Open Access
Add to BibSonomy
Abstract
Suspended particulate matter (SPM) affects the optical properties of water, which can be used as a marker of water quality. The water quality of Lake Taihu has changed immensely since the 1980's. However, despite the link between water quality and SPM, long-term systematic studies on SPM have not been conducted in this lake. Here, we used Landsat-5 TM and Landsat-8 OLI data to model changes in the SPM concentration of Lake Taihu from 1984 to 2020. Various models were generated, calibrated, and finally validated using in situ SPM, remote sensing reflectance (Rrs) data, and synchronous satellite data. After comparing various commonly used models, it was found that the exponential model based on band combination [Rrs(red) + Rrs(NIR)/Rrs(green)] had the highest accuracy, with an average unbiased relative error greater than 35%. Subsequently, the SPM products of Lake Taihu during 1984–2020 were generated. Overall, the SPM concentration showed a downward trend over the study period, which might be primarily attributable to a decline in wind speed. These findings may assist in the conservation of Lake Taihu and its associated water resources.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Suspended particulate matter (SPM) is an umbrella term that refers to insoluble particles suspended in a water body. In addition to suspended sediments, SPM includes plankton, humus, and organic debris [1–3]. SPM is one of the most important water quality parameters, as it affects the distribution of light and various optical properties of water, including transparency and turbidity. Through these properties, SPM influences the heat budget and primary productivity of lakes, as well as the growth of aquatic organisms [4–6]. Therefore, it is essential to explore the long-term spatiotemporal variation of SPM to monitor sediment transport patterns. This approach would facilitate the development and application of various aquatic treatments, which are intended to culminate in improved water quality management strategies [7,8].
Monitoring SPM usually involves field sampling and laboratory-linked efforts. Although measurements acquired using traditional methods are accurate, the associated costs remain high. Thus, limited sample data have been obtained, leading to a rarity of analyses that bring forth long-term spatiotemporal trends of SPM [9,10]. However, remote sensing can be used for the large-scale and continuous dynamic monitoring of water bodies; and this approach has been widely used in SPM monitoring [11–15]. The SPM concentration can directly affect the intensity of satellite signals. Therefore, understanding the spectral characteristics of water bodies with different turbidity levels forms the basis for establishing accurate remote sensing estimation algorithms [16]. Generally, the spectral reflectance of a water body increases with an increase in SPM concentration [14,17,18]; and the band most sensitive to such changes will move to a longer wavelength [10,18]. At present, there are some SPM estimation models built on the basis of different degrees of turbidity; for example, some of these models are specifically designed to retrieve SPM in moderately turbid waters, including in the Louisiana coast/Tampa Bay/micro-tidal river plume in the southeastern United States; such models work based on the AVHRR/MODIS/Landsat OLI red band model [19–21]. Other examples include estimation models fine-tuned for highly turbid environments, such as Lake Taihu, Poyang Lake, Hongze Lake, and the Gironde Estuary; these models work based on the HJ-CCD/ MODIS/MERIS/SPOT near-infrared (NIR) band or the red and NIR band ratio [8,9,22–26]. Nechad et al. [27] also developed a model for estimating SPM using the 520–885-nm band by constructing a band look-up table, which selects the reference band according to the turbidity of the water body.
Several types of satellite data can be used to analyze long-term changes in SPM. For example, MERIS data can be used to estimate the changes in SPM from 2003–2012 [8,28], while MODIS data can be used to estimate such changes from 2000 to the present [22,29]. Compared with these satellite data, Landsat data can be traced back to 1984, although with some limitations in the accuracy of atmospheric corrections (especially for Landsat 5 TM) and the complex optical characteristics of inland water bodies. Only a few regional SPM estimates have been derived based on Landsat data since the 1980s. For example, Du et al. [30] used Landsat TM/ETM + / OLI red and blue band ratios to retrieve the SPM of water bodies in the Songnen Plain, and Montanher et al. [31] used a multivariate linear model based on Landsat visible and NIR bands to retrieve the SPM of Amazonian white-water rivers.
Lake Taihu is the third largest freshwater lake in China, and is surrounded by the most developed and populous area in China. After China’s reform and opening up in the late 1970s, the economy developed rapidly; this development was accompanied by an increase in anthropogenic activities associated with economic productivity, which subsequently led to significant changes in the water quality of Lake Taihu [32,33]. Due to runoff and sediment resuspension, the concentration of SPM in Lake Taihu was great and highly variable [3,22,34]. Previously, Zhang et al. [8] used the red-edge band of MERIS to retrieve the SPM of Lake Taihu from 2003–2011, and Shi et al. [22] used the red band of MODIS to retrieve the SPM of Lake Taihu, extending the coverage from 2003 to 2013; meanwhile, Hou et al. [29] used MODIS data to retrieve the SPM of the lakes in the middle and lower basins of the Yangtze River Plains, including Lake Taihu, from 2000 to 2014 based on the ratio of red and green bands. However, high spatial-resolution research on the long-term variation of SPM in Lake Taihu based on Landsat data remains lacking.
Since 2018, the United States Geological Survey (USGS) has published surface reflectance (SR) data for Landsat TM, ETM+, and OLI [35]. This data can be used to estimate water quality parameters after applying a simple correction [4,36]; this includes estimating the water clarity of Lake Taihu from 1984–2019 [37]. In this study, we aim to establish an estimation model of SPM for Lake Taihu based on Landsat SR data, monitor the spatiotemporal changes in SPM concentration since 1984, and analyze the main factors that likely affected such variations.
2. Data and methods
2.1 Study area
Lake Taihu (30°56′-31°14′ N, 119°54′-120°36′ E) is the third largest freshwater lake in China. Located at the southern edge of the Yangtze River Delta and covering an area of ∼2338 km2, Lake Taihu is a typical shallow lake, with a mean water depth of 1.9 m [38,39]. To analyze the spatial distributions of SPM in different aqueous environments that characterize the lake, Lake Taihu was divided into six sub-regions, as shown in Fig. 1. Specifically, these were Meiliang Bay, Zhushan Bay, Gonghu Bay, East coast, Open area, and East Lake Taihu [22,38].
Fig. 1. Map of Lake Taihu in the lower basin of the Yangtze River Delta divided into six sub-regions, with field sample sites indicated by red dots.
