The rapid increase in both the quantity and complexity of data that are being generated daily in the field of environmental science and engineering (ESE) demands accompanied advancement in data analytics. Advanced data analysis approaches, such as machine learning (ML), have become indispensable tools for revealing hidden patterns or deducing correlations for which conventional analytical methods face limitations or challenges. However, ML concepts and practices have not been widely utilized by researchers in ESE. This feature explores the potential of ML to revolutionize data analysis and modeling in the ESE field, and covers the essential knowledge needed for such applications. First, we use five examples to illustrate how ML addresses complex ESE problems. We then summarize four major types of applications of ML in ESE: making predictions; extracting feature importance; detecting anomalies; and discovering new materials or chemicals. Next, we introduce the essential knowledge required and current shortcomings in ML applications in ESE, with a focus on three important but often overlooked components when applying ML: correct model development, proper model interpretation, and sound applicability analysis. Finally, we discuss challenges and future opportunities in the application of ML tools in ESE to highlight the potential of ML in this field.

Background Hyperspectral reflectance data in the visible, near infrared and shortwave infrared range (VIS–NIR–SWIR, 400–2500 nm) are commonly used to nondestructively measure plant leaf properties. We investigated the usefulness of VIS–NIR–SWIR as a high-throughput tool to measure six leaf properties of maize plants including chlorophyll content (CHL), leaf water content (LWC), specific leaf area (SLA), nitrogen (N), phosphorus (P), and potassium (K). This assessment was performed using the lines of the maize diversity panel. Data were collected from plants grown in greenhouse condition, as well as in the field under two nitrogen application regimes. Leaf-level hyperspectral data were collected with a VIS–NIR–SWIR spectroradiometer at tasseling. Two multivariate modeling approaches, partial least squares regression (PLSR) and support vector regression (SVR), were employed to estimate the leaf properties from hyperspectral data. Several common vegetation indices (VIs: GNDVI, RENDVI, and NDWI), which were calculated from hyperspectral data, were also assessed to estimate these leaf properties. Results Some VIs were able to estimate CHL and N (R 2 > 0.68), but failed to estimate the other four leaf properties. Models developed with PLSR and SVR exhibited comparable performance to each other, and provided improved accuracy relative to VI models. CHL were estimated most successfully, with R 2 (coefficient of determination) > 0.94 and ratio of performance to deviation (RPD) > 4.0. N was also predicted satisfactorily (R 2 > 0.85 and RPD > 2.6). LWC, SLA and K were predicted moderately well, with R 2 ranging from 0.54 to 0.70 and RPD from 1.5 to 1.8. The lowest prediction accuracy was for P, with R 2 < 0.5 and RPD < 1.4. Conclusion This study showed that VIS–NIR–SWIR reflectance spectroscopy is a promising tool for low-cost, nondestructive, and high-throughput analysis of a number of leaf physiological and biochemical properties. Full-spectrum based modeling approaches (PLSR and SVR) led to more accurate prediction models compared to VI-based methods. We called for the construction of a leaf VIS–NIR–SWIR spectral library that would greatly benefit the plant phenotyping community for the research of plant leaf traits.

No abstract