日光诱导叶绿素荧光（SIF）对于植物光合活性的动态变化极其敏感，它能够提供光合效率以及生态系统健康状态的相关信息，在监测植被光合生理方面具有独特优势。水汽吸收波段是640-800 nm叶绿素荧光窗口内最宽的吸收波段，其吸收线深度相较于O₂-B波段也更深；同时该波段处于红光和近红外波段两个荧光峰值之间，对于捕捉SIF光谱的整体形状也至关重要。从理论上分析该波段应具有较高的SIF探测潜力。然而，该波段的典型特征是大气水汽柱浓度变化显著，水汽含量的时间和空间动态变化可能会造成SIF反演较大的不确定性，也因此导致了该波段在实际应用中存在不小的局限性。基于此，本文通过模拟实验和地面塔基平台观测实验，利用目前常用的反演算法（sFLD、3FLD、iFLD、pFLD、SFM、SVD和DOAS），在多种观测环境下进行了水汽吸收波段的日光诱导叶绿素荧光（SIF）反演，并对不同算法的反演结果进行了对比，以评估水汽吸收波段的SIF反演能力。本文的主要研究结果总结如下：

（1）水汽吸收波段具备较高的SIF反演潜力。首先该波段受大气辐射传输效应影响较小，基于近地面（1 m高度）和30 m塔基观测高度下获取的SIF数据一致性较强，反演精度无显著差异。其次，多种算法在该波段可以提供可靠的SIF信息，基于常用的两个氧气吸收窗口的精度对比分析，水汽吸收波段的SIF反演精度整体上显著优于O₂-B波段，但次于O₂-A波段。

水汽波段0.3nm光谱分辨率条件下各算法的反演误差（RRMSE）基本维持在10%以内，随着光谱分辨率降低到3.0 nm，几种FLD算法和SVD算法会出现误差突增行为，但多种算法仍能提供较可靠的SIF反演信息，尤其是SFM算法其RRMSE在3.0 nm时仍保持在较低水平。其中sFLD算法在水汽波段的荧光反演中误差较高，存在严重的SIF高估行为，在考虑噪声影响时其RRMSE超过100%，不具备适用性。结合常用野外光谱观测设备的指标配置，QE65 Pro光谱仪获得的SIF数据精度较高，ASD FieldSpec 3光谱仪指标下SFM和DOAS表现较好，其RRMSE分别为9.08%和9.56%。大气水汽柱浓度对该波段的SIF反演影响显著，水汽含量与SIF反演精度呈正相关关系，其中sFLD、iFLD和DOAS三种算法对水汽含量变化较敏感。

（2）基于塔基SIF光谱观测系统（配置QE65 Pro光谱仪）的SIF反演中，除了sFLD种算法外的其他算法在该波段能够提供可靠的SIF反演信息，反演误差RMSE维持在0.5 mW/m²/nm/sr以内，其中同样以SFM的表现最优，而sFLD表现严重的SIF高估行为。研究验证了大气校正对于该波段的荧光反演影响较小，校正前后各算法反演的SIF数据与PAR之间的RMSE和R²精度变化不显著；而光照条件对于SIF反演的影响较大，晴天状态下各算法的反演精度显著优于阴天状态。各算法反演SIF与近红外植被反射辐亮度（NIRvR）的一致性评估方面，sFLD和SVD反演得到的SIF数据与NIRvR之间存在较高相关性，R²分别达到0.77和0.72，iFLD表现较差R²仅有0.3左右，其算法稳定性较差，而其余算法没有明显的统计检验差别，R²基本维持在0.6附近。基于DOY时间轴分析SIF日均值变化，研究发现水汽吸收波段反演的SIF数据在时空动态变化方面与O₂-A波段的荧光数据具有较强的一致性。