2.2 Datasets
2.2.1 In situ data
We conducted 14 field surveys in Lake Taihu from 2005–2020 and obtained 413 SPM measurements with corresponding remote sensing reflectance (Rrs) spectra and 22 SPM measurements without these spectra. The maximum, minimum, average, and standard deviation (S.D.) determined using in situ SPM measurements from the 435 sample sites were 285.6, 6.0, 68.3, and 54.9 mg/L, respectively. The spatial distribution of all sampling sites obtained in the field surveys is shown in Fig. 1.
In situ SPM data were obtained via filtering water samples using a pre-weighed Whatman GF/F filter with a diameter of 47 mm. Subsequently, the filter was dried at 105°C and weighed using an electronic balance to obtain the total SPM concentration. The filter was then heated at 550°C and reweighed to obtain the inorganic suspended particulate matter (ISPM) concentration [21,26]. Moreover, water-surface spectral measurements were performed using an ASD FieldSpec spectroradiometer following the above-water method [40]. The radiance of the standard gray panel (Lp(λ)), water body (Lsw(λ)), and skylight (Lsky(λ)) were measured sequentially and then were used to calculate Rrs(λ) as follows:
Fig. 2. Seasonal variation of in situ Rrs(λ) obtained for 413 sample sites in Lake Taihu during the study period.
2.2.2 Landsat data
Landsat TM/ETM+/OLI SR data were provided by the USGS, though the data could also be obtained from the Google Earth Engine (GEE). There have been some studies in which GEE was used to process long-term Landsat time series data [42,43]. Landsat TM and ETM+ SR data products were generated from the specialized Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS) [44]. For Landsat OLI SR data, the Landsat surface reflection code was used [45].
In this study, the Landsat SR data of Lake Taihu from 1984–2020 were obtained from the GEE (1984–2011: Landsat TM; 2013–2020: Landsat OLI). As ETM+ images were not used due to striping noise, there were no images for 2012; the lack of data for this year only had a marginal impact on the analysis of long-term spatiotemporal variations in SPM over the 37-year study interval. Through visual inspection, 226 Landsat SR data of good quality (TM: 168 scenes, OLI: 58 scenes) were finally obtained, followed by the elimination of scenes severely affected by clouds and sun glint. The number of images obtained in each season (Spring: Mach–May; Summer: June–August.; Autumn: September–November; Winter: December–February) of each year is shown in Fig. 3.
The revisit period for Landsat is long (16 d). Field data with a difference of ±1 d from the Landsat data were considered to be synchronized with satellite images if there was no precipitation in the adjacent two days and the wind did not exceed a light breeze (i.e., wind speed < 3.3 m/s). With this approach, the SPM of two adjacent days in Lake Taihu can be relatively stable. When the pixels did not detect ice, snow, clouds, or other potential noise, the median Rrs from a 3×3 window of the Landsat images centered on each sampling point was extracted as a matching sample, provided that the coefficient of variation (S.D. divided by the mean) in the window was less than 0.3 [46]. After testing 435 sampling points, 69/45 matched pairs were found between the in situ SPM and Landsat TM/OLI SR data.
2.2.3 Meteorological data
One of the main factors affecting the variation in the water quality of Lake Taihu is climate change (Yu, 2015). Wind speed and precipitation data from 1984 to 2018 were obtained from the China Meteorological Forcing Dataset (CMFD), which is a gridded dataset with high temporal (3 h) and spatial (0.1°) resolutions [47–49]. These data were used to analyze the driving factors of variations in SPM. Based on all of the data from one day, we calculated the daily, monthly, and annual mean meteorological data.
2.3 Calibration of the SPM model
Before modeling, we first converted the field Rrs(λ) data into satellite spectra. The relative spectral response function of Landsat TM and OLI were used to convert the field Rrs(λ) data to the Rrs(λi) of the Landsat TM and OLI equivalent bands. Briefly, from 435 in situ data, we used 321 pairs of in situ simulated TM/OLI Rrs(λi) and synchronized SPM data for calibration and 69/45 pairs of TM/OLI satellite Rrs(λi) and synchronized SPM for validation. The numbers of data pairs used to calibrate and validate the SPM models are shown in Table 1. Additionally, to construct the best SPM model, we compared the single band–band combination models commonly used in other studies. We conducted Pearson’s correlation analysis and significance testing for each model and selected the best-fit model to estimate the SPM in Lake Taihu.
Table 1. Data for calibration and validation of SPM estimation models
2.4 Model validation
To evaluate the performance of the model, the correlation coefficient (r) between the in situ and estimated values, as well as the root mean square error (RMSE) and the average unbiased relative error (AURE) of the estimation were calculated. The AURE and RMSE were calculated as follows [50,51]:
2.5 SPM products based on Landsat SR data
We applied the tested SPM estimation model to Landsat SR data, to produce SPM products in Lake Taihu. It was first necessary to extract the water boundary, which is dynamic due to the continuous changes in lake volume. Therefore, to extract the water boundary more accurately, we used monthly water history datasets from 1984–2020 (JRC/GSW1_3/MonthlyHistory, Pekel et al., 2016), provided by the European Commission’s Joint Research Centre (JRC) on GEE [52]. Additionally, we used the quality assessment band of Landsat SR data to remove clouds and cloud shadows.
We further processed the Landsat SR data, and converted it to remote sensing reflectance data (Rrs). As Landsat SR data are essentially land surface reflectances, Rrs is commonly used in ocean color remote sensing. We applied a simple method to convert SR to remote sensing reflectance in which we subtracted the minimum value of the shortwave infrared (SWIR) band from the visible and NIR bands and then divided it by π [37,53–55].
Cyanobacterial blooms can change the reflection characteristics of light on the surface of water, and it is impossible to accurately estimate SPM in areas with cyanobacterial blooms. Therefore, it was necessary to identify and remove such areas in this study. The floating algae index (FAI) proposed by Hu et al. (2010) [56] was used to extract cyanobacteria blooms, and the optimal threshold of FAI in each image was determined by visual interpretation to extract cyanobacteria scene-by-scene. Finally, the best-fit SPM model was used to produce the long-term SPM products of Lake Taihu from 1984–2020 on GEE.
