Hyperspectral determination of eutrophication for a water supply source via genetic algorithm-partial least squares (GA-PLS) modeling

A novel chlorophyll-a retrieval model based on suspended particulate matter classification and different machine learning

Chlorophyll-a (Chla) in inland waters is one of the most significant optical parameters of aquatic ecosystem assessment, and long-term and daily Chla concentration monitoring has the potential to facilitate in early warning of algal blooms. MOD09 products have multiple observation advantages (higher temporal, spatial resolution and signal-to-noise ratio), and play an extremely important role in the remote sensing of water color. For developing a high accuracy machine learning model of remotely estimating Chla concentration in inland waters based on MOD09 products, this study proposed an assumption that the accuracy of Chla concentration retrieval will be improved after classifying water bodies into three groups by suspended particulate matter (SPM) concentration. A total of 10 commonly used machine learning models were compared and evaluated in this study, including random forest regressor (RFR), deep neural networks (DNN), extreme gradient boosting (XGBoost), and convolutional neural network (CNN). Altogether, 41 basic bands and 820 band ratios between the 41 bands were filtered by measuring their correlation with Ln(Chla) and several bands brought into different machine learning models. Results demonstrated that the construction of Chla concentration remote estimation model based on SPM classification could significantly improve the correlation between Ln(Chla) and 41 basic spectral band combinations, the correlation between Ln(Chla) and 820 band ratios, and the model verification R2 from 0.41 to 0.83. Furthermore, B3, B20, and B32 were finally selected based on correlation with SPM to classify SPM and the classification accuracy could reach 0.9. Finally, we concluded that RFR model performed best via comparing the R2, RMSE, and MAPE. By comparing the relative contribution of input bands in different groups, B3 contributed most to three groups. The model constructed in this study has promising prospects for promotion and application in other inland waters, and could provide systematic research reference for subsequent research.

Winter Water Quality Modeling in Xiong'an New Area Supported by Hyperspectral Observation

Xiong'an New Area is defined as the future city of China, and the regulation of water resources is an important part of the scientific development of the city. Baiyang Lake, the main supplying water for the city, is selected as the study area, and the water quality extraction of four typical river sections is taken as the research objective. The GaiaSky-mini2-VN hyperspectral imaging system was executed on the UAV to obtain the river hyperspectral data for four winter periods. Synchronously, water samples of COD, PI, AN, TP, and TN were collected on the ground, and the in situ data under the same coordinate were obtained. A total of 2 algorithms of band difference and band ratio are established, and the relatively optimal model is obtained based on 18 spectral transformations. The conclusion of the strength of water quality parameters' content along the four regions is obtained. This study revealed four types of river self-purification, namely, uniform type, enhanced type, jitter type, and weakened type, which provided the scientific basis for water source traceability evaluation, water pollution source area analysis, and water environment comprehensive treatment.

Monitoring Optical Variability in Complex Inland Waters Using Satellite Remote Sensing Data

Optical classification for water bodies was carried out based on satellite remote sensing data, which avoided the limitation of having a limited amount of in situ measured spectral data. Unsupervised cluster analysis was performed on 53,815 reflectance spectra extracted at 500-m intervals based on the same season or quasi-same season Landsat 8 SR data using the algorithm of fuzzy c-means. Lakes and reservoirs in the study area were comprehensively identified as three optical types representing different limnological features. The shape and amplitude characteristics of the reflectance spectra for the three optical water types indicated that one corresponds to the clearest water, one corresponds to turbid water, and the other is moderate clear water. The novelty detection technique was further used to label the match-ups of the in situ data set collected during 2006 to 2019 in 12 field surveys based on mathematical rules of the three optical water types. The results confirmed that each optical water type was associated with different bio-optical properties, and the total suspended matter of the clearest, moderate clear and turbid water types were 14.99 mg/L, 41.06 mg/L and 83.81 mg/L, respectively. Overall, the clearest, moderate clear and turbid waters in the study area accounted for 49.3%, 36.7% and 14.0%, respectively. The spatial distribution of optical water types in the study area was seamlessly mapped. Results showed that the bio-optical conditions of the water distributed across the southeast region were roughly homogeneous, but in most of other regions and within some water bodies, they showed a patchy distribution and heterogeneity. This study is useful for monitoring water quality and provides a useful foundation to develop or tuning algorithms to retrieve water quality parameters.

