Lake Tana is the largest freshwater lake in Ethiopia and the source of the Blue Nile (Poppe et al., 2013). It was designated as a world biosphere reserve by UNESCO in 2015. The lake is located between 11°30′ N-12°30′ N and 37° E-37°30′ E (Figure 1). The mean depth is 9.7 m, the median depth is 10.5 m, and the maximum depth is 14.8 m (Kebedew et al., 2020b). The wave depth is nearly the same as the maximum lake depth, indicating that the lake is fully mixed. The adsorbed concentration in the bottom sediment affects the dissolved concentration in the entire water column (Kebedew et al., 2020b). The lake surface area varies between 2,945 and 3,090 km² depending on the time of the year, with the lowest water level in June and the highest at the end of the rain phase in September or October (Dersseh et al., 2020). The average water level observed is 1,786.2 m a.s.l. (Alemu et al., 2020). Historically, the lowest lake level was 1,784.2 m a.s.l. in June 2013, and the maximum lake level recorded in October 1998 was 1,787.72 m a.s.l. The lake's drainage area is around 12,000 km² (Zimale et al., 2018). The total storage volume of the lake is ~30 km³, and its morphometric shape is nearly cylindrical (Kebedew et al., 2020b).

Lake Tana has significant socio-economic importance for Sudan and Egypt downstream. More than 80% of the Nile water comes from Ethiopian highlands (Johnson and Curtis, 1994; Conway, 1997). There are four seasons, a main-rain phase from July to September; a post-rainy phase from October to November; the dry phase that extends from December to April; and a pre-rainy phase from May to June. From 1990 to 2007, the average annual rainfall on the lake was 1,230 mm (Alemu et al., 2020). More than 40 seasonal rivers and four perennial rivers flow into the lake. The main rivers are the Gilgel Abay (Little Blue Nile), Gumara, Rib, and Megech (Kebede et al., 2006). The four major rivers contribute 93% of the total flow (Kebede et al., 2006). Dessie et al. (2015) estimated that the Gilgel Abay in the southern part of Lake Tana contributed 58% of the flow, Gumara and Rib in the east 30%, and Megech in the north 10% and several small rivers in the west 2%. The lake outflow after 2010 has been directed through a tunnel to the Beles River to generate 460 MW of electricity with hydroelectric power. Excess flow leaves the lake over the Chara-Chara weir to the Blue Nile. The annual outflow of the lake varies between around 3 and 5 km³ a^–1 (Alemu et al., 2020). The lake is used for storage for hydropower generation, transportation, fishing, and recreation. It is a tourist attraction, with important religious sites on its islands.

Rainfed teff, maize, and sorghum are the main crops in the basin and cover three-fourths of the basin (Abera et al., 2021). During the dry monsoon phase, <20 % is cultivated using residual moisture (Abera et al., 2021). Nearly 20% of the basin is suitable for surface irrigation, but the water was only available for <1% of this potential area without constructing dams for surface storage (Worqlul et al., 2015). There are three major cities: Bahir Dar in south of the Lake, Debre Tabor in the upper watershed in the east, and Gonder in the north (Wondie, 2010). Most farmers grow eucalyptus trees (Tefera and Kassa, 2017). The wetlands have been severely affected by silt loads due to increased erosion from the degrading soil base due to deforestation followed by intensive agriculture (Wondie, 2010; Tebebu et al., 2017). Runoff is generated from saturation excess in the periodically saturated valley bottom and the degraded cropland with a hardpan (Moges et al., 2017b; Tebebu et al., 2017). The hydrology is complex, with many faults acting as conduits of subsurface flow and volcanic dikes as barriers (Adem et al., 2020).

Total phosphorus (TP) is the sum of all forms of phosphorus, including ortho-phosphates (soluble reactive phosphorus) and the phosphorus in plant and animal fragments suspended in lake water (Correll, 1998). Total nitrogen (TN) is the sum of nitrate, nitrite, ammonia, and organic nitrogen (Bremner, 1965; McGill and Figueiredo, 1993). Chlorophyll-a (Chl-a) is an indicator of water quality linked to productivity associated with nutrient levels in the lake (Koponen et al., 2001). The abundance of phytoplankton in freshwater lakes is strongly associated with TP (Guildford and Hecky, 2000).

