Non-destructive biomass estimation of vegetation has been performed via remote sensing as well as physical measurements. An effective method for estimating biomass must have accuracy comparable to the accepted standard of destructive removal. Estimation or measurement of height is commonly employed to create a relationship between height and mass. This study examined several types of ground-based mobile sensing strategies for forage biomass estimation. Forage production experiments consisting of alfalfa (Medicago sativa L.), bermudagrass [Cynodon dactylon (L.) Pers.], and wheat (Triticum aestivum L.) were employed to examine sensor biomass estimation (laser, ultrasonic, and spectral) as compared to physical measurements (plate meter and meter stick) and the traditional harvest method (clipping). Predictive models were constructed via partial least squares regression and modeled estimates were compared to the physically measured biomass. Least significant difference separated mean estimates were examined to evaluate differences in the physical measurements and sensor estimates for canopy height and biomass. Differences between methods were minimal (average percent error of 11.2% for difference between predicted values versus machine and quadrat harvested biomass values (1.64 and 4.91 t·ha⁻¹, respectively), except at the lowest measured biomass (average percent error of 89% for harvester and quad harvested biomass < 0.79 t·ha⁻¹) and greatest measured biomass (average percent error of 18% for harvester and quad harvested biomass >6.4 t·ha⁻¹). These data suggest that using mobile sensor-based biomass estimation models could be an effective alternative to the traditional clipping method for rapid, accurate in-field biomass estimation.

Forage nutritive value analysis is typically performed via laboratory methods. Developing a system that provides crude protein (CP) estimates more quickly would be valuable. For evaluating feasibility of this system, N rate experiments containing bermudagrass [Cynodon dactylon (L.) Pers.], tall fescue [Lolium arundinaceum (Schreb.) Darbysh.], and wheat (Triticum aestivum L.) were examined. A hyperspectral passive spectrometer collecting spectral data at 340 to 1030 nm and a sensor system developed for use from a mobile platform were employed to collect data for prediction of CP. The predicted CP from hyperspectral data regressed with those measured by near-infrared spectroscopy (NIRS) in a laboratory produced R² = 0.80. Bermudagrass CP predictions from data collected using the mobile system showed no seasonal influence and were characterized with R² = 0.85. Wheat predictions from the mobile sensors exhibited R² = 0.27 for both fall and spring wheat when modeled as one data set. When split into two data sets for fall (Feekes 1–7) and spring (Feekes 7–10) growth, wheat model predictions of CP bore R² = 0.65 and 0.01, respectively. Tall fescue exhibited a similar pattern for mobile data, whereas all predictions regressed with measured values exhibited an R² = 0.63, spring and early summer an R² = 0.83, and fall observations an R² = 0.41. The results of this study indicate prediction of CP using hyperspectral data are accurate enough to be used for reporting forage CP for bermudagrass and are seasonally dependent for reporting forage CP in tall fescue and wheat. Copyright © 2016. . Copyright © by the Crop Science Society of America, Inc.