Time Series Analysis of Water Quality Factors Enhancing Harmful Algal Blooms (HABs): A Study Integrating In-Situ and Satellite Data, Vaal Dam, South Africa

Globally, freshwater resources have been put under increasing pressure due to the rapid increase in the world population and their improvement of living standards [1]. It is indicated that an estimation of more than 83% of land surfaces that are surrounding freshwater systems have been significantly influenced by the footprint of human beings as a response to anthropogenic activities [1,2]. Contaminants from ineffective waste management, pesticides and fertilizers from agricultural areas, pollution from urban, industrial, and domestic wastewater can often be released to ground and surface water [3], and even the landfills or slag heap disposals may release pollutants seeping into nearby water resources [4,5]. The inland freshwater bodies are more vulnerable to such problems of pollution and contamination [6] resulting in the decline of quality and availability. Among the loaded pollutants, some serve as nutrients for the enhancing growth of algae. Such nutrients can be loaded from point or non-point sources, point sources, for example, wastewater treatment plants can easily be recognized and given more attention [7,8]. Non-point sources including urban areas, cultivated lands, natural forests, and pastures are more difficult to recognize and attention should be given to them according to the amounts of nutrients they can potentially release. For example, it was found that the estimated loads of nitrogen, phosphorus, and suspended sediments from urban areas and cultivated land are 10 to 100 times greater than idle and forested lands in the Great Lakes catchments of USA and Canada [9].

Eutrophication is a serious freshwater problem caused by the excessive growth of harmful algal blooms (HABs). In recent years, the factors enhancing such blooms have become interesting research subjects. Many studies have been conducted in this field, and most of them were related to the HABs to the loading of nutrients such as nitrogen and phosphorus into the water bodies [10,11].

In Africa, societies are highly dependent on their natural inland freshwater resources creating more pressure on them resulting in significant changes in their water quality [12]. South Africa has over 4000 freshwater storage dams, among which around 700 are public dams controlled and managed by the governments used for domestic, irrigation, and industrial water supply [13], they are underpinning the economic and social development of the country. Despite such a huge number of dams, South Africa is a water-stressed and scarce country facing critical challenges due to poor water management practices, inadequate infrastructure, and a relentless surge in water demands [14]. Owing to the fact that most South African dams are located downstream of metropolitan and urban areas, they have become more enriched with nutrients [4,15]. The climatic conditions of South Africa associated with nutrient loads resulted in extensive and widespread eutrophication and cyanobacterial blooms in the inland freshwater bodies [16,17]. This will continue to increase the cost of using these valuable resources. Nitrogen and phosphorus are considered the leading factors in accelerating the lakes’ and reservoirs’ eutrophication [18]. The Water Act of 1956, Section 21(1) (a) and the Department of Water Affairs in 1980 set exceptional standards of the phosphate concentration for effluent discharged to some mentioned rivers or their tributaries, it promulgated that the phosphate concentration should not be higher than 1 mg/L [14,19]. An investigation was conducted by CSIR and the Department of Water Affairs from 1985 to 1988 to predict the impact of the phosphorus standard (1 mg/L). Based on their findings, the eutrophication control aim was set to maintain the mean concentration of chlorophyll-a within receiving water bodies at the level that no conditions of severe nuisance would occur more than 20% of the time. To achieve this aim, it is required to keep the phosphorus concentration level in the reservoirs water less than 0.14 mg/L [20].

Vaal Dam is one of multiple water bodies affected by eutrophication and cyanobacterial blooms in South Africa. It was constructed by the Department of Irrigation of the national government and functionally completed in October 1936 at 20 km downstream of the confluence of the Vaal and Wilge rivers [21]. It is the second biggest dam by surface area in South Africa (about 320 Km²). The dam has undergone several rises to increase its storage capacity which ended up with a total capacity of approximately 2.603 × 10⁶ m³ [21]. The dam catchment area is approximately 38,000 km² holding various human activities including major agricultural activities (crop cultivation and cattle grazing), mining, and some industrial activities [22,23] as well as many formal and informal settlements. In many areas within the Vaal Dam catchment, mine dewatering and urban/industrial treated effluent discharges find their way to the streams causing serious water quality issues [22]. Such activities have direct or indirect effects on the dam water quality in terms of nutrient loading which enhances the growth of HABs. A study in 2013 stated that the Vaal Dam Reservoir was classified as a mesotrophic water body according to the South African Department of Water Affairs Classification System, the mean concentration of chlorophyll-a was 14.8 μg/L, the mean total phosphorus concentration was 0.077 mg/L, and the time percentage where chlorophyll-a exceeds 30 μg/L was found was 17% [24].