River inlets and lakeshore areas were sampled to determine the concentration. The river inlet concentrations were used to calculate the nitrogen (TN) and phosphorus (TP) load delivered by the four major rivers. Water samples were collected at each of the four major rivers (Figure 1) in June, July, August, September, and December 2019. The lake depth at the river inlets was between 3 and 11 m. The other samples along the shores were taken to characterize the relationship between water hyacinth distribution and the nitrogen, phosphorus, and chlorophyll-a concentrations in the nearshore area of Lake Tana. Nearshore samples for determining TP, TN, and Chl-a were collected at 27 stations roughly divided evenly between water hyacinth-infested and non-infested regions (Figure 1). The samples were collected in two campaigns. For the first campaign, samples were taken at 0.5 m depth in August 2016, December 2016, and March 2017. In the second campaign, water samples were collected in June, July, August, September, December 2019, and March 2020 at the 0.5 m surface and just above the lake bottom. The sampling dates were chosen to represent the main rainy season (July-September), the dry season (December–April), and a pre-rainy season (May–June) to understand how seasonality influences the water quality parameters.

The water samples in all rounds were collected using a water sampler (Kemmerer—Model 12.07—Water Sampler Set and Beta Water Samplers, Horizontal Acrylic—Water sampler only, transparent acrylic, 3.2 L). The water samples were filtered and transferred to polyethylene bottles, 100 ml for TP, 100 ml for TN, and 500 ml for Chl-a. Samples were stored at −4°C until analysis by the Amhara Design and Supervision Works Enterprise.

Annual inflow and outflow data for Lake Tana were taken from Alemu et al. (2020) study for a period with above and below-average rainfall. The proportion of mean annual flow rate from North, South, West and East side were calculated based on the measurements made by Dessie et al. (2017) and used in the nutrient loading calculation and balance.

An approximate P and N mass balance for the Lake was calculated by summing the products of discharges and measured concentrations near the four major rivers and subtracting that flowing through the outlets. The external phosphorus loading could not be estimated by the published functions in the literature by Larsen and Mercier (1976), Vollenweider (1976), Prairie (1989), and Hejzlar et al. (2006) because the lake was fully mixed, contrary to the assumptions on which those equations are based.

The average total phosphorus, TP, concentrations for the Rib, Gumara, and Gilgel Abay inlets was just under 2 mg/l in 2019 (Table 1). The TP concentration in the Megech that originated in the Simien mountains with little agriculture was half of the other three rivers (Table 1). The average total nitrogen, TN, concentrations of the three rivers Gumara, Rib and Megech, with a large flood plain, were just above 2 mg/l, and for the Gilgel Abay, 3.5 mg/l (Table 1). The difference was statistically not significant. The Gilgel Abay has only a very small flood plain, and thus the opportunity for the nitrate to denitrify is smaller. The maximum concentrations of TP and TN in the four rivers were observed in December. The minimum was in June (Table 1). Wondie et al. (2007) found a similar trend for nitrate but not for phosphates in Lake Tana in 2003 and 2004.

Water samples at the surface and the bottom were taken monthly from June to December in 2019 and March 2020. The average TP concentration in water samples at the lake bottom was, on average, 20% greater than in samples taken at 0.5 m from the surface (Table 2). Four-fifths of the water samples near the lake bottom exceeded the TP taken at the surface. Average TN was 16% lower at the bottom than at the surface, mainly because of the low concentration in the lake bottom samples for the sampling station near Bahir Dar (Stations 22–24, Figure 1; Table 2). Chl-a at the surface was greater than the bottom for two-thirds of the sampling locations (Table 2). Samples near the surface contain more Chl-a than the bottom samples following the pattern of TN. The small differences in surface and bottom concentration confirm the finding of Kebedew et al. (2020b) that the water column is mixed due to its shallow depth.

The water quality parameters of 27 stations (location shown in Figure 1) between 2016 and 2020 are given in Table 3. Mean TP concentrations increased about 5–10-fold over the 4 years. It increased from an average of 0.2 mg/l in 2016 to 1.8 mg/l in 2020, Table 3. Over the same period, TN concentrations in 2016 and 2020 were nearly the same, around 2 mg/l. The TN:TP ratio ranged from 0.4 to 29.5 (Table 3). The mean Chl-a concentrations did not show a clear trend. Average Chlorophyll-a concentrations ranged from 1 to 22 μg/l in 2016 and 2017. In 2019–2020 concentrations were between 4 and 9 μg/l (Table 3). Wondie et al. (2007) found the mean chlorophyll-a concentrations varied from 2.6 to 8.5 μg/l in 2003 and 2004.

