To document trends in land use and herbaceous production, 71 field sites sampled among cropped fields, fallow fields and rangelands in the Fakara region (Niger) were monitored from 1994 to 2006. The overall trend in land use confirmed the historical increase of the cropped areas since mid 20th century, at an annual rate of 2% from 1994 to 2006. This trend is the result of changes in the relative extent of fields permanently cropped and fields under shifting cultivation, and for the latter, the relative proportion of short (3 years) and long (10 years) duration fallows. Type of land use together with topography and soil type determine the herbaceous production and the resulting yield measured towards the end of the wet season. The variation in site yields between years is of the same order of magnitude as the variation in yields between sites within a year. There is an overall decreasing trend in site yields by 5% annually from 1994 to 2006 that is not explained by variations in rainfall. The decreasing trend is observed on fields under shifting cultivation, fallowed fields and rangelands, although not all sites are equally affected. Causes are likely to be multiple which might include changes in land use, decline of soil fertility and increased grazing pressure. Indeed, the remaining rangelands on marginal land and the fallows still accessible to livestock are subject to such a heavy grazing during the rainy season that the herbaceous standing mass measured at the end of the season reflects poorly the actual production. After the two first years of cropping, the herbaceous yield in fields under shifting cultivation with no fertilisation is negatively affected by the number of successive years of cropping. Moreover, clearing fallow after a decreasing number of years affects the mean herbaceous yield of fallowed fields by reducing the contribution of more productive old fallows. Changes in land use, grazing pressure and soil fertility also triggered changes in species composition with a strong reduction in diversity from rangelands to fallows, and again from fallows to cropland weeds. No correlations was found however between productivity and species composition. Cumulative rainfall does not explain between site or between year deviations in herbaceous yield even when sites are sorted by land use type or by soil type in the case of fallow and rangelands. Simulated production calculated with the STEP model does not explain herbaceous yields much better even when sites are grouped by land use and soil type. However, relative changes of herbaceous yields are reasonably predicted on sites that remained fallowed and were not heavily grazed for at least four consecutive years.