Spatiotemporal patterns of water hyacinth dynamics as a response to seasonal climate variability in Lake Tana, Ethiopia

Lake Tana, which is the largest Lake in Ethiopia, has been invaded by water hyacinths since 2011. Although the government and the community have devoted considerable time and energy over a long period to removing the invasive weed mechanically and manually, the weed has been increasing significantly. Accurate, reliable, and timely information on the spatiotemporal distribution and extent of water hyacinth is crucial to determine its evolution, propagation, and potentially vulnerable areas of the Lake. Therefore, comprehensive information on the spatial distribution of water hyacinths and their annual and seasonal variability is essential for Lake Tana’s water resource planning, development, and management. This study aims to evaluate the spatiotemporal pattern of water hyacinth and its dynamics with seasonal climate variability and impact on evapotranspiration. Landsat 7 ETM+, Landsat 8 OLI, and Sentinel 2 and meteorological datasets were employed. Supervised and manual digitization image classification methods were applied to prepare Land-use/ Land-cover in the Lake. The Mann–Kendall trend test and Pearson correlation coefficient were used to evaluate the trend of water hyacinth and the impact of climate variability on water hyacinth distribution respectively. Besides, the evapotranspiration and water losses were estimated using the FAO-56 Penman–Monteith method. The surface extent of the water hyacinth in Lake Tana has increased by 96% in 2019 from 2011. However, the surface area of the Lake has declined. That means 1603 ha of water surface area has been changed to land surface from 2011 to 2019. The average volume of water loss in Lake Tana was 0.21% of the volume of the Lake from September 2016 to December 2018.


Introduction
Water hyacinth is one of the worst weeds in the world (Holm et al. 1977). This is becoming a major challenge for water bodies in more than 50 countries of the world (Patel 2012). The major problems of the weeds are destroying fisheries and agricultural production, blocking the canals of hydroelectric power plants, a hindrance to water transport and rivers, creating environmental imbalance, causing human health problems, losing water by increasing the rate of evapotranspiration, reducing biodiversity, as well as affecting irrigation, navigation, and watering of livestock (Navarro and Phiri 2000).
In Ethiopia, water hyacinth was officially reported in Aba Samuel Reservoir on the outskirts of Addis Ababa in the 1950s (Stroud 1994). The weed was announced in Lake Koka andAwash River in 1956 (UNEP 2013;Samuel and Netsanet 2014). Lake Tana was officially recognized as infested with water hyacinth in September 2011 (Tewabe 1 3 170 Page 2 of 16 2015). Although the government and the community have been devoting considerable time to removing the invasive weeds mechanically and manually over a long period, the weed has been increasing significantly (Wassie et al. 2014(Wassie et al. , 2015Dersseh et al. 2020).
The main factors that increase the spread of water hyacinth in water bodies are climatic conditions, water body conditions, eutrophication, pH, and reproductive system (Dersseh et al. 2019). The most favorable conditions are standing water, shallow water depth (< 6 m), surface sediments rich in organic matter, and nutrients such as nitrogen and phosphorous (Dersseh et al. 2019). Climate change has profound effects on the distribution and location of water hyacinths in water bodies (Center et al. 2002). Rainfall and floods that swept agricultural runoff and nutrient-rich sediment to the lake could cause stimulate a new outbreak of water hyacinth (Fusilli et al. 2013).
Accurate, reliable, and timely information on the spatiotemporal distribution and extent of water hyacinth is crucial to understand its evolution, propagation, and potentially vulnerable areas of the Lake (Thamaga and Dube 2018b). The rate and direction of evolution and spread of invasive water hyacinths require a detailed understanding of the impacts of climate and other factors (Thamaga and Dube 2018a). Single-date weed information cannot provide a complete understanding of the spatiotemporal variability of water hyacinth. Therefore, comprehensive information on the spatial distribution of water hyacinths and their annual and seasonal variability is vital for water resource management (Molinos et al. 2015). For this reason, continuous observation and monitoring of the proliferation of aquatic weeds are essential for proper water resource management, the development of appropriate weed control strategies, and prioritization of the most infested areas (Albright et al. 2004).
Previous studies conducted in Lake Tana concerning spatial coverage of water hyacinths were dealt with in a single time even though Dersseh et al. (2020) tried to show the annual maximum and minimum area of water hyacinths and lake surface areas without the summer seasons. The longterm annual, monthly, and seasonal spatiotemporal distribution and extent of the weed and surface area variability of the Lake have not been well studied. Besides, the impact of climate variability on water hyacinth proliferation, and the implication of water hyacinth on evapotranspiration and water losses in Lake Tana have not been addressed by the previous studies. Moreover, previously reported weed area cover in Lake Tana is a source of conflict among the stakeholders (Dersseh et al. 2019). In this context, the 50,000 and 2279.4 ha areas covered by water hyacinth assessed by prior investigations in 2014 and 2018 October, respectively, show a substantial discrepancy. Therefore, detailed research is needed to examine the conflicting results of previous studies and to better understand the annual and seasonal spatiotemporal dynamics of water hyacinth and the water loss of Lake Tana. The objectives of this study were: (1) to assess the monthly, seasonal, and annual spatiotemporal distribution of water hyacinths in Lake Tana (2) to estimate evapotranspiration and water loss due to water hyacinths and (3) to investigate the association of water hyacinth area with climatic factors and Lake area. The information obtained from the results of the study is invaluable to the different stakeholders to understand the evolution of the invasion and the monthly and seasonal dynamics of the weed over Lake Tana to take corrective measures.