HABs have major effects on water quality and their aquatic system function; therefore, monitoring of their distribution in space and time is very important for water resources managers to address the issues related to it. Moreover, it is very important to address the question of which factors enhance and control the HABs in the Vaal Dam Reservoir. Historically, many studies were conducted to assess different methods to detect HABs and cyanobacteria in the Vaal Dam Reservoir, for example, an attempt was made to help the managers of the drinking water treatment facility with advanced prediction of Microcystis sp. concentration in the Vaal Dam water, by building a model using physical, chemical, and biological water quality records between 2000 and 2012. The model showed a promising result in estimating Microcystis sp. in 7 days in advance [24]. The composition of the phytoplankton in the reservoir was examined over the course of a year, it was found that Anabaena and Microcystis are the most abundant cyanobacteria that dominate the phytoplankton on the dam reservoir during summer time; during February and March, the study reviled that they are not necessary dominating throughout the year [25]. Few studies were conducted in the Vaal Dam using remote sensing data (Landsat 8 OLI and Sentinel 2 MSI), the two sensors have relatively high spatial resolution and a capability to distinguish very small differences in the optics of the water column as a function of water quality parameters such as Chl−a. Sakuno, Y et al., (2018) have obtained a strong correlation between the satellite estimation using unified red-to-near-infrared two-band and three-band models, and the in-situ measurements (see ref. [26]). Another study tested the red-to-near-infrared (NIR-red) bands by means of stepwise logistic regression (SLR) to classify Chl−a concentrations using Landsat 8 data for 2014–2016, the SLR gave overall accuracy ranged between 65 to 80% (see ref. [27]). A comparison of Landsat 8 and Sentinel 2 was conducted in mapping water quality at Vaal Dam, Sentinel 2 generated better results than Landsat 8 for estimating both chlorophyll-a and turbidity with R² = 0.86 comparing to 0.68 for Chl−a, and 0.59 and 0.50 for turbidity [26,27,28]. The latest remote sensing-based article published on the Vaal Dam was (Obaid et al., 2021) applied to high-resolution sensors that successfully used both the blue–green and red-infrared OC algorithms in estimating Chl−a concentrations [15,29].

Blue–green algorithms are typically utilized to estimate chlorophyll-a concentration in Case I waters, where water quality is predominantly influenced by phytoplankton [30,31]. However, in Case II waters, where the optical signature is influenced by various water quality constituents, employing blue–green models can result in inaccuracies when estimating chlorophyll-a concentration. The successful use of the blue–green OC algorithm in the Vaal Dam, a Case II water body, may be attributed to the strong signal of HABs arising from the high biomass concentration in the reservoir water column. While previous studies have focused on developing methods to detect HABs and cyanobacterial blooms in the reservoir, none have attempted to uncover the factors contributing to such blooms in terms of nutrient loading and environmental factors.

The data have been explored to uncover the important water quality properties. The perusal of the graphs reveals some extremely high values as well as some trends, it also reveals the seasonality of temperature and dissolved oxygen. The distribution frequency plot of chlorophyll−a (Figure 2) after excluding the very high values, shows that most of the data is centered near its lower range, below 10 μg/L. This illustrates that most of the time, the chlorophyll−a concentration in this measurement point remains below its average value. Algal blooms are spatially and temporarily varying, it might happen that an excess bloom might occur at the location of the measuring point during the time of sampling. This might explain the extremely high values of Chl−a concentration. The distribution plots of TP, NH₄_N, and KJEL_N after removing the very high values, also show that most of the data are centered relatively near their lower values in the reservoir water. Therefore, the extremely high to very high values probably indicate that some types of pollution ended up reaching the reservoir which apparently increases the capacity of the reservoir water to support high rates of biomass productivity during such high nutrient load times.

The productivity in the open water systems is a function of multi biophysical factors, such as light, temperature, nutrients, etc. In many aquatic systems, some limiting nutrients (mainly TP and N) are considered the main drivers of HABs [35,36]. However, the time series plot (Figure 3) showed that TP levels between 1990 and 2022 are relatively low except on some specific dates, a corresponding increase in Chl−a and KJEL_N concentrations at such specific dates were detected, which jumped to extremely high values. These are clearer when we look closely at the decadal time series plots in Figure 4b,c. Thus, this explains that the dam water HABs productivity is primarily driven by TP and KJEL_N when their concentrations suddenly jumped to high values. The water temperature time series in Figure 3 and Figure 4a,c shows no significant change, and its decadal trends show a slightly decreasing trend for the first decade (1990 to 2000) in Figure 5a while showing no increasing or decreasing trend during the last decade (2010–2020) in Figure 5c. The average temperatures for the first and last decades were 17.94 and 18.04 °C, respectively, which suggests a slight increase in the average water temperature between the first and last decades. This situation of increasing the decadal average of Chl−a during the study period while the temperature remains with no significant change suggests that the productivity in the Vaal Dam is not limited to the changes in the temperature during the study period, but its value ranges are ideal for algal growth.

