植被荧光如何预测小麦产量？

精准预测作物产量对农业管理和粮食安全至关重要。传统方法依赖人工采样和统计估算，不仅耗时耗力，而且精度有限。近年来，随着遥感技术的发展，太阳诱导叶绿素荧光（SIF）成为一种具有潜力的新型作物监测指标。

那么，SIF是如何反映作物生长状况的？它在产量预测中有哪些优势？又受到哪些因素影响？南京农业大学农学院智慧农业团队对这些问题展开了系统分析。

「实验设计与方法」

研究人员在中国进行了连续两年的小麦田间试验，设置了不同氮肥施用水平、种植密度和品种的小麦小区。在不同时间尺度（包括关键生育期、日变化和全生长季时间序列）采集冠层光谱数据，并同步测定叶面积指数（LAI）、叶绿素含量（Cab）等参数，最终结合收获实测产量进行分析。

图：小麦农学参数（LAI和Cab）、日均VIs、日均SIF参数和PPFD的季节变化

「主要发现」

1. 时间尺度很关键

研究发现，在较大的时间尺度上，累积的SIF数据通常在小麦产量估计方面表现更好，多个生长时期的累积SIF值显示出更强的相关性。

2. 非线性模型更优

非线性模型通常能更准确地描述SIF与小麦产量的关系。尤其在瞬时测量尺度下，非线性模型比线性模型拟合效果更好。但随着时间尺度增大，两者差异逐渐缩小。

3. optimal观测时期是开花期

在开花期测量的总近红外SIF（SIFNIR_tot）与产量相关性zui高，是该时期optimal产量预测指标。

图：关键生育期下SIF参数和植被指数与产量的相关性

4. 冠层结构影响显著

研究应用主成分分析法和偏最小二乘法的变量投影重要性分析，发现叶面积指数（LAI）和叶绿素含量（Cab）对SIF–产量关系有重要影响。其中，LAI的影响更大。两者都存在一个“optimal范围"。

表：偏最小二乘回归（PLSR）模型中影响因变量（yield/SIFypNIR或yield/SIFypNIR_tot的PC1）的因素（自变量的PC1：LAI、Cab和PAR）的变量投影重要性（VIP）得分

5. NDFI敏感但不强

归一化差异荧光指数（NDFI）对LAI和Cab的变化十分敏感，但在直接预测产量方面表现不如SIFNIR_tot。

「对农业的启示」

这项研究不仅明确了SIF在产量预测中的实用性，也指出了其受时间尺度、冠层结构和环境条件的综合影响。未来，通过多源数据融合与模型优化，SIF有望成为区域乃至global尺度作物产量监测的核心手段。

拓展阅读：如何获取高质量SIF数据？

想要开展SIF相关研究或应用，高精度、可定制的地面或无人机载监测设备是关键。爱博能研发生产的日光诱导叶绿素荧光监测系统（ABN-SIF-2），配备双波段光谱仪，可同步获取SIF信号与多种植被指数，支持在线监测与无人机载监测。相比卫星遥感，该系统具备更高的空间分辨率，适合田间尺度精准监测，为农情研判与作物模型验证提供可靠数据支持。

案例来源：The Relationship between Wheat Yield and Sun-Induced Chlorophyll Fluorescence from Continuous Measurements over the Growing Season.

How Does Vegetation Fluorescence Predict Wheat Yield?

Accurate prediction of crop yield is crucial for agricultural management and food security. Traditional methods rely on manual sampling and statistical estimation, which are not only time-consuming and labor-intensive but also have limited accuracy. In recent years, with the development of remote sensing technology, Solar-Induced Chlorophyll Fluorescence (SIF) has emerged as a promising new indicator for crop monitoring.

So, how does SIF reflect crop growth status? What are its advantages in yield prediction? And what factors influence it? The team from the College of Agriculture at Nanjing Agricultural University conducted a systematic analysis of these questions.

「Experimental Design and Methods」

Researchers conducted a two-year field experiment on wheat in China, establishing plots with different nitrogen application levels, planting densities, and wheat varieties. Canopy spectral data were collected at different temporal scales (including key growth stages, diurnal variations, and full-growth-season time series). Parameters such as the Leaf Area Index (LAI) and Chlorophyll Content (Cab) were measured synchronously, and the data were ultimately analyzed in conjunction with the actual yield measured at harvest.

Figure: Seasonal variations in wheat agronomic parameters (LAI and Cab), daily average Vegetation Indices (VIs), daily average SIF parameters, and Photosynthetic Photon Flux Density (PPFD).

「Key Findings」

1) Temporal Scale is Crucial:

The study found that on larger temporal scales, cumulative SIF data generally performed better for wheat yield estimation. Cumulative SIF values across multiple growth stages showed a stronger correlation with yield.

2) Nonlinear Models are Superior:

Nonlinear models generally described the relationship between SIF and wheat yield more accurately. This was especially true at the instantaneous measurement scale, where nonlinear models provided a better fit than linear models. However, the difference between the two model types diminished as the temporal scale increased.

3) The Optimal Observation Period is the Flowering Stage:

The total near-infrared SIF (SIFNIR_tot) measured at the flowering stage had the highest correlation with yield, making it the best yield predictor for that period.

Figure: Correlation between SIF parameters/Vegetation Indices and yield during key growth stages.

4) Canopy Structure Has a Significant Impact:

Using Principal Component Analysis and Variable Importance in Projection (VIP) scores from Partial Least Squares Regression (PLSR) analysis, the study found that Leaf Area Index (LAI) and Chlorophyll Content (Cab) significantly influenced the SIF-yield relationship. Among these, LAI had a greater impact. An "optimal range" was observed for both parameters.

Table: Variable Importance in Projection (VIP) scores from the Partial Least Squares Regression (PLSR) model, showing the influence of factors (PC1 of independent variables: LAI, Cab, and PAR) on the dependent variable (PC1 of yield/SIFypNIR or yield/SIFypNIR_tot).

5) NDFI is Sensitive but Not Strong for Direct Prediction:

The Normalized Difference Fluorescence Index (NDFI) was highly sensitive to changes in LAI and Cab. However, it was less effective than SIFNIR_tot for directly predicting yield.

「Implications for Agriculture」

This research not only confirms the practicality of SIF for yield prediction but also highlights that its effectiveness is influenced by a combination of temporal scale, canopy structure, and environmental conditions. In the future, through multi-source data fusion and model optimization, SIF is expected to become a core tool for crop yield monitoring at regional and even global scales.

「Further Reading: How to Obtain High-Quality SIF Data?」

Conducting SIF-related research or applications requires high-precision, customizable ground-based or UAV-borne monitoring equipment. The ABN-SIF-2 Solar-Induced Chlorophyll Fluorescence Monitoring System, developed by ExponentSci, features a dual-band spectrometer capable of simultaneously acquiring SIF signals and various vegetation indices. It supports online monitoring and UAV-based monitoring. Compared to satellite remote sensing, this system offers higher spatial resolution, making it suitable for precise monitoring at the field scale and providing reliable data support for agricultural condition assessment and crop model validation.

Sources:

The Relationship between Wheat Yield and Sun-Induced Chlorophyll Fluorescence from Continuous Measurements over the Growing Season.