Terrestrial laser scanning for crop monitoring. Capturing 3D data of plant height for estimating biomass at field scale

Terrestrial laser scanning (TLS) is a young remote sensing method, but the trustworthiness of such measurements offers great potential for accurate surveying. TLS allows non-experts to rapidly acquire 3D data of high density. Generally, this acquisition of accurate geoinformation is increasingly desired in various fields, however this study focuses on the application of TLS for crop monitoring. The increasing cost and efficiency pressure on agriculture induced the emergence of site specific crop management, which requires a comprehensive knowledge about the plant development. An important parameter to evaluate this development or rather the actual plant status is the amount of plant biomass, which is however directly only determinable with destructive sampling. With the aim of avoiding destructive measurements, interest is increasingly directed towards non-contact remote sensing surveys. Nowadays, different approaches address biomass estimations based on other parameters, such as vegetation indices (VIs) from spectral data or plant height. Since the plants are not taken it is feasible to perform several measurements across a field and across the growing season. Hence, the change of spatial and temporal patterns can be monitored. This study applies TLS for objectively measuring and monitoring plant height as estimator for biomass at field scale. Overall 35 TLS campaigns were carried out at three sites over four growing seasons. In each campaign a 3D point cloud, covering the surface of the field, was obtained and interpolated to a crop surface model (CSM). A CSM represents the crop canopy in a very high spatial resolution on a specific date. By subtracting a digital terrain model (DTM) of the bare ground from each CSM, plant heights were calculated pixel-wise. Manual measurements aligned well with the TLS data and demonstrated the main benefit of CSMs: the highly detailed acquisition of the entire crop surface. The plant height data were used to estimate biomass with empirically developed biomass regression models (BRMs). Validation analyses against destructive measurements were carried out to confirm the results. The spatial and temporal transferability of crop-specific BRMs was shown. In one case study, the estimations from plant height and six VIs were compared and the benefit of fusing both parameters was investigated. The analyses were based on the TLS-derived CSMs and spectral data measured with a field spectrometer. The important role of plant height as a robust estimator was shown in contrast to a varying performance of BRMs based on the VIs. A major benefit through the fusion of both parameters in multivariate BRMs could not be concluded in this study. Nevertheless, further research should address this fusion, with regard to the capability of VIs to assess information about the vegetation cover or biochemical and biophysical parameters