Description of Lake Tana Basin
Lake Tana Basin is located on the northwestern plateau of Ethiopia. Geographically, it is located from 10.9°N to 12.8°N latitude and 36.7°E to 38.2°E longitude. The Basin area of the Lake is approximately 15,123 km 2 , of which 3046 km 2 is the area of Lake Tana. It is a shallow lake with a maximum depth of 14 m and an average depth of 8 m. The volume of Lake Tana is 28.4 km 3 and it has a maximum length and width of 90 and 65 km, respectively. The Basin near the Lake is lowland and away from the Lake is a mountainous area. The average altitude of the Basin is 2035 m above the mean sea level with a large portion falling between 1686 and 2500 m (Fig. 1).
The total mean annual rainfall of the Lake Tana Basin was estimated to be 1331 mm, with a mean annual temperature of 19.5 °C from 2011 to 2018. The rainfall was highly seasonal: more than 78% occurred in summer and low rainfall was received in December, January, February, and March. The maximum mean monthly temperature was 29.4 ℃ in April followed by 29.3 ℃ in March with a minimum of 10 ℃ in winter. The mean monthly wind speed ranged from 0.8 m/s at Bahir Dar to 2.08 m/s at Aykel station The monthly wind speed showed an upward trend as of October/November (0.81/0.82 m/s) and peaks in April/May (1.25/1.23 m/s). The temporal variability of the sunshine hour was relatively high with a standard deviation of 1.7 h. The relative humidity of the Basin varied from 39.9% in April to 81.3% in August from 2011 to 2018.

Remote sensing data
The study used Landsat 7 ETM + and Landsat 8 OLI, which were accessed from the official website of the United States Geological Survey (USGS) (http:// glovis. usgs. gov/ web-link), and Sentinel-2 MSI satellite images from Sentinel Hub (https:// scihub. coper nicus. eu/) to analyze the annual spatiotemporal trend of the water hyacinth between 2011 and 2019 (excluding 2013). Landsat 7 was acquired on18/01/2011 and 23/01/2012; Landsat 8 was acquired on 16/01/2014 and 19/01/2015 and Sentinel-2 was acquired on 18/01/2016, 22/01/2017, 17/01/2018, and 22/01/2019. On the other hand, several Landsat 8 and Sentinel 2 images were also used from September 2016 to January 2019 period to analyze the monthly and seasonal spatiotemporal dynamics of water hyacinth and the surface area of the Lake, and the impact of climate variability on the spread of the weed. Most of the monthly images were acquired between the 20th and 31st days of the month. Two separate Landsat 7 Scenes with paths/rows of 170/51 and 170 /52, one Landsat 8 Scene with a path/row of 170/52, and three Sentinel-2 Tiles covered the study area. For this reason, a total of 83 and 60 Landsat images and Sentinel-2 images were downloaded, respectively, for annual and seasonal spatiotemporal analysis in this study.

Meteorological data
In the Ethiopian National Meteorological Agency (NMA), daily and monthly rainfall and temperature data were collected from 19 and 17 meteorological stations, respectively, between 2010 and 2018. Data on wind speed, relative humidity, and sunshine hours were obtained from the Class one level of four stations. Rainfall, temperature, and wind speed data were used to analyze the correlation between climate factors and changes in the water hyacinth area. Wind speed, relative humidity, and sunshine hours were used to estimate evapotranspiration, thereby helping to calculate the water hyacinth evapotranspiration and water losses in Lake Tana.