2.6 Spatiotemporal changes in SPM concentrations
Having a small number of images available each year might reduce the accuracy of analyses intended to illuminate the interannual fluctuations of SPM in Lake Taihu. Therefore, we considered every three years (e.g., 1984–1986, 1987–1989, etc.) as the analytical period to reduce the impact of insufficient data each year and to analyze the variation of SPM more accurately over time. There were only three years from 2011–2014 because of the data gap in 2012. Additionally, to reduce the influence of the uneven number of images in different seasons on the interannual variation in SPM, we first calculated the mean SPM in each season every three years and then calculated the three-year mean from the seasonally averaged values. To analyze the interannual variation of SPM across seasons, we divided the four seasons into two groups (i.e., spring/winter, summer/fall) to further mitigate the impact of few images in each season. The grouping of the seasons was done based on the growth of cyanobacteria. The three-year SPM outlier data were recognized as those outside the µ ± 2σ window, where µ and σ denote the mean and S.D., respectively; these outliers were excluded prior to the calculation of the mean. This approach minimized the impact of uncertainties caused by extreme weather events, thin clouds, sun glint, and noise, to obtain more stable and reliable results [57].
3. Results
3.1 Calibration and validation of the SPM model
Before constructing the SPM model, we first evaluated the accuracy of the remote sensing reflectance correction method mentioned in Section 2.5. This was done by subtracting the minimum value of the shortwave infrared (SWIR) band from the visible and NIR bands, and then dividing the resultant value by π. In our previous study [37], we have made a detailed accuracy evaluation of Landsat corrected Rrs based on the synchronous in situ Rrs. Here we cite some main results in Yin et al. [37]. It was found that the average AURE of corrected Landsat TM / OLI Rrs and in situ simulated Rrs were 29.29% and 25.28%, and the average R2 were 0.68 and 0.74 in the visible and near-infrared bands, respectively, indicating a high consistency between the absolute values of the corrected Rrs and in situ Rrs. Furthermore, we evaluated the similarity between the corrected Rrs and the in situ Rrs shapes. To accomplish this, we calculated the correlation coefficient between the corrected Rrs spectra of each synchronization point (TM: 1–4 bands, OLI: 1–5 bands) and the corresponding in situ simulated Rrs spectra. The results show that for TM and OLI, the mean values of the correlation coefficient of all synchronous sampling sites between the image and the in situ Rrs were 0.98 and 0.95, respectively, and the S.D. values of the correlation coefficient were 0.03 and 0.05, respectively, indicating a high consistency between the spectral shape of satellite and the in situ Rrs. The results showed that both the absolute value of Rrs and shape of Rrs are close to the in situ Rrs, indicating the effectiveness of the remote sensing reflectance correction method used in this study.
As shown in Table 1, we used 321 pairs of in situ SPM and simulated satellite Rrs(λi) data for SPM model calibration and 69/45 pairs of in situ SPM and synchronized TM/OLI satellite Rrs(λi) data for model validation. We compared 11 estimation models based on a single band or the spectral index, including the red band, NIR band, and NIR and green band ratio algorithms, among others. The model coefficients in these models have been calibrated using field data (321 pairs of in situ SPM and in situ Rrs(λ) data converted into Rrs(λi) of Landsat). The results are shown in Table 2.
Table 2. Calibration and validation of the various SPM models in Lake Taihu (the published models were further tuned using the in situ data in this study)
By comparing different models, we found that the coefficient of variation (R2) of the exponential model based on $({\textrm{R}_{\textrm{rs}}}({\textrm{red}} )+ {\textrm{R}_{\textrm{rs}}}({\textrm{NIR}} ))/{\textrm{R}_{\textrm{rs}}}({\textrm{green}} )$ was the highest (TM: 0.83 and OLI: 0.81) (Fig. 4). The validation accuracy of this model for Landsat SR images was also the highest (AURE of 34.4% and 23.5% for TM and OLI, respectively) (Fig. 5). Therefore, this model was chosen to produce the SPM products for Lake Taihu from 1984–2020.
Fig. 4. Calibration of the proposed SPM model using Landsat SR data from (a) Landsat 5 TM and (b) Landsat 8 OLI.
Fig. 5. Validation of the proposed SPM model fusing Landsat SR data from (a) Landsat 5 TM and (b) Landsat 8 OLI.
3.2 Spatial distribution of satellite-derived SPM
The remote sensing reflectance images of Landsat SR data from 1984–2020 were generated using the method described in Section 3.1. Subsequently, the spatiotemporal distribution of SPM was derived using these atmospheric-corrected satellite images and the best-fit SPM model. The three-year average SPM products for Lake Taihu, based on Landsat TM and OLI data from 1984 to 2020, can be downloaded from [62]. We first calculated the spatial distribution of the mean SPM of Lake Taihu from 1984 to 2020 (Fig. 6). The 37-year average SPM showed an obvious spatial pattern (Fig. 6(a)), and the SPM concentration across the lake ranged from 16.1 mg/L (for the clearest waters) to 116.5 mg/L (for turbid waters), with the mean value being 51.85 mg/L. Among the six sub-regions in Lake Taihu, the highest SPM concentration was observed in the open area, especially in the south; the SPM values in the five bays were comparatively lower. This pattern was mainly a result of the open area being more affected by winds and forming larger waves, resulting in sediment resuspension [22,63]. Importantly, the spatial distributions of SPM reported in this study were consistent with those reported previously [22,29,64].
Fig. 6. Spatial distribution (a) and S.D. (b) of SPM in Lake Taihu based on Landsat SR data gathered from 1984–2020.
We estimated the spatial distribution of SPM S.D. values from 1984–2020 (Fig. 6(b)). The 37-year mean S.D. of Lake Taihu was 27.1 mg/L. The S.D. in the south of the open area was the highest, while the lowest S.D. values were estimated for the northern bay areas, indicating that the variation in SPM in the open area was greater than that in other areas of the lake. We also calculated the spatial distribution of SPM during each of the four seasons (Fig. 7(a)–(d)). As can be seen in Fig. 7, the SPM of Lake Taihu varied seasonally. The multi-year mean SPM in the four seasons was close in spring and winter, and close in summer and fall. The mean SPM concentrations in spring (March–May), summer (June–August), fall (September–November), and winter (December–February) were 59.1, 37.6, 37.4, and 67.6 mg/L, respectively. Overall, the water of the entire lake was more turbid in spring and winter than in summer and fall. Notably, the regular structure in East Taihu Lake (Fig. 6 and Fig. 7) is due to the emerging and vanishing of cage culture area, with the rise and fall of water level and the change of relevant policies on cage culture. In this study, the water body boundary was extracted separately for each scene, and in some images, the cages above the water surface were removed from the water body area; in some images, the cages above the water surface were not seen and were retained in the water body area. When these images were superimposed together, the shape of the cage area still appeared, but this did not affect the calculation of the annual mean SPM.