The spatial distribution of water hyacinth and other aquatic vegetation is shown in Figure 1. The total area with water hyacinths was around 2,500 ha, and most was found on the eastern shore north of the Rib outlet (Figure 1). Some water hyacinths locations showed elevated TP concentrations. TP concentrations for samples collected in 2019 and 2020 of >1 mg/l were recorded for stations 4, 8, 10, 12, 13 on the eastern side with the main water hyacinth infestation, station 18 near the north shore with some water hyacinths, and stations 25 and 27 near the western shore (Figure 1). The TN/TP ratio was also less for the high P sites. Stations 14, 16, 17, 19 near the Megech and 21 and 22 near the Blue Nile outlet had an average TP concentration of <0.5 mg/l (Figure 1; Table 4). Except for station 14, water hyacinths were not present at these sites. The calculated spatial weight-based TN/TP ratios ranged from 1.3 to 6.4 (Table 4).

Total N concentrations over 2 mg/l were observed at stations 5, 7, 10, and 11 near the eastern shore, station 17 in the north, station 21 at the Chara-Chara weir near the outlet to the Blue Nile and 26 on the west shore. The lowest TN concentrations were observed at Stations 1–4 south of the Gumara (Table 4).

The maximum Chl-a concentrations were seen in parts of the lake that were not or minimally infested by water hyacinths (Stations 4, 5, 6 17 24, 25, and 26, Figure 1; Table 4). The minimum Chl-a concentrations in the lake were most severely infested by water hyacinths (Stations 8, 13, and 20, Figure 1; Table 4).

The TP concentrations in the Lake near Gumara, Rib and Gilgel Abay of 1.7 mg P/l (Table 4) were 30 times the lake water quality standard of the United States Environmental Protection Agency recommended level in streams entering lakes or reservoirs of 0.05 mg·P l⁻¹ (USEPA US Environmental Protection Agency, 1986, 2000) and was 10 times the European standard (Berzina and Sudars, 2010). The average TP lake concentrations were 1.1 mg P/l in 2019 and an order of magnitude greater than the USEPA standard. The average total nitrogen concentration near the rivers was 2.6 mg N/l which is 45% greater than the lake of 1.8 N mg/l in 2019 (Table 2) and approximately three times that USEPA US Environmental Protection Agency (2000) recommended level. The chlorophyll-a concentration of around 6 μg l⁻¹ in 2019 (Table 4) were about three times less than the concentration in shallow areas with a depth <20 m in Lake Victoria in 2005–2008 (Sitoki et al., 2010).

Annual inflow and outflow data were taken from Alemu et al. (2020) study for a period with above and below-average rainfall. To understand the future trend of the nutrient concentration, we calculated the annual incoming nutrient load from the rivers in Lake Tana and the load leaving either through the Tana Beles tunnel or the Blue Nile outlet. We approximated the concentrations in the rivers as the average annual concentration in 2019 presented in Table 1. The proportion of the annual flow originating in the south, west, north and east sides of the lake was based on the measurements made by Dessie et al. (2017). Finally, the annual lake elevation patterns were assumed the same so that the elevation changes in the lake could be neglected.

The load of TP and TN for the sides was calculated as the product of the observed concentration and the annual river flow (Table 5). The greatest TN and TP load was contributed to the lake in the south, where the Gilgel Abay entered the lake. The Gilgel Abay has the largest drainage basin and the highest rainfall over 2 m per year in the headwaters. The eastern side where the Rib and the Gumara enter the lake had the next highest load. It was followed by the Megech River in the north and a few minor streams in the west (Table 5). To calculate the load leaving the lake, we multiplied the discharge from the lake by the average concentration. We find an annual excess of ~5.5 Gg TN and TP (Table 5).