Ground truth data
Field data collection was conducted from May 07 to 09, 2019 to have the ground truth points of water hyacinth and other land cover classes using Handheld Global Position System (GPS Garmin 64) in three infested districts such as Libo Kemkem, Gondar Zuria, and Denbiya. A total of 51 Ground GPS points and 179 Google Earth points were collected to assess the accuracy of the final classified map. Google Earth Imagine was used as a supplementary reference to distinguish confusing water hyacinth classes from other aquatic vegetation.

Image pre-processing
All bands except band 6 were stacked together in Landsat 7 ETM+, whereas bands 2 to 7 were layer stacked in Landsat 8 (Topaloğlu et al. 2016). The Focal Analysis Tool in ERDAS Imagine software was used to remove the scan line errors in the Landsat 7 images. In the case of Sentinel-2 data, among the 13 multispectral bands, bands 2 to 8, band 8a, band 11, and 12 which have 10 and 20 m spatial resolution were stacked together (Shoko and Mutanga 2017;Thamaga and Dube 2018b). Then, the 10 and 20 m spatial resolution of the bands were resampled to 10 m spatial resolution using the nearest neighborhoods resampling method to make all the bands have a similar resolution (Marangoz et al. 2017). The projection transformation was carried out to make the remote sensing images compatible with the local datum system and assigned to the Adindan _UTM_ Zone_37N coordinate system. Moreover, histogram matching and equalization, brightness inversion, haze, and noise reduction were applied for images acquired in the summer season to increase the visual distinctions between features.

Mapping the spatio-temporal distribution of water hyacinth and water surface area in Lake Tana
Estimation of species-level biomass of invasive wetland vegetation using multispectral remote sensing is difficult because different invasive plant species have similar spectral reflectance (Ozesmi and Bauer 2002). In this study, manual visual classification (manual digitization) and maximum likelihood classification (MLC) algorithm methods were used to map the surface areas of water hyacinths and Lake Tana. Manual digitization was applied when water hyacinth biomass was dense, and the satellite imagery was overcast with thick clouds during the summer. This is because the method is suitable for mapping aquatic vegetation habitats (Cobbing 2006). On the other hand, the MLC algorithm was employed when water hyacinths were found scattered over the Lake due to difficulties in applying manual digitization techniques. The classification accuracy for images classified by MLC was assessed using the Kappa coefficient 'k' and overall accuracy.

Trend analysis of annual water hyacinth and surface area of Lake Tana
Mann-Kendall trend test was applied to assess the trend of water hyacinth and the surface area of Lake Tana (Sithranjan 2012). The mathematical formula for Mann-Kendall statistic (S) is as follows: The trend test is applied to a time series y i , i = 1,2,3, … n -1, and y j , j = i + 1, i + 2, i + 3, … n. Each data point y j is used as a reference point to be compared with the remaining data points.
Sen's slope estimator test Sen's slop nonparametric method was used to estimate the true slope of an existing trend such as the amount of change per year.
where β is Sen's slope estimate; β > 0 indicates an upward trend in a time series. Otherwise, the data series presents a downward trend during the study period. For all j > i and i = 1, 2… n−1 and j = 2, 3…, n (Rahman and Begum 2013).

Evapotranspiration of water hyacinth (ETc)
Many modeling techniques were used to estimate evapotranspiration in wetlands. These empirical formulas using climate data or models satisfy the water-energy balance equation. When using the water balance approach, the accuracy of wetland evaporation depends on other factors in water balance such as precipitation, runoff, and interaction with groundwater (Mohamed et al. 2012). In this study, the reference evapotranspiration (ETo) which is used to calculate evapotranspiration of water hyacinth was computed by the FAO Penman-Monteith method, using decision support software CROPWAT 8.0 (Sasaqi et al. 2019).

