Invasion of broadleved weeds cause competition with local grasses which are the main feed for herbivores. Chemical control should use selective herbicides that only kill the broadleaved weeds without harming the grass. The research was conducted in the savanna Bekol - Baluran National Park, East Java, from January 2015 until March 2015 to find the best rate of herbicide to control broadleaved weeds in the savanna and to estimate the soil residue of herbicide used. The experimental design was a randomized block with six treatments and four replications. The treatments were herbicide applications consisted of (1) the rate of triclopyr at 670 g a.i ha-1 (TA), (2) rate of triclopyr at 1340 a.i ha-1 (TB), (3) the rate of fluroxypyr at 200 g a.i ha-1 (FA), (4) rate of fluroxypir at 400 g a.i ha-1 (FB), (5) Weeding Manual (PM) and (6) Control (K).The result showed that there was a change in the composition of the vegetation after herbicide applied. The dominance of broadleaed weeds was replaced by the dominance of Brachiaria reptans and Sclerechnae punctata. The application of triclopyr at 1340 g ai ha-1 was able to reduce weed infestation to 0.48 g 0.25 m-2 compared to 12.66 g 0.25 m-2 before the herbicide application. It facilitated the growth of grasses up to 6.87 g 0.25 m-2 which was among the best of treatments applied because it has shown maximum effectieness for controlling broadleaved weeds.