Fig. 7. Spatial distribution of SPM in Lake Taihu in spring (a), summer (b), fall (c), and winter (d) based on Landsat SR data gathered from 1984–2020.
3.3 Interannual variation of SPM
We investigated the interannual (every three years) variation of SPM (Figs. 8 and 9). Overall, the SPM for the entire lake showed a significant downward trend from 1984–2020 (R2 = 0.67, P < 0.05). Among the six sub-regions, four showed significant declines from 1984–2020 (P < 0.05); however, this was not verified in Zhushan Bay or Meiliang Bay in the north. Possible reasons for this are that they have small areas and relatively enclosed shapes, which make them less affected by wind and waves that might influence their turbidity; they also have many rivers flowing into them (i.e., the Taige, Caoqiao, Yincun, and Shedu Rivers).
Fig. 8. Landsat SR estimated SPM in Lake Taihu, in three-year intervals from 1984–2020. The solid black line represents the interannual variation, across the entire study period and the six sub-regions of Lake Taihu.
Fig. 9. SPM estimates for the six sub-regions of Lake Taihu based on Landsat SR, in three-year intervals from 1984–2020.
We analyzed the interannual variations in SPM for the entire lake during spring/winter and summer/autumn (Fig. 8). With respect to the spatial variation among the seasons (Fig. 7), the SPM in spring and winter showed significant declines over time (Fig. 8; R2 = 0.78, P < 0.05), which were not observed in summer or autumn (Fig. 8; R2 = 0.28, P = 0.08). In each analytical period, the SPM concentrations in spring and winter were significantly higher than those in summer and autumn.
4. Discussion
4.1 Driving forces of variations in SPM concentrations
With respect to the spatial distribution of SPM in Lake Taihu, we observed that the SPM in the open area (especially in the south) was the highest among all sub-regions. This may be because this area is more susceptible to wind shear, leading to the formation of larger waves and the consequent resuspension of sediments [56,65]. Additionally, because the dominant wind directions in Lake Taihu are north-northwest [22], the path of the wind in the open area was longer than those in other areas. The longer a wind fetch is, the greater is its force and the greater is the sediment resuspension. Moreover, the Xitiaoxi River is the largest inflowing river into Lake Taihu, and its outlet is close to the south of the open area. The inflowing discharge from the Xitiaoxi River can reach 2.7 billion m3 in a year, accounting for approximately 60% of the total water input from all rivers into the lake [22,66]. Therefore, it is expected that a large amount of particles will flow into the south of the open area through this river, contributing for the significant increase in the SPM in the area. Conversely, the Zhushan Bay and Meiliang Bay areas in the north are small, and thus, the path of the wind is shorter than that in the open area, and the wind waves are smaller, which explains the low SPM concentrations in these areas.
The pattern of interannual variation in SPM in Lake Taihu showed a declining trend over 37 years. Many studies have shown that wind-induced solid resuspension is the most direct and important factor affecting changes in the SPM of shallow lakes [22,38,63]. Moreover, some studies have suggested that precipitation may also affect the spatiotemporal distribution of SPM in inland waterways [22,29,30]. Thus, we obtained three-year mean data on wind speed and precipitation in Lake Taihu using the CMFD dataset. Subsequently, the correlation between wind speed and precipitation and SPM concentration were analyzed (Fig. 10).
Fig. 10. The relationship between wind speed and SPM concentration (left); and precipitation and SPM concentration (right), in Lake Taihu.
Figure 10 shows that from 1984 to 2018, the mean wind speed decreased significantly from 3.4 to 1.8 m/s (P < 0.05). A significant positive correlation between SPM and wind speed was found (r = 0.85, P < 0.001), suggesting that the decline of SPM over the past 37 years may be closely related to a decrease in the wind speed over the same period. Moreover, the SPM concentration showed a significant negative correlation with precipitation (r = -0.70, P < 0.001). Precipitation can increase water levels and dilute sediment loads, thereby decreasing the SPM [30,67]. Due to the higher likelihood of high wind speeds causing sediment resuspension, we further analyzed the relationship between the mean number of days with a daily average wind speed greater than a light breeze (i.e., wind speed > 3.3 m / s) and SPM; this analysis was conducted in three year intervals (Fig. 11). The results showed that there was a significant positive correlation between high wind speeds and SPM concentrations (r = 0.87, P < 0.001). We also analyzed factors driving SPM variations in spring and winter, as they dominated the interannual changes in SPM (Fig. 8). We observed an extremely high positive correlation between wind speed and SPM concentration (r = 0.88, P < 0.001); in comparison, there was a lower but still significant negative correlation between precipitation and SPM (r = -0.61, P < 0.001). These results suggest that the changes in SPM concentration in Lake Taihu in spring and winter have mainly been driven by variations in wind speed. It can also be found that there is a peak in SPM from 2005 to 2007, which did not correspond to the positive correlation with wind speed. This might be due to the strong cyanobacterial bloom outbreak in Lake Taihu in 2007 [56,68], which increased SPM by increasing cyanobacterial organic suspended particles. It should be noted that the analysis of factors driving the spatiotemporal variation of SPM has been conducted over a long period; as such, a different set of factors could have been gleaned if we had performed a short-term study.
Suspended particulate matter can be divided into ISPM and organic suspended particulate matter (OSPM), with the former including abiotic amorphous particles and mineral debris, and the latter including plankton, aging or dead algae, microorganisms, and other biogenic particles [69,70]. Among the in situ SPM measurements from the 435 sampling sites, ISPM and OSPM concentrations were determined at 400 sites. To further explore the composition of SPM in Lake Taihu, we analyzed the correlation between the ISPM vs. OSPM concentration and the total SPM concentration. The ISPM concentration accounted for the majority of the SPM in Lake Taihu (mean: 87%) (Fig. 12; R2 = 0.99, P < 0.001). These results suggest that the SPM is mainly composed of inorganic matter, which is extremely sensitive to resuspension caused by wind and is consistent with our results that suggest that wind speed is the most important factor affecting changes in SPM in Lake Taihu.