Since the average volume of the lake is 30 km³, the excess load is equivalent to an increase in the concentration of 0.18 mg P/l/a. Over the 3.5 years of the study, we would expect an increase of 0.6 mg P/l. As shown in Table 3, the concentration difference between August 2016 and September 2019 is 0.8 P/mg/l, which is in the same order as the load calculations in Table 5. However, TP concentrations increased steeply in December 2019 and March 2020 to 1.8 mg P/l. The load calculation underestimated this rapid rise. One of the reasons that the P loads were underestimated over the 3.5 years was because the samples for concentration determination were taken in the lake near the mouth of the rivers, and lake water might have mixed with river water resulting in lower concentrations. Another reason could be the arithmetic averages calculation of the concentrations. The highest flows have the greatest concentrations, as the Moges et al. (2016) data showed for the Awramba watershed.

Since TP and TN loads were approximately equal (Table 5), we expected, based on the load calculations, to see a similar increase in TN concentrations of 0.18 mg N/l/a. The expected increase in nitrogen concentrations due to the surplus of nitrogen added did not materialize as the concentration of TN during the last 4 years remained nearly the same (Table 4; Figure 3).

The difference between P and N accumulation in the lake can be explained by considering the difference in sinks. The phosphorus concentration in the water column is in equilibrium with the bottom sediment (Table 2) because the lake is shallow and fully mixed (Kebedew et al., 2020b). The incoming sediment from the rivers that have settled on the lake bottom will mix with the bottom sediment with a lower P concentration deposited earlier, lowering the P concentration in the water column. Therefore, the lake's increase in concentration is delayed compared to that in the rivers. This trend is visible in Figure 2, where the TDP concentrations around the same date during the rain phase were greater in the river than in the lake. The lake P concentration catches up with the river after 3–4 years. Thus, unlike deep lakes, where sediment deposition takes P out of the water column, there is no sink term for P in lake Tana because of its shallow depth. Consequently, the P concentration in the lake can only decrease when the load in the outflow is greater than the load from the rivers.

The lake nitrogen concentration can be lowered by denitrification of nitrate to nitrogen gasses, plant uptake, and outflow (Abrahamsson and Håkanson, 1998; Ayele and Atlabachew, 2021) in the lake and surrounding wetlands. Abrahamsson and Håkanson (1998) show that denitrification is essential in nitrogen assimilation in the lake. The assimilation capacity of the Lake has been changing over the 17 years that data were collected. In 2003 and 2004, denitrification brought down the nitrogen concentrations to near zero at the end of the dry phase (Figure 3). In March 2010, the nitrate concentration was 0.25 mg N/l (Goshu and Aynalem, 2017), indicating that not all sources (i.e., fixing atmospheric nitrogen loads) delivered by the rivers could not be assimilated. Therefore, it does not seem accidental that in 2010, when the lake and its wetlands could not fully assimilate the nitrogen, the first water hyacinths were noted in Lake Tana.

We found that the greatest infestation of water hyacinths occurred in the northeastern part of the Lake. To investigate whether the location of water hyacinth was related to TN, TP concentrations or the TN/TP ratio, we plotted the average values of these parameters over the measurement period (shown in Table 4) counterclockwise starting at the Gumara in Figure 5. There is not a unique relationship between any of the variables. It could be argued that the phosphorus concentrations are greater in the water hyacinth-infested area. But at the same time, phosphorus concentrations are just high without water hyacinth on the western shore (sampling locations 27, 26, 25) and location 24 on the southern shore near the flower greenhouse. The nitrate concentrations and the TN/TP ratios are not different in the water hyacinth-infested areas than elsewhere.

Water hyacinth is a floating plant that will move with the water currents and be pushed by the wind. With a TN/TP ratio below the critical value of 7 and TN concentrations >1 mg N/l, water hyacinths can potentially emerge anywhere in Lake Tana (Center and Spencer, 1981). Thus, it should not be limited to the northeast. Kebedew et al. (2020b) show that the water of the Gumura and the Rib moves to the north and then to the west, where it will eventually evaporate. The flow from the Gilgel Abay flows directly to the outlet of the Tana Beles hydroelectric power plant or when the lake is at its maximum level to the Blue Nile inlet. The wind over the lake is predominantly southeast. Thus, water hyacinths emerging in the northern paper part of the lake will eventually end up in the northeast. It will not flow westward with the lake current due to the wind. Water hyacinth in the southern part of the lake will likely leave through one of the outlets. Thus, the elevated phosphorus concentration in the water hyacinths area might be due to dying off the water hyacinths and not the cause of the water hyacinths in these locations.