Referenced evapotranspiration (ET 0 )
The formula for daily ETo according to Penman Monteith's FAO56 method (Allen et al. 1998) is given as follows: where ET 0 is referenced evapotranspiration in mm/day, Δ is the slope of saturation vapor pressure curve (kPa/°C), γ is a psychometric coefficient (kPa/°C), Rn is the net radiation at the surface (MJm −2 /d −1 ), G is soil heat flux in MJm −2 d −1 , assumed zero on daily basis, T is the mean daily air temperature in °C, U 2 is the mean wind speed at 2 m (m/s), (es−ea) is vapor pressure deficit (kPa).
The average daily evapotranspiration (ETc) of water hyacinth was calculated using the following equation (Rashed 2014;Ali and El-Din Khedr 2018;Sasaqi et al. 2019).
where ETo: reference evapotranspiration in mm/day, Kc: crop coefficient of water hyacinth.
Crop coefficient (Kc) The Kc value for water hyacinth computed in different kinds of literature ranged from 0.65 to 1.90. The study by Meleha (2005) Dooenboss and Pruitt (1992), the Kc value is 1.1 (light to moderate wind) and 1.15 (strong wind), and Rashed, (2014) found a minimum Kc value of water hyacinth as 0.65 using FAO Penman-Monteith. In this study, a minimum Kc value of 0.65 and a maximum Kc value of 1.90 were adopted from Rashed (2014) and Jiménez-Rodríguez et al. (2019), respectively, to estimate the evapotranspiration of water hyacinth and water loss due to the Kc value of water hyacinth in Lake Tana has not been determined yet. If the Kc value of the water hyacinth is less than 1, the value of evapotranspiration of the water hyacinth is less than the evaporation from the free water surface. The average temperature and wind speed for Lake Tana, Ethiopia is less than Batujai Reservoir, Indonesia. Certainly, the Kc value of water hyacinth in Lake Tana could not be less than 0.65 and greater than 1.90.

Estimation of water losses due to evapotranspiration of water Hyacinth
The volume of water lost due to the evapotranspiration of water hyacinth was calculated using the following equation (Rashed 2014).
Where, reach water loss is in m 3 , reach the surface area is in m 2 , ETc is evapotranspiration of water hyacinth in mm/day.

Correlational analysis
Pearson's correlation coefficient was used to analyze the relationship between the monthly/seasonal area of water hyacinth with climatic elements and surface area of the Lake, and the relationship between rainfall and the surface area of the Lake. The monthly and seasonal correlational analysis between water hyacinth and climatic elements covered the periods from September 2016 to December 2018 whereas the annual correlation was covered from 2010 to 2019.
where, r xy is the simple correlation coefficients of water hyacinth and climate elements, x i is the area of water hyacinth for the ith year/month/season, y i is climatic elements of the ith year/month/season; x m is the area coverage of the weed for all years/month/season, y m is the average climate element for all years/month/season (Li et al. 2004).
The overall methodological flow chart of the study is presented in Fig. 2.

Accuracy assessment results
The accuracy assessment results show that the overall classification accuracy ranges from 95.78 to 99.47% and the kappa statistic is from 0.92 to 0.99 (Table 1; Fig. 3).

Annual spatiotemporal distribution of water hyacinth
The growth of the water hyacinth infestation increased in Lake Tana between its existence in 2011 and 2019. The weed area coverage increased by 96% from 2011 to 2019. However, it declined by 30% from 2015 to 2016 (Table 2). As seen in Fig. 4, water hyacinth began in the Gondar Zuria District and the infestation was relatively higher in the Denbiya District (between points 13 and 14).
The result of the Mann-Kendall trend test indicates that there was a statistically significant increment of the annual area of water hyacinth (p-value, 0 < alpha value) at a 95% confidence level (Table 3). However, the annual surface area of Lake Tana was deceased for the last 9 years by 335.54 ha per year, although it is not statistically significant (p-value, 0.95 > alpha value).
Gondar Zuria and Libo Kemkem were the most affected districts, whereas Denbiya and Dera were the least affected by the water hyacinth. However, the Denbiya district was the most adversely affected by water hyacinth invasive species compared with other districts from 2011 to 2017. Gondar Zuria was the most infested district in 2018 and this shows that the severity of the infestation of water hyacinth was expanded from the northern shore towards the eastern shore of the Lake (Fig. 5). Fikra Dangurie, Lemba Arbaytu, Mitriha Abawarka Kebeles (the lowest administrative unit in Ethiopia) from Gondar Zuria, and Agid Qiregna and Kab Abo Kebeles from Libo Kemkem Districts were heavily infested with the weed (Fig. 6).