Fig. 12. Relationship between in situ ISPM concentration and the overall SPM concentration measured at 400 sampling sites in Lake Taihu.
With the rapid development of the Chinese economy and society since the 1980s, Lake Taihu has been increasingly affected by human activities. Among them, sand mining is the key factor that could affect SPM concentrations [14]. However, because sand content in the lake sediments is not as rich as that of other lakes, such as Lake Poyang, and due to the strict management and control by relevant departments, Lake Taihu does not appear to have been significantly affected by sand mining activities [71].
4.2 Applicability of the SPM model
In this study, we used Landsat SR data obtained from the USGS. However, as these were SR data, which cannot meet the requirements of water quality inversion, a simple remote sensing reflectance correction method was employed, and the corrected Rrs data were used to estimate SPM concentrations. Our results showed that the Rrs correction method was effective, with mean AURE values of 29.3% and 25.3% for TM and OLI in visible and NIR bands, respectively. Additionally, we established an SPM model based on $({{R_{rs}}(red) + {R_{rs}}(NIR)/{R_{rs}}(green)} )$. The band ratio method was not easily affected by thin clouds, residual aerosols, or bidirectional reflectance effects, and the accuracy of the algorithm was mainly affected by the spectral shape of the TM and OLI derived Rrs. After calculation, for TM and OLI, the mean spectral correlation coefficients at all synchronous sampling points were as high as 0.98 and 0.95, while the S.D. values were as low as 0.03 and 0.05, respectively. Overall, these results indicate that the Rrs shapes after correction were close to the measured Rrs shapes, thus meeting the accuracy requirements of SPM inversion.
Some studies have shown that there are clear differences in the Rrs spectra of clear and turbid waters. With increases in SPM, the Rrs of water in the visible and NIR bands can also increase; in particular, the Rrs peaks in the red and NIR bands can increase significantly [14,17,18]. When the SPM concentration is moderate, the Rrs peaks may be in the range of 550–670 nm; in this case, it is suitable to use the red band to estimate SPM. Conversely, when the SPM concentration is relatively high, the Rrs peak may shift to a wavelength of 780–830 nm; in this case, it is better to use the NIR band for SPM modeling. Lake Taihu has both very turbid and slightly turbid waters (in situ SPM: 6.0–285.6 mg/L). Therefore, we selected the band combination $({{R_{rs}}(red) + {R_{rs}}(NIR)/{R_{rs}}(green)} )$ to construct the SPM model, which can be applied to very turbid and slightly turbid waters.
5. Conclusions
With rapid socioeconomic development in China since the 1980s, notable changes have occurred in the aqueous environment of Lake Taihu. In this study, we established and validated an SPM model of Lake Taihu based on Landsat TM/OLI satellite data and the spatiotemporal changes in SPM concentrations from 1984–2020, and the possible influencing factors were investigated. Using the 321 in situ SPM and TM/OLI band-simulated Rrs out of synchronization with TM and OLI images, we tested and optimized various commonly used SPM models. For Landsat SR data, Rrs correction was conducted to convert SR into remote sensing reflectance by subtracting the minimum value of the SWIR band from the visible and NIR bands and then divided it by π. The corrected satellite Rrs and synchronous in situ SPM data were then used to validate the models. Our results showed that the exponential model based on $({{R_{rs}}(red) + {R_{rs}}(NIR)/{R_{rs}}(green)} )$ had the highest accuracy in estimating SPM in Lake Taihu, with the AURE < 35%. Therefore, we used this model to produce SPM estimates from 1984–2020.
We analyzed the spatial variation of SPM in Lake Taihu. Among the six sub-regions, the SPM concentration in the open area was the highest, while those in East Lake Taihu and the northern bays were the lowest. This spatial pattern likely occurred due to the wind speed, wind direction, and rivers entering the lake. To reduce the impact of insufficient data within one year, we calculated the three-year mean SPM and then analyzed the interannual variations. From 1984–2020, the SPM concentration in the entire lake and in four of the six sub-regions (East Coast, Open Area, East Lake Taihu, and Gonghu Bay) showed significant downward trends, which might be primarily attributable to the decrease in wind speed over the same period. The SPM model based on $({{R_{rs}}(red) + {R_{rs}}(NIR)/{R_{rs}}(green)} )$ proposed in this study could be applied to moderate- to high-turbidity lakes with similar SPM concentrations. Additionally, the spatiotemporal variations in SPM observed in this work could provide auxiliary support for decision-making and a critical reference for environmental management departments responsible for the conservation of Lake Taihu.
Funding
National Natural Science Foundation of China (41971318, 41901272, 41901304); Strategic Priority Research Program of the Chinese Academy of Sciences (XDA19080304).
Acknowledgments
We would like to thank the U.S. Geological Survey (USGS) for providing Landsat data.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are available in Ref. [62].
References
1. C. E. Binding, D. G. Browers, and E. G. Mitchelson-Jacob, “Estimating suspended sediment concentrations from ocean colour measurements in moderately turbid waters; the impact of variable particle scattering properties,” Remote Sensing of Environment 94(3), 373–383 (2005). [CrossRef]
2. C. E. Binding, J. H. Jerome, R. P. Bukata, and W. G. Booty, “Spectral absorption properties of dissolved and particulate matter in Lake Erie,” Remote Sensing of Environment 112(4), 1702–1711 (2008). [CrossRef]
3. L. Feng, C. M. Hu, X. L. Chen, and Q. J. Song, “Influence of the Three Gorges Dam on total suspended matters in the Yangtze Estuary and its adjacent coastal waters: Observations from MODIS,” Remote Sensing of Environment 140, 779–788 (2014). [CrossRef]
4. B. Zhang, J. S. Li, Q. Shen, Y. H. Wu, F. F. Zhang, S. L. Wang, Y. Yao, L. N. Guo, and Z. Y. Yin, “Recent research progress on long time series and large scale optical remote sensing of inland water,” National Remote Sensing Bulletin 25(1), 37–52 (2021).
