Open Access
Add to BibSonomy
Abstract
Optical remote sensing provides optimal technical support for the detection and quantification of floating macroalgae. Although the spatial scale effect on optical estimation of floating macroalgae coverage or biomass from different images has been clarified, the directional effect on them has not been investigated until now. In this study, synchronous multi-angle imaging spectroradiometer (MISR) and MODIS images were collected to investigate the multi-angle remote sensing of green tides. A dual thresholding method, based on the difference vegetation index (DVI) and scaled algae index, was employed to determine algae pixels. In addition, piecewise empirical models were developed for MISR and MODIS images to estimate the total biomass of green tides based on laboratory measurements and DVI values. Comparative analysis of DVI histograms and total biomass shows that the sensor zenith angle has a significant impact on the quantification of green tides. Under the same solar conditions, as the sensor zenith angle increases, the optical signals received from algae pixels weaken, resulting in a decrease in the quantification of green tides. In future research, the observation geometry (including the solar/sensor zenith angle and the solar/sensor azimuth angle) needs to be considered to improve the accuracy of optical remote detection and quantification of floating macroalgae.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Floating macroalgae (e.g., Ulva prolifera or Sargassum) are important components of marine ecosystems as well as key focal points for ocean color remote sensing [1–5]. Green tides, a marine ecological phenomenon, caused by the outbreak of Ulva (U.) prolifera, has occurred annually in the Yellow Sea (YS) of China since 2008 [6–8]. Large-scale green tides have profound implications for marine ecosystems and coastal ecological environments, resulting in significant social impacts and economic losses [9–10]. Satellite remote sensing is a critical approach for the monitoring of green tides as it can provide accurate, wide-range, and long-time series green tides data [11–12].
Various remote sensing techniques have been used for the monitoring of floating macroalgae, including optical remote sensing, thermal remote sensing [13], and synthetic aperture radar (SAR) [14]. Optical remote sensing is commonly used due to its high-revisit frequency and credible capacity for detection and quantification, and different algae indices, such as Floating Algae Index (FAI) [15], Alternative Floating Algae Index (AFAI) [16], Virtual-Baseline Floating macro-Algae Height (VB-FAH) [17], Scaled Algae Index (SAI) [18], and Normalized Difference Algae Index (NDAI) [19], were designed to extract floating macroalgae from optical images. By incorporating the reflectance spectra of floating U. prolifera, a statistical relationship between biomass (wet weight) per unit area (BPA) and algae indices can be established [11]. Then a piecewise unmixing model can be employed to enable the estimation of total biomass from optical remote sensing images with different spatial resolution. Although spatial scale effect on remote estimation of the biomass of green tides with different optical images has been clarified and resolved [8,10], the uncertainty arising from bidirectional reflectance distribution function (BRDF) of floating green tides has not been illuminated. Considering the green tides drift on the sea surface, the multi-angle synchronous optical observation could help to figure out the BRDF characteristic of green tides.
Multi-angle Imaging SpectroRadiometer (MISR) Instrument is an optical sensor equipped with the Terra satellite. Nine cameras onboard MISR are arranged with one pointing toward the nadir (designated AN (0°)), one bank of four pointing in the forward direction (AF (+26.1°), BF (+45.6°), CF (+60.0°), and DF (+70.5°) in order of increasing off nadir angle), and one bank of four pointing in the backward direction (AA (-26.1°), BA (-45.6°), CA (-60.0°), and DA (-70.5°)) (Fig. 1). Each camera has four spectral bands, three visible (443 nm, 555 nm, and 670 nm) and one near-infrared (865 nm) with spatial resolution ranging from 275 m to 1100 m. It should be noted that the nine cameras capture images within 7 minutes, so the quasi-simultaneous optical images of MISR with different sensor zenith angles are adopted to investigate the effect of viewing angle difference on the estimation of green tides biomass.
Fig. 1. MISR instrument imaging diagram. Nine push-broom cameras point at discrete angles along the spacecraft ground track, and data in four spectral bands are obtained for each camera.
