In the years 1994 to 1996 14 herbicides were tested independently and in combination which were applied on 10 ruderal competitively strong species. They were in different facies of phytocenosis Rumicetum obtusifolii (Smarda et al., 1963) on ruderal grassland with feeding value Kh less than or equal to 4. The site was formed by eutrophized soil which in a 0 to 200 mm depth (Tab. II) contained extremely high values of the total nitrogen (from 3,534 to 4,767 mg.kg(-1)) and potassium (from 533.75 to 641.87 mg.kg(-1)). Secondarily synantropic plant community was formed on the original association of Lolio-Cynosuretum cristati R. Tx. 1937 by excessive over-fertilizing by cattle excrements during grazing season. By intensive footing, particularly after rains, the sod was completely disturbed and ruderal species were established at skips. The results of the research confirmed that in specific conditions of environment the control of undesirable perennial species was substantiated. Some selective system herbicides were proved to be good. Of them, 71.42% were manifested by their excellent and very good effect after application on the studied competitively strong weed species which prevailed before in the stand. After applied minimum dose of herbicide, they receded to be close to the level of maximum marginal occurrence. Competitively stronger were those weed species which stayed in the stand (Tabs III, IV). It follows from changes in floristic composition that by comparison of the treated stand with the control floristic group of grasses (Tab. III) was marked significantly by increase of its share. Among them it was particularly Poa trivialis which by its above-ground shoots filled the skips very fast (14.34 to 60.63%). The herbicides used did not cause damage to those grasses which formed the basis of the grassland. This group participated the most in thickening of the sod cover after receding of controlled species and in most cases their percentage increased. Floristic group of leguminoses was comparatively stable and of herbicides it was protected only by Asulox 40. Floristic group of other herbs was most abundant. Of feeding-valuable species with Kh greater than or equal to 5 the greatest loss was in eight variants in the species Taraxacum officinale. Achillea millefolium and Plantago lanceolata receded in five variants, Alchemilla xantochlora, Carum carvi in two, and Leontodon autumnalis did not occur in four variants. Among frequently occurring weed species with feeding value Kh less than or equal to 4 (Tab. IV) only variant 3 was surface-treated with the species Rumex obtusifolius where Asulox 40 was applied (100% effect). Slightly poisonous Ranunculus repens receded in seven of 14 tested variants (Tab. I). The problem species seemed to be Anthriscus sylvestris on which Starane 250 EC and Banvel 480 had the best effects. The most problem controlled species was Aegopodium podograria which did not recede even after application of herbicides Garlon 4, Duplosan DP and Garlon 4 + Lontrel 300 (93.50% effect). Phytotoxicity caused maximum damage of 40.07% to valuable species in the grassland after application of selective system herbicides (Tab. V). In weed species after control the threshold of noxiousness was recorded. Among selective system herbicides the greatest reduction of diversity was recorded after application of herbicides Duplosan DP (12 species), U 46 M Fluid (11 species), Banvel 480 and Aniten I (10 species).