Variables' classification via equivalence relations for the trophic state of a Mediterranean ecosystem

The trophic state of an aquatic body is influenced by many biotic and abiotic factors. When lots of parameters affect a phenomenon, such as eutrophication, it is difficult to distinguish which are the ones that affect the ecosystem the most. In this paper, we propose an alternative way for data analysis, in order to avoid complex systems with many variables. For the examined Mediterranean shallow lake, the studied parameters are water temperature (°C), ammonia (NH₄ -N) (mg/L), dissolved oxygen (mg/L), turbidity (NTU), pH, conductivity (mS/cm), and chlorophyll-a (μg/L). We formed groups with the variables above based on fuzzy equivalence relations and from each group we chose the parameter influence the studied phenomenon the most. Numerical results of fuzzy linear regression showed strong agreement with the proposed method above and pH, NH₄ -N, and dissolved oxygen are the variables influence this ecosystem more than the others. PRACTITIONER POINTS: When having many parameters in a studied ecosystem, we propose a way we can distinguish the most representative ones, the parameters that influence more the phenomenon we study each time. Formation of groups in variables can be applied to many case studies in order to have a clear idea of our data and the relevance of each of them in our dependent one. Fuzzy linear regression can be used in order to check the final results and ensure that the equivalence relations are a such a good method used in the data analysis while the researchers save time in long procedures of analyzing parameters not very close involved to the phenomenon investigated every time.

Evaluation of Chlorophyll-a Estimation Approaches Using Iterative Stepwise Elimination Partial Least Squares (ISE-PLS) Regression and Several Traditional Algorithms from Field Hyperspectral Measurements in the Seto Inland Sea, Japan

Harmful algal blooms (HABs) occur frequently in the Seto Inland Sea, bringing significant economic and environmental losses for the area, which is well known as one of the world's most productive fisheries. Our objective was to develop a quantitative model using in situ hyperspectral measurements in the Seto Inland Sea to estimate chlorophyll a (Chl-a) concentration, which is a significant parameter for detecting HABs. We obtained spectra and Chl-a data at six stations from 12 ship-based surveys between December 2015 and September 2017. In this study, we used an iterative stepwise elimination partial least squares (ISE-PLS) regression method along with several empirical and semi-analytical methods such as ocean chlorophyll, three-band model, and two-band model algorithms to retrieve Chl-a. Our results showed that ISE-PLS using both the water-leaving reflectance (RL) and the first derivative reflectance (FDR) had a better predictive ability with higher coefficient of determination (R²), lower root mean squared error (RMSE), and higher residual predictive deviation (RPD) values (R² = 0.77, RMSE = 1.47 and RPD = 2.1 for RL; R² = 0.78, RMSE = 1.45 and RPD = 2.13 for FDR). However, in this study the ocean chlorophyll (OC) algorithms had poor predictive ability and the three-band and two-band model algorithms did not perform well in areas with lower Chl-a concentrations. These results support ISE-PLS as a potential coastal water quality assessment method using hyperspectral measurements.

Quantification of dissolved organic carbon (DOC) storage in lakes and reservoirs of mainland China

As a major fraction of carbon in inland waters, dissolved organic carbon (DOC) plays a crucial role in carbon cycling on a global scale. However, the quantity of DOC stored in lakes and reservoirs was not clear to date. In an attempt to examine the factors that determine the DOC storage in lakes and reservoirs across China, we assembled a large database (measured 367 lakes, and meta-analyzed 102 lakes from five limnetic regions; measured 144 reservoirs, and meta-analyzed 272 reservoirs from 31 provincial units) of DOC concentrations and water storages for lakes and reservoirs that are used to determine DOC storage in static inland waters. We found that DOC concentrations in saline waters (Mean/median ± S.D: 50.5/30.0 ± 55.97 mg/L) are much higher than those in fresh waters (8.1/5.9 ± 6.8 mg/L), while lake DOC concentrations (25.9/11.5 ± 42.04 mg/L) are much higher than those in reservoirs (5.0/3.8 ± 4.5 mg/L). In terms of lake water volume and DOC storage, the Tibet-Qinghai lake region has the largest water volume (552.8 km³), 92% of which is saline waters, thus the largest DOC (13.39 Tg) is stored in these alpine lake region; followed by the Mengxin lake region, having a water volume of 99.4 km³ in which 1.75 Tg DOC was stored. Compared to Mengxin lake region, almost the same amount of water was stored in East China lake region (91.9 km³), however, much less DOC was stored in this region (0.43 Tg) due to the lower DOC concentration (Ave: 3.45 ± 2.68 mg/L). According to our investigation, Yungui and Northeast lake regions had water storages of 32.14 km³ and 19.44 km³ respectively, but relatively less DOC was stored in Yungui (0.13 Tg) than in Northeast lake region (0.19 Tg). Due to low DOC concentration in reservoirs, especially these large reservoirs having lower DOC concentration (V > 1.0 km³: 2.31 ± 1.48 mg/L), only 1.54 Tg was stored in a 485.1 km³ volume of water contained in reservoirs across the entire country.