In this study, MISR images at 9 viewing angles and the synchronous MODIS image on 25 June 2022 were collected. Algae pixels in these images can be successfully extracted using a dual thresholding method. Moreover, the biomass estimation models of green tides were generated based on the laboratory measurements of Difference Vegetation Index (DVI) and BPA. The relationship between DVI values and the biomass estimation results indicates that the sensor zenith angle has a significant impact on the quantification of green tides. Specifically, as the sensor zenith angle increases, the optical signals received from algae pixels weaken, resulting in a decrease in the quantification results of green tides.
2. Data and methods
2.1 MISR and MODIS data processing
MISR and MODIS Level 1B data of green tides in the YS on 25 June 2022 were obtained from NASA’s Goddard Space Flight Center. MISR and MODIS Level 1B data contain top-of-atmosphere (TOA) radiance (Lt, in the unit of W·m-2·sr-1·nm-1). TOA reflectance (R, dimensionless) of MISR and MODIS images can be determined using Eq. (1).
where F0 is the extraterrestrial solar irradiance and θ0 is the solar zenith angle. All of the MISR and MODIS images were resampled to 275 m and 250 m resolution, respectively. MISR data at 865, 670, and 555 nm, and MODIS data at 859, 645, and 555 nm were used as the Red, Green, and Blue channels to compose the false color images, respectively (Fig. 2).Fig. 2. False color images of MISR at nine viewing angles and synchronous MODIS Terra of green tides in the YS on June 25, 2022.
2.2 DVI and BPA model
DVI was employed to enhance the difference between green tides and seawater for MISR and MODIS images. Reflectance-based DVI can be defined using the following equation:
R is the TOA reflectance (dimensionless) and the subscripts NIR and RED represent the near-infrared and red bands, respectively. Based on laboratory measurements of U. prolifera BPA and the corresponding spectral reflectance (Fig. 3(a)), robust estimation relationships have been established to link BPA to DVI values of green tides detected from MISR and MODIS images. And then piecewise empirical models to estimate BPA form DVI values for MISR and MODIS images were developed (Fig. 3(b)).Fig. 3. (a) Reflectance spectra of floating U. prolifera macroalgae from a water tank experiment in Hu et al. [8]. (b) Green tides biomass estimation model of MISR and MODIS.
2.3 Determination of the algae pixels
The SAI algorithm was used to detect algae pixels from the DVI images [13,20]. SAI algorithm can effectively remove complex interference information caused by the differences in seawater and atmosphere background in the DVI image. This is done by subtracting the DVI value of a local pixel from each pixel in the DVI images using a kernel of an odd-numbered pixel square region (21 × 21 pixels in this study). The median DVI value within the square region is computed and subtracted from the DVI value of the central pixel. The SAI equations can be written the Eqs. (3) and (4).
where f(x,y) is the DVI value for each pixel generated from MISR and MODIS, g(x,y) is the median DVI value of the pixels within the window (W), and SAI(x,y) is the final subtracted DVI value of this pixel, i.e., the SAI value of the pixel. As shown in Fig. 4, the extraction of algae pixels was performed using a dual thresholding method. The first threshold (T1) was applied to DVI images to delineate the large clusters of algae pixels. These algae pixels have image features of high aggregation, thus are relatively easy to extract. For smaller patches of algae pixels, the complex background in DVI images may result in false detection. Then the SAI images (Eqs. (3) & (4)) were calculated for the remaining pixels. The second threshold (T2) segmentation was applied to SAI images to extract small patches of algae pixels. A more accurate extraction of algae pixels was finally determined with this two-step threshold segmentation.MISR is capable of capturing nine images within ±3 minutes, with the imaging time of AN as the starting point. These nine MISR images can thus be considered as synchronous observations, and the algae pixels should be kept the same for these nine images. In this study, the extraction pixels from the AN image were used as referenced extraction results for all these nine images.
