Leaf area index (LAI) serves as a pivotal parameter in crop monitoring, significantly impacting agricultural applications. Empirical models are one of the commonly used methods for estimating LAI, they are often dependent on vegetation indices (VIs), predominantly derived from low-to-moderate spatial resolution satellite sensors. A critical limitation of these VIs is their tendency to saturate at elevated LAI values. Additionally, the interplay between chlorophyll content (Cab), LAI, as well as average leaf inclination angle (ALA), particularly in reflectance spectra from red to red-edge regions, has been underexplored in past research. Based on Sentinel-2 satellite, with three red-edge bands and enhanced spatio-temporal resolution, this study introduced a new three red-edge vegetation index (VI), comprising NDVI and CI, which leverages the unique spectral characteristics of these bands and related differential response to LAI and Cab variations. We investigated VI’s efficacy through a threefold approach: firstly, by conducting sensitivity analyses using noise equivalent (NE)ΔLAI ((NE) ΔLAI=1.37 and 1.59 for NDVI and CI; (NE)ΔLAI = 1.75 – 3.25 for other VIs) and extended Fourier amplitude sensitivity test (EFAST) (the contribution of LAI: FOI > 30 % and TOI > 40 % for VI; FOI ∼ 17 % − 28 % and TOI ∼ 25 % − 34 % for other VIs, except for NDVI with TOI ∼ 50 %), we demonstrated VI’s heightened sensitivity to LAI and its improved capability for seasonal LAI estimation compared to other VIs. Secondly, by comprehensively considering the simplicity, interpretability, AIC and R values of various models, the linear regression model was selected for analyzing the relationships between LAI and various VIs. We established that VI, particularly NDVI, exhibits a robust correlation with LAI, thereby enhancing LAI estimation accuracy (R = 0.72, RMSE = 1.39). Lastly, we applied the LAI-VI regression models to generate crop LAI imageries throughout the growing season, subsequently validating by field measured LAI data. The results affirmed VI’s superior performance in seasonal LAI estimation of crops, notably during peak growth and sowing phases. We conclude that the newly formulated VI offers a universal, highly precise, and reliable model for LAI prediction across various crop types and phenological stages. However, the broader application of VI may be limited, due to the three red-edge bands are not commonly found in most satellite platforms. In addition, some other aspects are not considered in this study, such as phenological information, meteorological factors, sampling strategies, etc., which should be meticulously considered in the future studies.

中文翻译：

---

使用 Sentinel-2 数据预测作物季节性 LAI 的新三红边植被指数 (VI3RE)


叶面积指数（LAI）是作物监测的关键参数，对农业应用产生重大影响。经验模型是估算 LAI 的常用方法之一，它们通常依赖于植被指数 (VI)，主要源自中低空间分辨率卫星传感器。这些 VI 的一个关键限制是它们在 LAI 值升高时趋于饱和。此外，叶绿素含量 (Cab)、LAI 以及平均叶片倾角 (ALA) 之间的相互作用，特别是在从红到红边缘区域的反射光谱中，在过去的研究中尚未得到充分探索。本研究基于Sentinel-2卫星，具有三个红边波段和增强的时空分辨率，引入了新的三红边植被指数（VI），包括NDVI和CI，该指数利用了这些波段独特的光谱特征，对 LAI 和 Cab 变化的相关差异响应。我们通过三重方法研究了 VI 的功效：首先，使用噪声当量 (NE)ΔLAI 进行敏感性分析（对于 NDVI 和 CI，(NE) ΔLAI=1.37 和 1.59；对于其他 VI，(NE)ΔLAI = 1.75 – 3.25) 并扩展傅立叶振幅灵敏度测试 (EFAST)（LAI 的贡献：对于 VI，FOI > 30 % 和 TOI > 40 %；对于其他 VI，FOI ∼ 17 % − 28 % 和 TOI ∼ 25 % − 34 %，除了 NDVI 和 TOI ∼ 50%），我们证明了与其他 VI 相比，VI 对 LAI 的敏感性更高，并且其季节性 LAI 估算能力有所提高。其次，综合考虑各种模型的简单性、可解释性、AIC和R值，选择线性回归模型来分析LAI与各种VI之间的关系。 我们确定 VI，特别是 NDVI，与 LAI 表现出很强的相关性，从而提高了 LAI 估计的准确性（R = 0.72，RMSE = 1.39）。最后，我们应用 LAI-VI 回归模型生成整个生长季节的作物 LAI 图像，随后通过现场测量的 LAI 数据进行验证。结果证实了 VI 在作物季节性 LAI 估算中的卓越性能，特别是在生长高峰和播种阶段。我们的结论是，新制定的 VI 为各种作物类型和物候阶段的 LAI 预测提供了一个通用、高精度和可靠的模型。然而，由于大多数卫星平台中并不常见三个红边频段，VI 的更广泛应用可能会受到限制。此外，本研究没有考虑到其他一些方面，如物候信息、气象因素、采样策略等，应在今后的研究中仔细考虑。