5. D. Knighton, Fluvial Forms and Processes: A New Perspective (Arnold, 1998).
6. K. A. Moore, R. L. Wetzel, and R. J. Orth, “Seasonal pulses of turbidity and their relations to eelgrass (Zosteramarina L.) survival in an estuary,” J. Exp. Mar. Biol. Ecol. 215(1), 115–134 (1997). [CrossRef]
7. J. R. Jensen, Remote Sensing of the Environment: An Earth Resource Perspective (Prentice Hall, 2000).
8. Y. L. Zhang, K. Shi, X. H. Liu, Y. Q. Zhou, and B. Q. Qin, “Lake topography and wind waves determining seasonal-spatial dynamics of total suspended matter in turbid Lake Taihu, China: Assessment using long-term high-resolution MERIS data,” PLoS One 9(5), e98055 (2014). [CrossRef]
9. J. S. Li, Q. Shen, B. Zhang, and D. M. Chen, “Retrieving total suspended matter in Lake Taihu from HJ-CCD near-infrared band data,” Aquat,” Aquat. Ecosyst. Health Manage. 17(3), 280–289 (2014). [CrossRef]
10. Y. H. Wang, Z. D. Deng, and R. H. Ma, “Suspended solids concentration estimation in Lake Taihu using field spectra and MODIS data,” Acta Scientiae Circumstantiate 27(3), 509–515 (2007).
11. E. O. John, S. Maritorena, B. G. Mitchell, D. A. Siegel, K. L. Carder, S. A. Garver, M. Kahru, and C. McClain, “Ocean color chlorophyll algorithms for SeaWIFS,” J. Geophys. Res. 103(C11), 24937–24953 (1998). [CrossRef]
12. H. T. Duan, L. Feng, R. H. Ma, Y. C. Zhang, and S. A. Loiselle, “Variability of particulate organic carbon in inland waters observed from MODIS Aqua imagery,” Environ. Res. Lett. 9(8), 084011 (2014). [CrossRef]
13. J. C. Ritchie, F. R. Schiebe, and J. R. McHenry, “Remote sensing of suspended sediment in surface water. Photogramm,” Eng. Remote Sens. 42, 1539–1545 (1976).
14. L. Feng, C. Hu, X. L. Chen, L. Q. Tian, and L. Q. Chen, “Human induced turbidity changes in Poyang Lake between 2000 and 2010: observations from MODIS,” J. Geophys. Res. 117(D20), 1–19 (2012). [CrossRef]
15. L. Sokoletsky, X. Yang, and F. Shen, “MODIS-based retrieval of suspended sediment concentration and diffuse attenuation coefficient in Chinese estuarine and coastal waters,” Proc. SPIE 9261, 1–25 (2014).
16. P. Forget, S. Ouillon, F. Lahet, and P. Broche, “Inversion of Reflectance Spectra of Nonchlorophyllous Turbid Coastal Waters,” Remote Sensing of Environment 68(3), 264–272 (1999). [CrossRef]
17. C. Petus, G. Chust, F. Gohin, D. Doxaran, J. M. Froidefond, and Y. Sagarminaga, “Estimating turbidity and total suspended matter in the Adour Riverplume (South Bay of Biscay) using MODIS 250-m imagery,” Cont. Shelf Res. 30(5), 379–392 (2010). [CrossRef]
18. R. L. Miller and B. A. McKee, “Using MODIS Terra 250 m imagery to map concentrations of total suspended matter in coastal waters,” Remote Sensing of Environment 93(1-2), 259–266 (2004). [CrossRef]
19. A. Ody, D. Doxaran, Q. Vanhellemont, B. Nechad, S. Novoa, G. Many, F. Bourrin, R. Verney, I. Pairaud, and B. Gentili, “Potential of High Spatial and Temporal Ocean Color Satellite Data to Study the Dynamics of Suspended Particles in a Micro-Tidal River Plume,” Remote Sens. 8(3), 245–281 (2016). [CrossRef]
20. C. M. Hu, Z. Q. Chen, T. Clayton, P. Swarzenski, J. C. Brock, and F. E. Muller-Karger, “Assessment of estuarine water-quality indicators using MODIS medium-resolution bands: Initial results from Tampa Bay, FL,” Remote Sensing of Environment 93(3), 423–441 (2004). [CrossRef]
21. S. W. Myint and N. D. Walker, “Quantification of surface suspended sediments along a river dominated coast with NOAA AVHRR and SeaWiFS measurements: Louisiana, USA,” International Journal of Remote Sensing 23(16), 3229–3249 (2002). [CrossRef]
22. K. Shi, Y. L. Zhang, G. W. Zhu, X. H. Liu, Y. Q. Zhou, H. Xu, B. Q. Qin, G. Liu, and Y. M. Li, “Long-term remote monitoring of total suspended matter concentration in Lake Taihu using 250 m MODIS-Aqua data,” Remote Sensing of Environment 164, 43–56 (2015). [CrossRef]
23. D. Doxaran, J. M. Froidefond, S. Lavender, and P. Casting, “Spectral signature of highly turbid waters Application with SPOT data to quantify suspended particulate matter concentrations,” Remote Sensing of Environment 81(1), 149–161 (2002). [CrossRef]
24. D. Doxaran, J. M. Froidefond, and P. Casting, “Remote-sensing reflectance of turbid sediment-dominated waters. Reduction of sediment type variations and changing illumination conditions effects by use of reflectance ratios,” Appl. Opt. 42(15), 2623–2634 (2003). [CrossRef]
25. G. F. Wu, L. J. Cui, J. J. He, H. T. Duan, T. Fei, and Y. L. Liu, “Comparison of MODIS-based models for retrieving suspended particulate matter concentrations in Poyang Lake, China,” International Journal of Applied Earth Observation and Geoinformation 24, 63–72 (2013). [CrossRef]
26. Z. G. Cao, H. T. Duan, L. Feng, R. H. Ma, and K. Xue, “Climate- and human-induced changes in suspended particulate matter over Lake Hongze on short and long timescales,” Remote Sensing of Environment 192, 98–113 (2017). [CrossRef]
27. B. Nechad, K. G. Ruddick, and Y. Park, “Calibration and validation of a generic multisensor algorithm for mapping of total suspended matter in turbid waters,” Remote Sensing of Environment 114(4), 854–866 (2010). [CrossRef]