Finally, to investigate the relationship between Chl-a and the remaining nutrients, we plotted the average Chl-a concentrations as a function of the ratio of TN and TP (all listed in Table 4) for the 17 sampling points that were located in areas without water hyacinth and the 10 with water hyacinths. Figure 6 shows that Chl-a concentrations in water hyacinth-infested areas were generally lower than elsewhere. The lower concentration is a consequence of water hyacinth shading the water. Consequently, the algae that need sunlight for growth do not multiply. In the non-shaded areas, the CHL-a concentrations depended on the nutrient concentration. Only in the locations shaded by the water hyacinth (Figure 5) do the Chl-a concentrations decrease linearly with the TN and TP ratio. Wondie et al. (2007) showed the algae group of diatoms was most abundant in 2003, but currently, the blue-green algae (in the non-shaded areas) could dominate, which needs future research.

Water quality has decreased slowly in the once pristine lake (Cheesman, 1936). Water hyacinth appeared first in the lake around 2010 when the assimilation capacity of the lake and wetland for nitrogen was exceeded, and nitrate was available for plant uptake year around. Currently, the area coverage of water hyacinth is 1% of the lake area, which is manageable. However, measures should be taken to prevent further expansion.

To consider what management practices are effective, we note that the N and P ratio is below the Redfield ratio of 6–8 (Figure 4), and the lake is nitrogen-limited. Thus, water hyacinths growth depends on the nitrogen concentration that, according to our findings (Figure 3; Table 3), increases slightly throughout the measurement period but has not increased significantly in the last few years. Therefore, to manage the water hyacinth expansion, the nitrogen concentration in the lake needs to be controlled, which is problematic because blue-green algae will fix atmospheric nitrogen when the N and P ratio is <6–8, as currently is the case. When the lake was free of water hyacinth in 2003, the N and P ratio was greater than during the rain and post-rain phase, diatoms instead of blue-green algae were dominant, and nitrogen was not fixed from the air (Wondie et al., 2007). Thus, to prevent nitrogen fixation, the growth of blue-green algae, both P and N concentrations in the lake need to be reduced and the P and N ratio increased (Bulgakov and Levich, 1999).

Several practices can decrease the N and P concentrations in the lake. They consist of reducing nitrogen and phosphorus input by the rivers, removing water hyacinths from the lake, and enhancing denitrification in case of nitrogen (Mucheye et al., 2018; Worku and Sahile, 2018). Since phosphorus cannot be removed from the lake other than with the water flowing out, phosphorus concentration can only be reduced by decreasing the P load.

Our results indicate that the largest source of nutrients is the fertilizers applied on the cropland, as large flows that originate from the croplands have the greatest P and N concentrations and contribute the greatest loads. Leaching of nutrients from the soil cannot be prevented with the high rainfall amounts. Therefore, fertilizers should be applied ideally in small quantities throughout the rainy period at a time so that crops need to reduce the concentrations. Point source pollution and animal manure (Goshu et al., 2020) result in a steady increase during the whole year. It might also be occurring and can become a greater source in the future.

To evaluate the effectiveness of water hyacinth harvesting on improving the water quality in Lake Tama, we calculated the nitrogen and phosphorus mass harvested from the 25 km² water hyacinth infestation. Tewabe et al. (2017) found that 82 Mg/ha of fresh weight can be harvested during the dry phase and 270 Mg/ha during the rainy phase for a total of 352 Mg/ha/year in Lake Tana. The dry mass is ~5% of the fresh weight. Su et al. (2018) found that the P content was 0.5% and N was 2.8% on a dry weight basis. Based on these numbers, 88 kg P and 492 kg N can be removed per hectare per year. For an area of 25 km² of water hyacinth, 0.22 Gg P/year and 1.2 Gg N/year can be harvested. It is <5% of the excess P added to the lake and around 20% of the N added (Table 5).

Finally, nitrate can be reduced by enhancing denitrification. Lake water levels perhaps could be managed to optimize denitrification. To do this will require more research. Moreover, wetlands within the watershed should be preserved to denitrify the nitrates before entering Lake Tana. Thus, it is not too late to prevent the expansion of water hyacinths by carefully controlling the nitrogen and phosphorus sources for both point and non-point sources. Finding the most effective measures will require more research, and implementation might be politically difficult. However, if no measures are taken, Lake Tana's water quality will decrease further due to the imbalance of nutrient input and that taken out of the Lake.