We obtained a complete census of vegetation in a 30- x 1,900-m plot in
the Sonoran Desert (Tumamoc Hill, Pima County, Ariz.). We then evalua
ted the accuracy of 5 sampling techniques commonly used to sample vege
tation in deserts: step-point, point-quarter, and 3 line-intercept met
hods, We compared presence and percent occurrence for each method with
the census, The line-intercept (method 3) most closely estimated the
census: it was closer for all species occurring on >1% of the site, re
vealed the least variability relative to sample size for dominant spec
ies, and accounted for more (20 of 23) plant species on the study site
than the other methods.