3. Results and discussion
3.1 Floating algae difference against viewing angle in MISR Images
The algae pixels within the MISR and MODIS images could be discriminated from the background seawater and clouds using the dual thresholding method. The MISR detected green tides can be used for the simultaneous verification of multi-angles remote sensing of green tides, and the DVI values of algae pixels on 25 June 2022 are displayed in Fig. 5. All of the DVI values derived from MISR images fall within the range of -0.05 to 0.3. In order to visually represent the variations more effectively, DVI values are categorized as positive (greater than 0) and negative (less than 0) and represented with distinct color bars (Fig. 5).
The near-synchronous MISR images at nine viewing angles have identical solar zenith angles, solar azimuth angles, and solar radiation intensity. Therefore, the spectral variation of algae pixels of nine different MISR images can be caused by their different viewing angles (sensor zenith angles). As shown in Fig. 5, the percentage of negative DVI values (green pixels in Fig. 5) increases with the increase of the viewing zenith angle. For instance, the nadir viewing image (i.e., AN) has a relatively low percentage of 40.5% of negative DVI values compared with the other eight observations whose negative DVIs all exceeded 66%. It indicates that variations in sensor zenith angles represent the different sensor's reception of optical signals from green tides, resulting in such a detection variation.
3.2 Directional quantification difference of green tides
MODIS image on 25 June 2022 was used for comparison to evaluate the quantitative capability of MISR. Most published quantification studies of floating macroalgae are based on MODIS measurements because of its high-revisit frequency and long-term coverage [8]. Moreover, MODIS onboard Terra satellite has nearly identical image acquisition time, observation geometry (including the solar/sensor zenith angle and the solar/sensor azimuth angle), and spatial resolution (275 m for MISR AN and 250 m for MODIS Terra) with MISR AN image. So, in this study, it was effective to use MODIS images as a reference.
MISR and MODIS derived DVI is an important process parameter for estimating the variations in quantification of green tides caused by sensor zenith angle. Statistical histograms of derived DVI of each image were given to show the distribution of DVI values (Fig. 6). For MISR images, as the sensor zenith angle increases, the DVI histogram shifts to the left, indicating an overall decrease in DVI values (Fig. 6(a) and Fig. 6(b)). This is also demonstrated by the DVI values corresponding to the peaks (DVIP) of the histogram. The DVIP values for DF, CF, BF, AF, AN, AA, BA, CA, and DA are -0.021, -0.019, -0.016, -0.011, -0.002, -0.013, -0.015, -0.020, and -0.023, respectively. This suggests that as the sensor zenith angle increases, the signals received from green tides weaken. For MODIS image, algae pixels extraction was performed using the same method (Fig. 4) and its comparison with MISR AN image is shown in Fig. 6(c). The statistical histograms and accumulative percentage (Acc pct) curves between MISR AN and MODIS are similar. Their DVIP values are -0.002 and -0.003, respectively. In this study, the extraction pixels from the AN image were used as referenced extraction results for all nine MISR images. However, the overall decrease in DVI values (Fig. 6(a) and Fig. 6(b)) suggests that, if algae pixels are extracted independently, it is necessary to set different thresholds for each MISR image. The larger the sensor zenith angle, the lower the required DVI threshold. And the decrease in DVI values inevitably results in further variations in quantitative calculations of green tides(e.g., biomass and coverage area).
Fig. 6. (a) and (b) Comparison of DVI values distribution of algae pixels between MISR images at nine viewing angles. The DVI value corresponding to the peak value (DVIP) at nine viewing angles has shown respectively. (c) Comparison of DVI histogram of algae pixels between MISR AN and MODIS image.
