Spatial-Temporal Evaluation of Plant Phenotypic Traits Via Imagery Collected By Unmanned Aerial Systems (UAS).

Current challenges in crop production relates to environment sustainability and productivity on site-specific management. Detailed spatial and temporal data is critical to overcome these limitations. Final yield is the outcome of a complex interaction among genotype  (G) x environment (E) x management practices (M) – (G x E x M interaction). Conventional methods (via plant sampling procedure) for estimating biomass are labor-intensive (temporal-analysis), time-consuming, and related to small-scale (hard to be extrapolated to large field-scale). Recent incursion of small-unmanned aerial vehicle systems (sUAVS) technology seems promising for overcoming the scale issue. The utilization of low cost sensors accompanying sUAVS technology allowed imagery collection for producing high-resolution digital models. Implementation of these models allowed us to estimate biophysical parameters from both spatial and temporal scales. The main goals of the project are: 1) to create a crop surface model (CSM) and correlate this model with ground-truthing data (biophysical parameters) for plant height trait (spatial-temporal scales) and 2) to utilize imagery collected for biomass prediction at varying growth stages and for final yield prediction. The experiment was located at Ashland Bottoms Research Farm, Kansas State University (Manhattan, KS). Four experiments were carried out during the 2015 corn growing season. The studies were related to investigation of fertilizer N rates, hybrid variation, plant density, and stand uniformity (random-gaps). PhotoScan Agisoft and ArcGIS 10.2 were utilized for imagery processing workflow and CSM generation. Correlation between absolute plants height values estimated by CSM versus ground truth measurements were evaluated at varying growth stages; and absolute plant height prediction looks promising. Several biophysical plant traits (e.g. stand counts, plant height, biomass, stalk diameter, among others) were collected throughout the entire corn growing season to generate plant growth models for biomass prediction. Plant growth estimation at varying growth stages (temporal-scale) will be utilized for yield prediction; and final yield validation will be implemented from data collected from the field studies.