Monthly spatiotemporal distribution of water hyacinth
The analysis of temporal satellite images showed that the spatial extent and distribution of water hyacinth fluctuate both monthly and seasonally from September 2016 to January 2019.     in the spatiotemporal distribution of water hyacinths in the monthly study period (Fig. 7). The area of the weed has increased steadily from July to November in 2017 and 2018 years while it began to decline in December and reached its lowest point again in July. Therefore, November and July were the turning points where the area of water hyacinths increased or decreased. The magnitude of the change of area in the water hyacinth varied from 43 to 1004 ha between June and July, and October and November 2017, respectively. It also varied from 17 to 1057 ha between February and March, and September and October in 2018. This shows the highest monthly variations seen between September and December. However, within a year, the highest change was observed between November and July in both 2017 and 2018, which accounted for 85 and 79% change, respectively (Fig. 8).

Seasonal spatiotemporal distribution of water hyacinth
Three seasons have been identified by the NMSA (1996) as summer (June to September), winter (October to January), and autumn (February to May). Summer is the main rainy season, autumn is the small rainy season and winter is the dry season in Ethiopia. The maximum and minimum mean seasonal water hyacinth areas were observed during the winter and summer seasons between October 2016 and January 2019. The weed was scattered in the summer season and moved from the coast of the Lake to the inside of the Lake with the help of flood. The highest concentrations of water hyacinth were observed along the north shore around Lamba Arbaytu Kebele of the Lake in the winter season. The quantity of water hyacinth observed in the autumn season shows a decrease in magnitude compared to the winter season, and the geographic locations were also shifted evidently to the eastern direction (from Gondar Zuria District in the north to Libo Kemkem District to the east shoreline of the Lake). The spatial extent and distribution of water hyacinth declined considerably as the season changed from winter to summer in the same year. The area of the weed decreased from winter to autumn and summer by 74 and 85%, respectively from October 2016 to September 2017 period. However, the proliferation of water hyacinths increased again by 64% from 2017 summer to winter (October 2017 to January 2018). Similarly, the abundance of water hyacinths was reduced by 42 and 93% from winter to autumn and summer, respectively between October 2017 and September 2018, and then increased by 65% in winter (October 2018 to January 2019) (Table 4).

Spatiotemporal change of surface area of lake tana
The annual maximum and minimum area of Lake Tana was 306,399 ha in 2015 and 302,952 ha in 2019 during the study period with an average of 304,555 ha ( Table 2). The highest annual surface area change of the Lake was 3447 ha between 2015 and 2019, while the lowest change was 22 ha between 2017 and 2018. This may be due to the highest annual mean rainfall of 1453 mm recorded in 2014 in the Basin and 3019 ha of water surfaces were covered by water hyacinth in 2019. A minimum lake surface area was also observed in 2016 next to 2019 possibly due to the lowest annual mean rainfall of 1214 mm recorded in 2015 in the Basin. A total of 1603 ha of the surface area of the Lake (0.5% of the average surface area of the Lake) was lost due to water hyacinths over the last nine years. Maximum changes were observed around the northern and eastern parts of the Lake due to areas occupied by weeds. The average monthly surface area of the Lake computed for the three years; 2016, 2017, and 2018 was 303,783, 304,382, and 305,749 ha, respectively. The highest surface area of Lake Tana was 317,257 ha observed in July 2018, while the lowest was 298,392 ha in June 2016 for the period from January 2016 to December 2018. The minimum area of Lake Tana was observed in June while the maximum area of the Lake was seen in August for 2016 and 2017 years and in July for 2018 years. The monthly surface area change of Lake Tana between the highest and lowest areas varied from 8699 ha (between June and August 2017) to 16,388 ha (between June and July 2018). This means that the area of the Lake has increased from 8699 ha (2.82%) to 16,88 ha (5.20%) per year due to rainfall. The annual water inflow to the lake is estimated to be 3843 million m 3 yr −1 from rainfall, 3970 million m 3 yr −1 from gauged rivers, and 2729 million m 3 yr −1 from ungauged rivers. The annual evaporation from the Lake and annual outflow at the outlet of the Blue Nile River are estimated as 5182.5 million and 4714 million m 3 yr −1 , respectively (Wale et al. 2009). More than ninety percent of the flow into Lake Tana is contributed by four major rivers: Megech, Rib, Gumara, and Gilgel Abay (Kebede et al. 2006). The surface area variation of Lake Tana reached its maximum between July and June during the study period (Fig. 9).
The seasonal surface area of Lake Tana varied from 300,906 ha in 2016 autumn to 310,157 ha in 2018 summer.  weeds was 2.39 and 6.99 mm/day at Kc values of 0.65 and 1.90, respectively. The evapotranspiration of water hyacinth was less than the reference evapotranspiration for the Kc value of 0.65 (Fig. 10). The water loss due to water hyacinth evapotranspiration at a Kc value of 0.65 ranged from 8032 m 3 /day on July 21, 2017, to 84,148 m 3 /day on November 11, 2018, with an average of 35,678 m 3 /day, while, the water loss at a Kc value of 1.90 varied from 23,478 m 3 /day on 21, July 2017 to 245,971 m 3 /day on November 11, 2018, with an average of 104,290 m 3 /day. Although the highest evapotranspiration of water hyacinth was observed in April 2017, greater water loss was observed in November 2018 due to the highest area of water hyacinth infestation in this month. The total amount of water loss in Lake Tana was 30,397,918 m 3 and 88,855,452 m 3 at Kc values of 0.65 and 1.90, respectively from September 2016 to December 2018 due to the evapotranspiration of water hyacinth. The average was about 60,000,000 m 3 , which is 0.21% of the volume (28.4 km 3 ) of Lake Tana (Fig. 11).