3.3 Directional reflection characteristics of floating green tides
To further compare the optical signal variations induced by the sensor zenith angle, four algae pixels (P1, P2, P3, and P4 labeled in Fig. 2) were selected from MISR images. Four points are located within the range of solar zenith angle from 20.5 to 21.5 degrees, and their differences in solar angles are negligible. Then DVI values of the four points from MISR images are compared. Moreover, based on the piecewise empirical models between BPA (kg/m2) and DVI values in Fig. 3(b), the total biomass of green tides of MISR and MODIS images can be calculated. The comparative results of DVI values and biomass are presented in Fig. 7. Among all viewing angles, the DVI values observed in the AN direction exhibited the highest values. However, as the sensor zenith angle increased, the DVI values gradually decreased (Fig. 7(a)). The estimation results of the total biomass of green tides also followed this pattern (Fig. 7(b)). The MODIS derived total biomass is 140.18 kilotons, close to 137.71 kilotons derived from MISR AN image (Fig. 7(b)). This is due to the fact that they have nearly identical image acquisition time, observation geometry, and spatial resolution. However, there are significant numerical differences among the MISR-derived biomass results. The maximum value (140.18 kilotons from AN) of the estimation results is nearly 2.7 times the minimum value (51.89 kilotons from DA), indicating that maximum variations in sensor zenith angle result in approximately 170% uncertainty. The main reasons for the difference in biomass estimation results among MISR images are as follows. 1) The decrease of DVI values caused by sensor zenith angle. 2) The decrease of DVI values leads to the exclusion of numerous algae pixels from the biomass estimation, as their DVI values fall below the lower limit (DVI = -0.003 in Fig. 3(b)) of the piecewise empirical model (Fig. 6(a) and Fig. 6(b)).
Fig. 7. (a) The DVI value corresponding to algae pixels labeled in MISR AN image in Fig. 1. (b) Total biomass of green tides derived from MISR and MODIS images on 25 June 2022.
4. Conclusion
Optical remote sensing provides the most effective technical support for the detection and quantification of floating macroalgae. However, the impact mechanism of sensor zenith angles on the quantification of floating macroalgae has not been clarified. In this study, simultaneous MISR and MOIDS images on 25 June 2022 were used to investigate the multi-angle remote sensing of green tides. These data were processed to generate DVI and SAI images. A dual thresholding method was then employed to detect algae pixels from MISR and MODIS images to obtain the DVI values of green tides. Based on laboratory measurements of BPA and the corresponding spectral reflectance, piecewise models to estimate BPA from DVI values were developed. These models were then applied to MISR and MODIS images to calculate the total biomass of each image. Comparative analysis of DVI histograms and total biomass can reveal the impact of sensor zenith angle on the quantification of green tides. Variations in sensor zenith angles have a significant impact on the quantification of green tides. Specifically, under the same solar conditions, as the sensor zenith angle increases, the optical signals received from algae pixels weaken, resulting in a decrease in the quantification results of green tides. Therefore, when detecting and quantifying the floating targets (including but not limited to macroalgae) over the ocean surface, the observation geometry (including the solar/sensor zenith angle and the solar/sensor azimuth angle) is an important factor that needs to be considered. And further studies are needed to explore methods for mitigating or normalizing the impacts caused by observation geometry.
Funding
National Natural Science Foundation of China (42071387).
Acknowledgments
