This integrated field and model study explains low effectiveness of non-fertilized 5 m wide harvested grass buffer strips (BS) to reduce nutrient loads from agricultural lowland fields to ditches. Buffer strip effectiveness (BSE) was defined as the relative nutrient load reduction, compared to a normally fertilized reference strip (REF). We collected nutrient loads from paired BS and REF strips in separate reservoirs in the ditch for three or four years at five field sites characteristic for the hydrogeology of the Netherlands. No statistically significant BSE was found at three out of five sites. A statistically significant BSE of 10-15% was found for N at the peat soil site and of 57-61% for P at the site with pure shallow flow. Dynamic 2D modelling shows BSE for N slightly increases after four years, the increase of BSE for P depends on the rate of net P withdrawal from the BS by the harvested grass and on the buffer capacity of the soil. BSE increases with BS width or ditch density. A steady-state model predicts BSE for N between 7 and 22% for sandy soils, and between 14 and 25% for peat soils. Highest BSE for N is expected on fields with high, but not pure shallow flow. High BSE for P is expected with high surface runoff or shallow flow, and with high original P status of the soil (i.e. P leaking soils). Buffer strips are ineffective on pipe drained fields (clay soils). According to our cost analysis BS may be cost-effective under specific circumstances, but more cost effective alternatives exist to reduce nutrient loads. ; De effectiviteit is bestudeerd van 5 meter brede onbemeste bufferstroken (BS) met gemaaid gras om de belasting van het naastgelegen oppervlaktewater met nutriënten vanuit landbouwpercelen te verminderen.