Detection and application of sunlight-induced chlorophyll fluorescence based on spaceborne hyperspectrometer

Figure 1. Distribution of SIF in Northern Nigeria, Africa, as of April 8, 2025.

Currently, most sunlight-induced chlorophyll fluorescence (SIF) data are in discrete point form with sparse spatial coverage.

We used a hyperspectral instrument with a spectral resolution of 0.48 nm to conduct imaging high-resolution chlorophyll fluorescence exploration, hoping to fill the gap in the current satellite platform for sunlight-induced chlorophyll fluorescence with higher spatial resolution.

Table 1. Key performance indicators of some spaceborne sensors with SIF detection capabilities

*Reference: Zhang, Lifu; Wang, Siheng; Huang, Changping. 2018. Satellite remote sensing inversion method for solar-induced chlorophyll fluorescence. Journal of Remote Sensing, 22(1): 1-12

SIF Calculation Fundamentals

A portion of the light energy absorbed by vegetation is not used for photosynthesis but is emitted as fluorescence in the form of long wavelengths. Chlorophyll fluorescence is located in the red and far-red light regions of vegetation (650-800 nm). As a byproduct of photosynthesis, fluorescence is organically linked to the mechanism of photosynthesis and can directly reflect the photosynthetic capacity of vegetation.

Fluorescence contributes to the apparent reflectance of the canopy; reflection peaks at 685 and 760 nm can be observed using a sub-nanometer resolution spectrometer. At 685 nm, fluorescence contributes 10%–20% to the apparent reflectance, and at 740 nm, it contributes 2%–6%.

In the red border region of vegetation (650-800 nm), fluorescence contributes to apparent reflectance.

Advantages of SIF compared to vegetation indices: Reflectance-based vegetation indices cannot promptly capture early signs of abnormal vegetation growth in the initial stages of environmental anomalies, exhibiting a lag. SIF, on the other hand, can detect anomalies more sensitively.

Figure 2. Location of the hidden lines of Fraunhofer and Fehème

Fraunhofer dark line filling method: Due to atmospheric absorption, the solar spectrum contains many fine dark lines, known as Fraunhofer dark lines. Fluorescence signals can be quantitatively measured at specific wavelengths attenuated by the solar spectrum.

Solar radiation has three main absorption bands in the red and near-infrared regions: the Hα line at 656.3 nm (represented by atmospheric hydrogen absorption), and the O2 absorption lines in the Earth's atmosphere at 687 nm and 760 nm. These bands typically overlap with the chlorophyll fluorescence emission spectrum. Fluorescence remote sensing technology compares the depth of solar radiation spectral lines with the depth of plant radiation spectral lines, measuring the amount of fluorescence emitted from short-wavelength excitation to fill a Fraunhofer "well" to a certain extent. A Fraunhofer "well" represents an absorption line in a specific wavelength band, resembling a "well".

Assuming the canopy spectral reflectance R of vegetation in the Fraunhofer dark line and adjacent spectral regions is equal, the fluorescence intensity f of vegetation in the Fraunhofer dark line and adjacent spectral regions is calculated using the following formula:

Figure 3. Fraunhofer dark line fluorescence detection principle

a, b: The intensity of the solar irradiance spectrum measured using a reference plate in the Fraunhofer dark line and adjacent spectral regions.

c, d: The intensity of the irradiance spectrum reflected by the vegetation canopy in the Fraunhofer dark line and adjacent spectral regions.

λ0: The center wavelength of the Fraunhofer dark line.

Since R is independent of fluorescence, chlorophyll fluorescence can be measured simply by measuring the spectral reflectance, solar irradiance, and canopy reflectance in the Fraunhofer dark line band.

Different vegetation species and their conditions result in different ratios of relative fluorescence intensity values ​​at 760nm and 687nm calculated by FLD, which can be used as a basis for judging the physiological status of plants.

Data Sample

Figure 4. Example of vegetation pixel spectrum

Figure 5. O2A absorption peak

The image above shows the typical vegetation spectrum of satellite 04. An O2B absorption peak near 687 nm, an O2A absorption peak near 760 nm, and an H2O absorption peak near 720 nm can be observed. Compared with the standard Fraunhofer absorption spectrum, the spectral positions are accurate, and the effect of spectral shift is negligible (

SIF calculation

Improved Spline Interpolation Fraunhofer Dark Line Filling Method (iFLD)

Based on the Fraunhofer dark line filling method (FLD), improvements were made in the selection of reference bands and adjacent bands. A total of 8 bands were selected according to the absorption peak width, and spline functions were used to fit the background spectrum.

Figure 6. Band range used for fitting

Figure 7. Spline function fitting and interpolation results (Pixel1, 2, and 3 represent shrub, farmland, and building pixels, respectively).

As shown in the figure above, the dashed line represents the fitting results using eight bands near the O2A absorption peak. Since bands 211 and 212 are located on both sides of the O2A absorption peak, the average value, 760.43195 nm, is used as the radiation value affected by SIF, and also as the target wavelength for spline function fitting. In the figure above, at the bottom of the quadratic function, the difference between the original spectral value and the spline function fitted value is the part filled by SIF, which is the SIF intensity.

Based on actual terrain conditions, the vegetation cover of Pixel 1 shrub pixels is significantly lower than that of farmland, and the photosynthetic intensity is also lower than that of crops during their vigorous growth period. Therefore, the SIF value is significantly lower than that of Pixel 1 farmland. Pixel 3, being a built-up area, provides SIF with only a small amount of landscaping and green trees, resulting in the lowest SIF value.

SIF Result Verification

Figure 8. Comparison of calculation results for different absorption bands

The figure above shows the SIF results using different band positions and compares them with the NDVI from the same Sentinel-2 data. As shown, the SIF using the O2A band has the least noise, and the data distribution is close to that of the NDVI. The O2B band, due to its shallower absorption depth, shows significant noise, but a distribution trend is still discernible. The results calculated from non-SIF characteristic bands provide almost no useful information. This comparison demonstrates the effectiveness of our method.

Application Cases

Figure 9. Distribution map of SIF in farmland and forest areas

 Figure 10. NDVI Distribution Map of Farmland and Forest Areas 

Figure 11. High-resolution distribution map of NDVI in farmland area

Figure 12. Statistical comparison of root zone soil moisture and precipitation in crop and forest regions.

The SIF values ​​in the aforementioned forest and crop areas show significant differences and differ slightly from the NDVI distribution. In the crop area, because the original resolution of the sentinel image is 10 meters, after resampling to match the SIF resolution, the mixing of crops and soil in the lower left farmland area lowered the overall NDVI value. However, the high SIF phenomenon in the farmland is clearly observed in the SIF results. The forest area, with its mountainous terrain and vegetation consisting mainly of low shrubs and some trees, has a lower vegetation cover than the farmland area, resulting in a significantly lower SIF value. Furthermore, according to the statistical charts above, rainfall decreased in both areas after March, leading to a significant drop in soil moisture, with a greater decrease in the forest area than in the crop area. Since the crop area typically has irrigation facilities, while the mountainous area generally does not, it is speculated that the forest area may be experiencing drought stress, leading to a further decrease in SIF. This indicates that, compared to NDVI, SIF can more sensitively observe crop stress phenomena and identify stressed areas, providing a more powerful tool for ecological environment assessment and precision agriculture.