We thank NASA for providing MISR and MODIS data (https://search.earthdata.nasa.gov/, https://ladsweb.modaps.eosdis.nasa.gov/), and thank Dr. Lianbo Hu from Ocean University of China for providing experimental data.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. C. Hu, D. Johnson, M. Wang, J.-P. Maréchal, D. Goodwin, and B. Lapointe, “Sargassum Watch Warns of Incoming Seaweed,” EOS Trans. Am. Geophys Union 97(22), 10–15 (2016). [CrossRef]
2. J. Gower and S. King, “The distribution of pelagic Sargassum observed with OLCI,” Int J Remote Sens 41(15), 5669–5679 (2020). [CrossRef]
3. S. Jin, Y. Liu, C. Sun, X. Wei, H. Li, and Z. Han, “A study of the environmental factors influencing the growth phases of Ulva prolifera in the southern Yellow Sea, China,” Mar. Pollut. Bull. 135(April), 1016–1025 (2018). [CrossRef]
4. C. Hu and M. He, “Origin and offshore extent of floating algae in Olympic sailing area,” Eos Trans. AGU 89(33), 302–303 (2008). [CrossRef]
5. M. Wang and C. Hu, “Predicting Sargassum blooms in the Caribbean Sea from MODIS observations,” Geophys. Res. Lett. 44(7), 3265–3273 (2017). [CrossRef]
6. L. Qi, C. Hu, Q. Xing, and S. Shang, “Long-term trend of Ulva prolifera blooms in the western Yellow Sea,” Harmful Algae 58, 35–44 (2016). [CrossRef]
7. D. Liu, J. K. Keesing, P. He, Z. Wang, Y. Shi, and Y. Wang, “The world’s largest macroalgal bloom in the Yellow Sea, China: Formation and implications,” Estuarine, Coastal Shelf Sci. 129, 2–10 (2013). [CrossRef]
8. C. Hu, L. Qi, L. Hu, T. Cui, Q. Xing, M. He, N. Wang, Y. Xiao, D. Sun, Y. Lu, C. Yuan, M. Wu, C. Wang, Y. Chen, H. Xu, L. Sun, M. Guo, and M. Wang, “Mapping Ulva prolifera green tides from space: a revisit on algorithm design and data products,” Int J Appl Earth Obs. 116, 103173 (2023). [CrossRef]
9. Q. Xing, C. Hu, D. Tang, L. Tian, S. Tang, X. Wang, M. Lou, and X. Gao, “World's largest macroalgal blooms altered phytoplankton biomass in summer in the Yellow Sea: satellite observations,” Remote Sens. 7(9), 12297–12313 (2015). [CrossRef]
10. T. Lu, Y. Lu, L. Hu, J. Jiao, M. Zhang, and Y. Liu, “Uncertainty in the optical remote estimation of the biomass of Ulva prolifera macroalgae using MODIS imagery in the Yellow Sea,” Opt. Express 27(13), 18620–18627 (2019). [CrossRef]
11. L. Hu, C. Hu, and M. He, “Remote estimation of biomass of Ulva prolifera macroalgae in the Yellow Sea,” Remote Sens. Environ. 192, 217–227 (2017). [CrossRef]
12. L. Hu, K. Zeng, C. Hu, and M. He, “On the remote estimation of Ulva prolifera areal coverage and biomass,” Remote Sens. Environ. 223, 194–207 (2019). [CrossRef]
13. Q. Song, C. Ma, J. Liu, and H. Wei, “Quantifying ocean surface green tides using high-spatial resolution thermal images,” Opt. Express 30(20), 36592–36602 (2022). [CrossRef]
14. L. Qi, M. Wang, C. Hu, and B. Holt, “On the capacity of Sentinel-1 synthetic aperture radar in detecting floating macroalgae and other floating matters,” Remote Sens. Environ. 280, 113188 (2022). [CrossRef]
15. C. Hu, “A novel ocean color index to detect floating algae in the global oceans,” Remote Sens. Environ. 113(10), 2118–2129 (2009). [CrossRef]
16. M. Wang and C. Hu, “Mapping and quantifying Sargassum distribution and coverage in the Central West Atlantic using MODIS observations,” Remote Sens. Environ. 183, 350–367 (2016). [CrossRef]
17. Q. Xing and C. Hu, “Mapping macroalgal blooms in the Yellow Sea and East China Sea using HJ-1 and Landsat data: Application of a virtual baseline reflectance height technique,” Remote Sens. Environ. 178, 113–126 (2016). [CrossRef]
18. J. K. Keesing, D. Liu, P. Fearns, and R. Garcia, “Inter- and intra-annual patterns of Ulva prolifera green tides in the Yellow Sea during 2007-2009, their origin and relationship to the expansion of coastal seaweed aquaculture in China,” Mar. Pollut. Bull. 62(6), 1169–1182 (2011). [CrossRef]
19. W. Shi and M. Wang, “Green macroalgae blooms in the Yellow Sea during the spring and summer of 2008,” J. Geophys. Res. 114(C12), C12010 (2009). [CrossRef]
20. J. Liu, J. Liu, J. Ding, and Y. Lu, “A refined imagery algorithm to extract green tide in the Yellow Sea of China from HY-1C CZI measurements,” Haiyang Xuebao 44(5), 1–11 (2022).
