Many variables were found to be highly correlated with erosion but intercorrelation 

 among most of these was also quite high. For each area, the most promising variables 

 were chosen. In these analyses, the general objective was to reduce standard error to 

 a minimum using, as much as possible, those variables most easily measured or estimated. 



These analyses produced seven regression equations, one for each study area. Each 

 of the following variables appears in at least one equation: 



A - Proportion of the soil surface protected from direct raindrop impact by 

 vegetation and litter. 



B - Proportion of soil surface protected from direct raindrop impact by 

 vegetation, litter, and stone. 



D - Sand content of the surface inch of soil (proportion by weight). 



E - Organic matter content of the surface inch of soil (proportion by weight). 



F - Organic matter content of the surface 2 inches of soil 

 (proportion by weight). 



G - Slope gradient in percent. 



H - Bulk density of the surface 4 inches of soil (g./cc). 

 L - Air-dry weight of litter (pounds per milacre). 



Of course, these are not the only variables that affect erosion. Most of the other 

 measured variables had some effect on erosion but they did not explain sufficient addi- 

 tional variance to merit their inclusion in any multiple regression equations. For 

 example, erosion was closely correlated with soil aggregation characteristics in some 

 areas but the relations are rather complex and variable. Large water-stable aggregates 

 resist erosion but stable aggregates smaller than 0.5-mm. diameter seemed to be more 

 easily eroded than unaggregated material. Organic matter content served to explain as 

 much, or more, variance as aggregation. When organic matter content was included in 

 the equations, the additional variance explained by aggregation was small. Erosion was 

 also affected by soil moisture content but these relations were also complex and vari- 

 able. Some soils are more erodible when wet and some are more erodible when dry. Since 

 surface soil moisture content was so variable, changed so rapidly, and generally had 

 minor effects, it was not included in any of the final equations. 



RESULTS AND DISCUSSION 



Without exception, protection of the soil surface from direct raindrop impact 

 proved to be the most important means of controlling erosion on the study areas. On 

 four of the study areas, the logarithm of soil eroded was more closely correlated with 

 proportion of soil surface protected from raindrop impact by vegetation and litter than 

 with any other measured variable. On the other three, the highest correlation was ob- 

 tained with proportion of the soil surface protected by plants, litter, a?id stone. On 

 these study areas, the presence of stone on soil surface not otherwise protected con- 

 tributed significantly to protection against erosion. It is probable that stone had a 

 protective influence in all cases but its effects were negligible on four of the seven 

 areas, possibly because stone was not very prevalent on those areas or did not provide 

 protection that was as effective as plants and litter. 



The effects of plot slope gradient are important on all seven study areas and 

 appear in all but one regression equation. This exception is the equation for an area 

 where there were few observations and little variation in slope. The direct relation 

 between erosion and slope tends to be greatest on the sandy soils. 



5 



