Abstract
Grassland yield estimates from remote sensing often rely on Leaf Area Index (LAI) or LAI-derived variables. We hypothesize that LAI may saturate at high yield levels resulting in inaccurate estimates due to plant parts contributing more to yield than to LAI, such as the stems. In a multi-site field experiment studying the effects of organic fertilization on the vegetation of moderately species-rich mountain permanent meadows, we measured dry matter yield, Leaf Area Index (with the sensor AccuPAR LP-80) and the yield proportion of grasses, legumes and forbs at the time of the first cut over three growing seasons. We evaluated the effect of the yield proportion of grasses, which were expected to provide the most relevant contribution of non-leafy plant material, on the accuracy of predicting dry matter yield by means of a linear mixed models accounting for LAI and design factors (site, year and site x year). Including the yield proportion of grasses into the statistical model allowed to slightly improve the accuracy of the prediction from 0.615 to 0.635 R².