Chl−a decadal trends (Figure 5a–c) showed a slight decrease in the first decade (1990 to 2000) and remained constant within the second decade (2000 to 2010) before it increased noticeably during the last decade (2010 to 2020). The average decadal concentrations were 4.75, 10.51, and 16.7 μg/L, respectively. Such an increasing trend of the Chl−a during the last decade alongside the increasing decadal average through the study period raises a significant warning to the situation of algal blooms in the dam. It will become a serious challenge facing the authorities to control this rising impact of HABs in the future. The significant increases in the Chl−a decadal average might be affected by the erratic extremely high values that occurred at individual dates during the second and third decades, for example, when excluding all Chl−a values above 50 μg/L, the decadal averages will decrease to 6.26 and 10.38 μg/L for the second and third decades, respectively. It still has high jumps on the decadal averages through the last three decades.

TP, KJEL_N, and NH₄_N decadal trends followed the same trend behavior of the Chl−a in the first and last decades, where they were associated with general upward and downward trends of Chl−a. The average decadal concentrations of TP were 0.1043, 0.1096, and 0.1119 for the first, second, and third decades, respectively. These concentrations show that the TP concentrations were low except on the above-mentioned individual dates where they jump to high levels greater than the hypertrophic TP threshold level which is 0.25 mg/L [15]. For KJEL_N, the decadal average concentrations were 0.8033 and 1.1417 mg/L for the first and last decade, respectively, and NH₄_N were 0.0397 and 0.0431 mg/L, respectively. However, the corresponding behavior of the Chl−a, TP, and KJEL_N decadal trends alongside the erratic high Chl−a values recorded at the same individual dates of very high TP and KJEL_N records, support that they are the driving factors of the algal blooms within the Vaal Dam.

On the other hand, the NO₃NO₂_N concentrations showed a negative trend compared to the Chl−a decadal trends during the first and last decade and DO also showed a negative trend during the last decade. The NO₃NO₂ decadal average concentrations were 0.2455 and 0.2248 mg/L, respectively, while the average concentration of DO was 7.928 mg/L. The trend of the DO is well understandable because dissolved oxygen is usually consumed during excessive algal blooms which have been known through the last few decades [37], but this situation might not be applicable to the behavior of NO₃NO₂_N trends. The analysis of the results suggests that it has no direct relation with the Chl−a trend, but a paper [38] focused on sources and forms of the most important nitrogen substrate for blooms in eutrophic Lake Erie suggested that the NO₃⁻ was the most important source of N except in late blooming stages where phytoplankton relays on recycled N derived from dissolved organic nitrogen. Their study further showed that the NO₃ˉ depletion was related to the consumption by phytoplankton during its blooms showing a negative relationship. In this study, NO₃NO₂_N also showed a negative correlation with Chl−a, but the assumption of consumption by the HABs needs more detailed investigations. The regression between the targeted WQ parameters showed a strong correlation between Chl−a and TP, Chl−a and KJEL_N in great agreement with the trend analysis. It also showed a positive correlation between Chl−a and temperature while there was no correlation between Chl−a and NH₄_N which explains that NH₄_N is not a driving factor for HAB in the Vaal Dam. Like the decadal trend plots, the regression showed a negative correlation between the Chl−a, DO and NO₃NO₂_N.

The need for more effective monitoring of the temporal and spatial distribution of HABs in the Vaal Dam Reservoir has led to the use of satellite remote sensing techniques. The satellite data gives a synoptic view of the spatial distribution and the temporal frequency of HABs. Thus, it is an effective tool for the detection, mapping, and monitoring of the blooms. The spatial distribution of HABs in the reservoir has been analyzed using Chl−a concentrations derived from satellite data. Despite the presence of only one ground station for validation, the use of remote sensing data is justified based on the robustness of the employed algorithms. The accuracy of the retrieved data have been confirmed through rigorous evaluation of these algorithms. This assessment included correlating the in-situ data point with 16 distinct satellite images obtained across varying dates, seasons, and weather conditions. From each image, a singular value corresponding to the measuring point coordinates was extracted. The resulting R² value indicates a reasonable fit, thereby affirming the reliability of the algorithms for chlorophyll retrieval in the Vaal Dam. Additionally, the observed accuracy of the retrieved data underscores the effectiveness of this approach, despite the limited number of in-situ data points utilized for evaluation. Notably, observations indicate temporal variations in Chl−a concentrations, including instances of elevated levels of Chl−a concentrations in 2015 and the following summers. Such high concentrations are strongly correlated with the in-situ measured values of Chl−a, TP, and KJEL_N during April and May 2015, and January 2016. High productivity was also noticed in images of 2019, 2021, 2022, and 2023. These high productivities may relate to the periods where TP and KJEL_N levels increase due to the increase in human activities such as the discharge of partially treated or untreated wastewater, or through runoff from urban and agricultural areas around the dam. For example, the media reported some waste found its way to the reservoir on 23 July 2015, a report mentioned that an uncontrolled sewage discharge overwhelmed the Deneysville town which is located on the Vaal Dam bank, just next to the dams wall (https://mg.co.za/article/2015-07-23-sewage-in-gautengs-drinking-water/), accessed 7 March 2023.