperature for May) + 24.61 X 2 ( precipitation 



62 