28. F. Gao, Y. P. Wang, X. Y. Hu, C. Xu, and N. Horning, “Spatiotemporal variations of total suspended matter (TSM) in the Pearl River estuary using MERIS full-resolution (FR) level-2 TSM product,” Mar. Freshwater Res. 70(8), 1065 (2019). [CrossRef]
29. X. J. Hou, L. Feng, H. T. Duan, X. Chen, D. Y. Sun, and K. Shi, “Fifteen-year monitoring of the turbidity dynamics in large lakes and reservoirs in the middle and lower basin of the Yangtze River, China,” Remote Sensing of Environment 190, 107–121 (2017). [CrossRef]
30. Y. X. Du, K. S. Song, G. Liu, Z. D. Wen, C. Fang, Y. X. Shang, F. R. Zhao, Q. Wang, J. Du, and B. Zhang, “Quantifying total suspended matter (TSM) in waters using Landsat images during 1984–2018 across the Songnen Plain, Northeast China,” J. Environ. Manage. 262, 110334 (2020). [CrossRef]
31. O. C. Montanher, E. M. Novo, C. C. Barbosa, C. D. Renno, and T. S. Silva, “Empirical models for estimating the suspended sediment concentration in Amazonian white water rivers using Landsat TM,” International Journal of Applied Earth Observation and Geoinformation 29, 67–77 (2014). [CrossRef]
32. B. Q. Qin, G. W. Zhu, G. Gao, Y. L. Zhang, W. Li, H. W. Paerl, and W. W. Carmichael, “A drinking water crisis in lake taihu: China: linkage to climatic variability and lake management,” Environmental Management 45(1), 105–112 (2010). [CrossRef]
33. B. Q. Qin, L. Wei, G. W. Zhu, Y. L. Zhang, T. Wu, and G. Gao, “Cyanobacterial bloom management through integrated monitoring and forecasting in large shallow eutrophic Lake Taihu (China),” J. Hazard. Mater. 287, 356–363 (2015). [CrossRef]
34. M. H. Wang, W. Shi, and J. W. Tang, “Water property monitoring and assessment for China’s inland Lake Taihu from MODIS-Aqua measurements,” Remote Sensing of Environment 115(3), 841–854 (2011). [CrossRef]
35. U.S. Geological Survey, “Landsat surface reflectance data (ver. 1.1, March 27, 2019): U.S. Geological Survey Fact Sheet,” 1, 3034 (2015).
36. C. A. Baughman, B. M. Jones, K. K. Bartz, D. B. Young, and C. E. Zimmerman, “Reconstructing Turbidity in a Glacially Influenced Lake Using the Landsat TM and ETM+ Surface Reflectance Climate Data Record Archive, Lake Clark, Alaska,” Remote Sens. 7(10), 13692–13710 (2015). [CrossRef]
37. Z. Y. Yin, J. S. Li, Y. Liu, Y. Xie, F. F. Zhang, S. L. Wang, X. Sun, and B. Zhang, “Water clarity changes in Lake Taihu over 36 years based on Landsat TM and OLI observations,” International Journal of Applied Earth Observation and Geoinformation 102, 102457 (2021). [CrossRef]
38. K. Shi, Y. L. Zhang, G. W. Zhu, B. Q. Qin, and D. L. Pan, “Deteriorating water clarity in shallow waters: evidence from long term MODIS and in-situ observations,” International Journal of Applied Earth Observation and Geoinformation 102, 102457 (2021).
39. R. H. Ma and J. F. Dai, “Investigation of chlorophyll-a and total suspended matter concentrations using Landsat ETM and field spectral measurement in Taihu Lake, China,” International Journal of Remote Sensing 26(13), 2779–2795 (2005). [CrossRef]
40. J. L. Mueller, A. Morel, R. Frouin, C. Davis, R. Arnone, K. Carder, Z. Lee, R. G. Steward, S. Holben, B. Hooker, C. D. Mobley, S. McLean, M. Miller, C. Pietras, G. S. Fargion, K. D. Knobelspiesse, J. Porter, and K. Voss, “Ocean optics protocols for satellite ocean color sensor validation, revision 4, volume III: radiometric measurements and data analysis protocols,” National Aeronautical and Space Administration Report 21621, 1–72 (2003).
41. C. D. Mobley, “Estimation of the remote-sensing reflectance from above-surface measurements,” Appl. Opt. 38(36), 7442–7455 (1999). [CrossRef]
42. X. Zhang, L. Y. Liu, X. D. Chen, Y. Gao, and M. H. Jiang, “Automatically Monitoring Impervious Surfaces Using Spectral Generalization and Time Series Landsat Imagery from 1985 to 2020 in the Yangtze River Delta,” Journal of Remote Sensing 2021, 1–16 (2021). [CrossRef]
43. G. W. Chen, R. J. Jin, Z. J. Ye, Q. Li, J. L. Gu, Y. M. Luo, G. Christakos, J. Morris, J. Y. He, D. Li, H. W. Wang, L. Song, Q. X. Wang, and J. P. Wu, “Spatio-temporal Mapping of Salt Marshes in the Intertidal Zone of China during 1985–2019,” J. Remote Sens.9793626 (2022).
44. J. G. Masek, E. F. Vermote, N. E. Saleous, R. Wolfe, F. G. Hall, and K. F. Huemmrich, “A Landsat surface reflectance dataset for North America, 1990-2000,” IEEE Geosci. Remote Sensing Lett. 3(1), 68–72 (2006). [CrossRef]
45. E. Vermote, C. Justice, M. Claverie, and B. Franch, “Preliminary analysis of the performance of the Landsat 8/OLI land surface reflectance product,” Remote Sensing of Environment 185, 46–56 (2016). [CrossRef]
46. G. Zibordi and K. J. Voss, “Field radiometry and ocean color remote sensing,” In: Oceanography from space. Springer, Dordrecht. 307-334 (2010).
47. J. He, K. Yang, W. J. Tang, H. Lu, J. Qin, Y. Y. Chen, and X. Li, “The first high-resolution meteorological forcing dataset for land process studies over China,” Sci. Data 7(1), 1–11 (2020). [CrossRef]
48. K. Yang, J. He, W. J. Tang, J. Qin, and C. C. Cheng, “On downward shortwave and longwave radiations over high altitude regions: Observation and modeling in the Tibetan Plateau,” Agricultural and Forest Meteorology 150(1), 38–46 (2010). [CrossRef]
49. K. Yang and J. He, “China meteorological forcing dataset (1979-2015),” A Big Earth Data Platform for Three Poles.2016.