Spatiotemporal dynamics of chlorophyll-a in a large reservoir as derived from Landsat 8 OLI data: understanding its driving and restrictive factors

Chlorophyll-a (Chla) is an important indicator of water quality and eutrophication status. Monitoring Chla concentration (C _Chla ) and understanding the interactions between C _Chla and related environmental factors (hydrological and meteorological conditions, nutrients enrichment, etc.) are necessary for assessing and managing water quality and eutrophication. An acceptable Landsat 8 OLI-based empirical algorithm for C _Chla has been developed and validated, with a mean absolute percentage error of 14.05% and a root mean square error of 1.10 μg L^-1. A time series of remotely estimated C _Chla was developed from 2013 to 2015 and examined the relationship of C _Chla to inflow rate, rainfall, temperature, and sunshine duration. Spatially, C _Chla values in the riverine zone were higher than in the transition and lacustrine zones. Temporally, mean C _Chla value were ranked as spring > summer > autumn > winter. A significant positive correlation [Pearson correlation coefficient (r) = 0.88, p < 0.001] was observed between the inflow rate and mean C _Chla in the northwest segment of the Xin'anjiang Reservoir. However, no significant relation was observed between mean C _Chla and meteorological conditions. Mean (± standard deviation) value for the ratio of total nitrogen concentration to total phosphorus concentration in our in situ dataset is 75.75 ± 55.72. This result supports that phosphorus is the restrictive factor to algal growth in Xin'anjiang Reservoir. In addition, the response of nutrients to Chla has spatial variabilities. Current results show the potential of Landsat 8 OLI data for estimating Chla in slight turbid reservoir and indicate that external pollution loading is an important driving force for the Chla spatiotemporal variability.

Monitoring spatiotemporal variations in nutrients in a large drinking water reservoir and their relationships with hydrological and meteorological conditions based on Landsat 8 imagery

Nutrient enrichment is a major cause of water eutrophication, and variations in nutrient enrichment are influenced by environmental changes and anthropogenic activities. Accurately estimating nutrient concentrations and understanding their relationships with environmental factors are vital to develop nutrient management strategies to mitigate eutrophication. Landsat 8 Operational Land Imager (OLI) data is used to estimate nutrient concentrations and analyze their responses to hydrological and meteorological conditions. Two well-accepted empirical models are developed and validated to estimate the total nitrogen (TN) and total phosphorus (TP) concentrations (C_TN and C_TP) in the Xin'anjiang Reservoir using Landsat 8 OLI data from 2013 to 2016. Spatially, C_TN decreased from the transition zone to the riverine zone and the lacustrine zone. On the other hand, C_TP decreased from the riverine zone to the transition zone and the lacustrine zone. Temporally, C_TN displayed elevated values during the late fall and winter and had lower values during the summer and early fall, whereas C_TP was higher during the spring and lower during the winter. Among the environmental factors, the rainfall and the inflow rate have strong positive correlations with the nutrient concentrations. TN is more sensitive to meteorological factors (wind speed, temperature, sunshine duration), and the spatial driving forces vary among the different sections of the reservoir. However, TP is more easily influenced by human activities, such as fishery and agricultural activities. Current results would improve our understanding of the drivers of nutrients spatiotemporal variability and the approach in this study can be applicable to other similar reservoir to develop related strategies to mitigate eutrophication.

Remote estimation of phycocyanin (PC) for inland waters coupled with YSI PC fluorescence probe

Nuisance cyanobacterial blooms degrade water resources through accelerated eutrophication, odor generation, and production of toxins that cause adverse effects on human health. Quick and effective methods for detecting cyanobacterial abundance in drinking water supplies are urgently needed to compliment conventional laboratory methods, which are costly and time consuming. Hyperspectral remote sensing can be an effective approach for rapid assessment of cyanobacterial blooms. Samples (n=250) were collected from five drinking water sources in central Indiana (CIN), USA, and South Australia (SA), which experience nuisance cyanobacterial blooms. In situ hyperspectral data were used to develop models by relating spectral signal with handheld fluorescence probe (YSI 6600 XLM-SV) measured phycocyanin (PC in cell/ml), a proxy pigment unique for indicating the presence of cyanobacteria. Three-band model (TBM), which is effective for chlorophyll-a estimates, was tuned to quantify cyanobacteria coupled with the PC probe measured cyanobacteria. As a comparison, two band model proposed by Simis et al. (Limnol Oceanogr, 50(11): 237-245, 2005; denoted as SM05) was paralleled to evaluate TBM model performance. Our observation revealed a high correlation between measured and estimated PC for SA dataset (R (2) =0.96; range: 534-20,200 cell/ml) and CIN dataset (R (2) =0.88; range: 1,300-44,500 cell/ml). The potential of this modeling approach for imagery data were assessed by simulated ESA/Centinel3/OLCI spectra, which also resulted in satisfactory performance with the TBM for both SA dataset (RMSE % =26.12) and CIN dataset (RMSE % =34.49). Close relationship between probe-measured PC and laboratory measured cyanobacteria biovolume was observed (R (2) =0.93, p<0.0001) for the CIN dataset, indicating a stable performance for PC probe. Based on our observation, field spectroscopic measurement coupled with PC probe measurements can provide quantitative cyanobacterial bloom information from both relatively static and flowing inland waters. Hence, it has promising implications for water resource managers to obtain information for early warning detection of cyanobacterial blooms through the close association between probe measured PC values and cyanobacterial biovolume via remote sensing modeling.