Monitoring and predicting above ground biomass yield of grasslands are of key importance for grassland management. Established manual methods such as clipping or rising plate meter measurements provide accurate estimates of forage yield, but are time consuming and labor intensive, and do not provide spatially continuous data as required for precision agriculture applications. Therefore, the main objective of this study is to investigate the potential of sward height metrics derived from low-cost unmanned aerial vehicle-based image data to predict forage yield. The study was conducted over a period of 3 consecutive years (2014–2016) at the Rengen Grassland Experiment (RGE) in Germany. The RGE was established in 1941 and is since then under the same management regime of five treatments in a random block design and two harvest cuts per year. For UAV-based image acquisition, a DJI Phantom 2 with a mounted Canon Powershot S110 was used as a low-cost aerial imaging system. The data were investigated at different levels (e.g., harvest date-specific, year-specific, and plant community-specific). A pooled data model resulted in an R² of 0.65 with a RMSE of 956.57 kg ha⁻¹, although cut-specific or date-specific models yielded better results. In general, the UAV-based metrics outperformed the traditional rising plate meter measurements, but was affected by the timing of the harvest cut and plant community.

监测和预测草地的地上生物量产量对于草地管理至关重要。已建立的手动方法（例如修剪或升高平板计的测量）可以准确估算草料产量，但既费时又费力，并且不能提供精确农业应用所需的空间连续数据。因此，这项研究的主要目的是研究从低成本的无人机图像数据中得出的草料高度度量潜力，以预测草料产量。该研究连续3年（2014-2016年）在德国的伦根草原实验（RGE）进行。RGE成立于1941年，此后一直采用相同的管理制度，采用随机区组设计和每年两次减产五种处理方法。对于基于无人机的图像采集，已安装佳能Powershot S110的DJI Phantom 2被用作低成本的航空成像系统。在不同级别（例如特定于收获日期，特定年份和植物群落）对数据进行了调查。汇总数据模型导致R ²为0.65，RMSE为956.57 kg ha ^-1，尽管切割特定或日期特定的模型产生了更好的结果。总的来说，基于无人机的指标优于传统的上升平板仪表的指标，但是受收获时机和植物群落时间的影响。