Reaching new heights: can drones replace current methods to study plant population dynamics?

Abstract

Spatially explicit data on heterogeneously distributed plant populations are difficult to quantify using either traditional field-based methods or remote sensing techniques alone. Unmanned Aerial Vehicles (UAVs) offer new means and tools for baseline monitoring of such populations. We tested the use of vegetation classification of UAV-acquired photographs as a method to capture heterogeneously distributed plant populations, using Jacobaea vulgaris as a model species. Five sites, each containing 1–4 pastures with varying J. vulgaris abundance, were selected across Schleswig–Holstein, Germany. Surveys were conducted in July 2017 when J. vulgaris was at its flowering peak. We took aerial photographs at a 50 m altitude using three digital cameras (RGB, red-edge and near-infrared). Orthomosaics were created before a pixel-based supervised classification. Classification results were evaluated for accuracy; reliability was assessed with field data collected for ground verification. An ANOVA tested the relationship between field-based abundance estimations and the supervised classifications. Overall accuracy of the classification was very high (90.6%, ± 1.76 s.e.). Kappa coefficients indicated substantial agreement between field data and image classification (≥ 0.65). Field-based estimations were a good predictor of the supervised classifications (F = 7.91, df = 4, P = 0.007), resulting in similar rankings of J. vulgaris abundance. UAV-acquired images demonstrated the potential as an objective method for data collection and species monitoring. However, our method was more time consuming than field-based estimations due to challenges in image processing. Nonetheless, the increasing availability of low-cost consumer-grade UAVs is likely to increase the use of UAVs in plant ecological studies.