Solar-induced chlorophyll fluorescence (SIF) is highly sensitive to the dynamic changes in plant photosynthetic activity, providing crucial information on photosynthetic efficiency and ecosystem health. It plays a pivotal role in monitoring vegetation physiological and biochemical processes. The water vapor absorption band is the broadest absorption window within the 640–800 nm chlorophyll fluorescence emission range. It also exhibits deeper absorption lines compared to the O₂-B band. Moreover, this band lies between the two fluorescence peaks in the red and near-infrared spectral regions, making it essential for capturing the overall shape of the SIF spectrum. Theoretically, this band holds significant potential for SIF retrieval. However, its characteristic feature is the substantial variation in atmospheric water vapor column concentrations, leading to considerable temporal and spatial fluctuations in water vapor content. These variations introduce significant uncertainties in SIF retrieval and impose practical limitations on the band’s real-world applicability. To address these challenges, this study conducted simulated experiments and ground-based tower observations, applying seven commonly used retrieval algorithms (sFLD, 3FLD, iFLD, pFLD, SFM, SVD, and DOAS) to retrieve solar-induced chlorophyll fluorescence (SIF) in the water vapor absorption band under various observational conditions. The retrieval results of different algorithms were compared to assess the feasibility of SIF retrieval using this band. The primary findings of this study are summarized as follows:

(1) The results demonstrate that the water vapor absorption band possesses considerable potential for SIF retrieval. The impact of atmospheric radiative transfer effects on this band is relatively small, as evidenced by the strong consistency of SIF data obtained at tower observation heights of 1 m and 30 m, with no significant differences in retrieval accuracy. Furthermore, multiple algorithms provide reliable SIF retrievals within this band. When compared with the commonly used atmospheric absorption windows, the retrieval accuracy of SIF in the water vapor absorption band is significantly higher than that in the O₂-B band but slightly lower than that in the O₂-A band.

Under high spectral resolution conditions, the relative root mean square error (RRMSE) of retrieval across different algorithms remains below 10%. However, as the spectral resolution decreases to 3.0 nm, several FLD-based algorithms and the SVD algorithm exhibit a sharp increase in retrieval error. Despite this, multiple algorithms still yield relatively reliable SIF retrievals, with the SFM algorithm maintaining a low RRMSE even at 3.0 nm. Notably, the sFLD algorithm exhibits substantial overestimation errors in SIF retrieval within the water vapor absorption band, with its RRMSE exceeding 100% under noise conditions, rendering it unsuitable for practical application. Considering the spectral configurations of commonly used field spectrometers, the QE65 Pro spectrometer provides high-precision SIF data, while the ASD FieldSpec 3 spectrometer demonstrates optimal performance with the SFM and DOAS algorithms, achieving RRMSE values of 9.08% and 9.56%, respectively. The influence of atmospheric water vapor column concentration on SIF retrieval in this band is substantial, with retrieval accuracy positively correlated with water vapor content. The sFLD, iFLD, and DOAS algorithms are particularly sensitive to variations in water vapor concentration.

(2) In ground-based tower platform SIF retrieval, multiple algorithms demonstrate reliable SIF retrieval capability in this band, with retrieval errors (RMSE) consistently remaining below 1.0 mW/m²/nm/sr. Among them, the SFM algorithm achieves the best performance, whereas the sFLD algorithm continues to exhibit severe overestimation. The study also verifies that atmospheric correction has minimal influence on SIF retrieval in this band. The RMSE and R² values between retrieved SIF data and photosynthetically active radiation (PAR) show no significant differences before and after atmospheric correction. However, illumination conditions exert a significant impact on retrieval accuracy, with clear-sky conditions yielding significantly higher retrieval accuracy than cloudy conditions. Notably, the sFLD algorithm exhibits abnormal behavior under cloudy conditions, likely due to the suppression of its overestimation effect under reduced illumination. In algorithm evaluation, SIF retrievals from the sFLD and SVD algorithms exhibit strong correlations with the near-infrared reflectance of vegetation (NIRvR), with R² values of 0.77 and 0.72, respectively. The iFLD algorithm performs relatively poorly, with an R² value of approximately 0.3, indicating low algorithm stability. Other algorithms show no significant differences in retrieval accuracy, with R² values stabilizing around 0.6. Analysis of daily SIF variations along the day-of-year (DOY) axis reveals strong temporal and spatial consistency between SIF retrievals in the water vapor absorption band and those in the O₂-A band.