DATA COLLECTION 



F'eld data were collected July 1-17, 1957, midway in the second growing season. 

 I'" . ^^mg fo^jj. of the treatment replications--64 plots--were sampled (two replications 



ch aspect road). One year later, July 1-17, 1958, all of the 128 plots were 

 ^ *led although 2 weeks before this sampling began, 21 plots on the south aspect road 

 i'blJ'been grazed by a band of sheep in trespass. 



The density of grass seedlings (live plants per square foot) and the percent of 

 •rnund cover were used to judge effectiveness of treatment. Sampling in each plot was 

 5ne on four permanent 72-inch transects staked randomly crosswise to the axis of the 

 Toadbed--two in the inboard half and two in the outboard half. A previous survey found 

 that the roadbeds averaged 16.5 feet in width (5). The notched side of a 4- by 72Tinch 

 sanpling frame was anchored to the transect. Live plants within the frame were 

 identified by species and counted, after which the notched side of the frame was used 

 as a point sampler by insertion of a needlepoint gage at each notch successively. 

 Percent ground cover was estimated from the number of basal hits on live vegetation and 

 Iplant litter. 



When plots were selected for treatment, none was eliminated to reduce experimental 

 error caused by variations in soil, topography, aspect, and roadbed compaction. It was 

 Ji?' deemed impractical to do this with roads built on irregular terrain. Accordingly, site 

 '^^roperties not common to every plot and hence not measured contributed to experimental 

 .fjXTOr. However, site properties common to every plot were indexed empirically if they 

 ivtre thought to be related to seedling emergence and establishment, independent of 

 ftreatment effects. For example, the percentage of each roadbed plot shaded at noon by 

 ■^surrounding trees and topography was estimated, and later used in an analysis of 



covariance. 



ANALYSES 



ig;'^? The data on individual species, on treatment effects, and on certain site properties 

 '.;Were examined statistically by analyses of variance and covariance (8) . 



s.^.' Some sources of variation that either could not be measured or were not deserve 

 discussion: 



'1. Some seed broadcast in the fall were lost through extraneous causes before 

 endnation in late winter. Had losses been uniform from plot to plot, the experimental 

 ^sequences would not have been as serious. However, during the seeding operation 

 ill mammals were observed gathering and storing seed taken from some road surfaces 

 _ . not from others. Obviously, such feeding would leave a nonuniform distribution of 

 j^^^in the roadbed plots. Therefore, the sparseness of plants in some transects may 

 well be related to an initial lack of seed and not to differences in treatment and 



te. 



Furthermore, in the time between seeding and germination, two major periods of 

 ^land flow occurred when seed may have been washed away completely or redistributed 

 the plot. The first such period was during the heavy rainstorms of December 

 1955. The second was the normal snowmelt runoff in late March 1956. 



~2, Such soil properties as compaction, texture, and fertility, had they been 

 ined directly rather than indexed empirically, might have explained more site 



between plots. An intensive study of soil properties should be given 

 -****'^^ion in any new reseeding trials. 