50. Z. P. Lee, S. L. Shang, C. M. Hu, K. P. Du, A. Weidemann, W. Hou, J. Lin, and G. Lin, “Secchi disk depth: a new theory and mechanistic model for underwater visibility,” Remote Sensing of Environment 169, 139–149 (2015). [CrossRef]
51. Z. P. Lee, S. L. Shang, L. Qi, J. Yan, and G. Lin, “A semi-analytical scheme to estimate Secchi-disk depth from Landsat-8 measurements,” Remote Sensing of Environment 177, 101–106 (2016). [CrossRef]
52. J. F. Pekel, A. Cottam, N. Gorelick, and A. S. Belward, “High-resolution mapping of global surface water and its long-term changes,” Nature 540(7633), 418–422 (2016). [CrossRef]
53. S. L. Wang, J. S. Li, B. Zhang, Q. Shen, F. F. Zhang, and Z. Y. Lu, “A simple correction method for the MODIS surface reflectance product over typical inland waters in China,” International Journal of Remote Sensing 37(24), 6076–6096 (2016). [CrossRef]
54. S. L. Wang, J. S. Li, B. Zhang, E. Spyrakos, A. N. Tyler, Q. Shen, F. F. Zhang, T. Kuster, M. K. Lehmann, Y. H. Wu, and D. L. Peng, “Trophic state assessment of global inland waters using a MODIS-derived Forel-Ule index,” Remote Sensing of Environment 217, 444–460 (2018). [CrossRef]
55. S. L. Wang, J. S. Li, B. Zhang, Z. P. Lee, and X. Zhang, “Changes of water clarity in large lakes and reservoirs across China observed from long-term MODIS,” Remote Sensing of Environment 247, 111949 (2020). [CrossRef]
56. C. M. Hu, Z. P. Lee, R. H. Ma, K. Yu, D. Li, and S. L. Shang, “MODIS observations of cyanobacteria blooms in Taihu Lake: China,” J. Geophys. Res. Oceans 115, 261–263 (2010).
57. S. L. Wang, J. S. Li, W. Z. Zhang, C. Cao, F. F. Zhang, Q. Shen, X. Zhang, and B. Zhang, “A dataset of remote-sensed Forel-Ule Index for global inland waters during 2000–2018,” Sci. Data 8(1), 26–37 (2021). [CrossRef]
58. C. D. Liu, B. Y. He, M. T. Li, and X. Y. Ren, “Quantitative modeling of suspended sediment in middle Changjiang River from MODIS,” Chin. Geograph.Sc. 16(1), 79–82 (2006). [CrossRef]
59. X. W. Jiang, J. W. Tang, M. W. Zhang, R. H. Ma, and J. Ding, “Application of MODIS data in monitoring suspended sediment of Taihu Lake, China,” Chin. J. Oceanol. Limnol. 27(3), 614–620 (2009). [CrossRef]
60. P. E. Tarrant, J. A. Amacher, and S. Neuer, “Assessing the potential of Medium Resolution Imaging Spectrometer (MERIS) and Moderate-Resolution Imaging Spectroradiometer (MODIS) data for monitoring total suspended matter in small and intermediate sized lakes and reservoirs,” Water Resour. Res. 46(9), W09532 (2010). [CrossRef]
61. H. Q. Wang, C. M. Hladik, W. R. Huang, K. Milla, L. Edmiston, M. A. Harwell, and J. F. Schalles, “Detecting the spatial and temporal variability of chlorophyll-a concentration and total suspended solids in Apalachicola Bay, Florida using MODIS imagery,” International Journal of Remote Sensing 31(2), 439–453 (2010). [CrossRef]
62. Z. Y. Yin and J. S. Li, “Suspended particulate matter concentrations in Lake Taihu from 1984 to 2020 based on Landsat TM and OLI (V 1.0),” Zenodo: Version 1.0, 17 January 2022, https://doi.org/10.5281/zenodo.5864677
63. J. G. Booth, R. L. Miller, B. A. McKee, and R. A. Leathers, “Wind-induced bottom sediment resuspension in a microtidal coastal environment,” Cont. Shelf Res. 20(7), 785–806 (2000). [CrossRef]
64. M. H. Wang, S. Song, Y. L. Zhang, and W. Shi, “Remote sensing of water optical property for China's inland Lake Taihu using the SWIR atmospheric correction with 1640 and 2130 nm bands,” IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing 6(6), 2505–2516 (2013). [CrossRef]
65. L. Feng, X. J. Hou, and Y. Zheng, “Monitoring and understanding the water transparency changes of fifty large lakes on the Yangtze plain based on long-term MODIS observations,” Remote Sensing of Environment 221, 675–686 (2019). [CrossRef]
66. X. H. Liu, Y. L. Zhang, M. Z. Wang, and Y. Q. Zhou, “High-frequency optical measurements in shallow Lake Taihu, China: Determining the relationships between hydrodynamic processes and inherent optical properties,” Hydrobiologia 724(1), 187–201 (2014). [CrossRef]
67. D. Liu, H. T. Duan, S. Loiselle, C. M. Hu, G. Q. Zhang, J. L. Li, H. Yang, T. R. Thompson, Z. G. Cao, M. Shen, R. H. Ma, M. Zhang, and W. X. Han, “Observations of water transparency in China’s lakes from space,” International Journal of Applied Earth Observation and Geoinformation 92, 102187 (2020). [CrossRef]
68. L. Guo, “Doing battle with the green monster of Taihu Lake,” Science. 317(5842), 1166 (2007). [CrossRef]
69. D. R. Corbett, B. McKee, and D. C. Dan, “An evaluation of mobile mud dynamics in the Mississippi River deltaic region,” Mar. Geol. 209(1-4), 91–112 (2004). [CrossRef]
70. Y. C. Zhang, S. Lin, J. P. Liu, X. Qian, and Y. Ge, “Time-series MODIS Image-based Retrieval and Distribution Analysis of Total Suspended Matter Concentrations in Lake Taihu (China),” Int. J. Environ. Res. Public Health 7(9), 3545–3560 (2010). [CrossRef]
71. Y. H. Yang, “Decadal inundated area dynamics monitoring and analysis of three typical fresh lakes of China based on Landsat time-series satellite images,” Nanjing University.2017.
