Flower color index for detecting and monitoring warm-colored flowering across scales
Year: 2025
Type: article
Abstract: • The novel flower color index (FCI) can detect flowering events. • FCI is simple and fast to assess abundance, distribution, and dynamics of flowering. • Robust to different environments, scales, and geographic locations. • Applicable to any optical remote sensing data that has RGB-NIR. Flowers provide significant ecological and economic benefits. Despite rapid technological progress and broad availability of remote sensing data, flower detection and monitoring are still challenging because many phenological studies rely on vegetation indices (e.g., normalized difference vegetation index, NDVI) which are not well-suited for flowers. Although some studies have focused on detection of flowering, the development and application of a comprehensive and robust remote sensing method for flowering remains largely unexplored. In this study, we introduce the flower color index (FCI) for detecting four major warm flower colors—red, pink, purple, and yellow—by leveraging remote sensing bands including blue (B), green (G), red (R) and near-infrared (NIR); these bands are widely available across a variety of remote sensing platforms. Here, we demonstrate the application of FCI across species representing a range of vegetation forms (herbaceous, shrubs, and trees) and study locations, corresponding to global biodiversity hotspots. We also tested FCI performance utilizing imagery with different spatial resolution—SkySat (0.5 m), PlanetScope (3 m), and Landsat (30 m). We show that FCI effectively detects and quantifies flowering events, which can be widely applied in ecology (e.g., flowering phenology, species identification, biodiversity conservation), agriculture (e.g., pollination management, yield estimation), and ecotourism (e.g., cherry blossom festival). Our proposed FCI has a simple formulation, is fast to calculate, and robust to different environments (urban or natural), spatial scales (spatial resolution and spatial extent), and background effects. Optimal FCI threshold—though variable across environments and settings—consistently lies on the right side of the distribution, facilitating automatic flower detection.
Authors: Bina Thapa, Brady S. Hardiman, Songlin Fei
Institution: Purdue University West Lafayette
Language: English
Subfield: Ecology
Field: Environmental Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,550
OpenAlex ID: W4416772294