The correlation of water hyacinth area with climate elements and surfaces area of Lake Tana
The total annual rainfall in the Basin had an insignificant effect on the annual distribution of water hyacinth. However, the rainfall had moderate negative significant and insignificant effects on the monthly and seasonal distribution of water hyacinth, respectively. The annual rainfall in the Basin also showed insignificant positive effects on the annual surface area of Lake Tana, while, the monthly and seasonal rainfall had moderate positive significant and insignificant effects on the monthly and seasonal surface area of the Lake. This implies that the seasonal and monthly surface area of Lake Tana is more importantly affected by the total amount of rainfall in the Basin than the area of the water hyacinth ( Table 5). The temperature in the Basin had a negative insignificant effect on the monthly and seasonal expansion of water hyacinth. Besides, the monthly and seasonal wind speeds had a significant and insignificant moderate negative effect on the monthly and seasonal distribution of the weed, respectively.
There was an inverse spatial correlation between rainfall in the Basin and the location of water hyacinth in Lake Tana. The highest mean total annual rainfall value was seen in the southern and southeastern parts of the Lake, and relatively low rainfall was observed in the northern and western parts of the Lake throughout the study period. However, the greatest spread and change in water hyacinth was found on the north and east shores of the Lake. There was no water hyacinth to the west, south, or low in the southeast of the Lake. A large portion of the weed was found on the north and east shores of the Lake, in which the area exhibited a higher temperature.

Discussion
The analysis of consecutive annual satellite image results indicates that the annual area of water hyacinth has increased significantly from 2011 to 2019 as the computed p-value < 0.05 at a 95% confidence level. However, the surface area of Lake Tana showed a decreased trend while the rainfall availability in the Basin did not show any trend. The annual maximum and minimum lake surface area was 306,399 ha in 2015 and 302,952 ha in 2019 with an average surface area of 304,555 ha. The previous studies conducted in 2012, 2014, and 2015 that estimates the area of water hyacinth in Lake Tana were overestimated and extremely higher than the estimates in this study. BoEPLAU (2012) reported the area of water hyacinth was 20,000 ha in 2012 in three districts such as Denbiya, Gondar Zuria, and Libo Kemkem. However, the current study estimated the area of water hyacinth as 512 ha in Lake Tana in the same year. According to Wassie et al. (2014Wassie et al. ( , 2015, the area of water hyacinth in the Lake was 50,000 ha and 34,500 ha in 2014 and 2015, respectively which disagrees with this study. It is difficult to explain why the area of water hyacinth reported in 2012, 2014, and 2015 had a large deviation from this study because of no clear procedure stated in their reports. However, one thing could be understood from the reports. They were taking a GPS reading while traveling using a Boat on a free water surface and ground. A polygon was created based on these GPS points to compute the area of the water hyacinth. The problem may be the polygon includes not only the area covered by water hyacinth but also the free water surface, and the land surface even has no potential to be infested in the future. According to Dersseh et al. (2019), the area of water hyacinth in Lake Tana was less than 5,000 ha even in the peak growing   Dersseh et al. (2020) for similar months. The areal extent and distribution of water hyacinth in Lake Tana fluctuated both monthly and seasonally during the study period. The peak monthly area of water hyacinth was 3830 ha in November 2018, while, the minimum was 396 ha in July 2017. This result confirms the study of Ongore et al. (2018), the spatial-temporal dynamics of water hyacinth in Lake Victoria from 2014 to 2017 exhibited periodic cyclical patterns of emergence, growth, disappearance, and reappearance within a year, reaching the highest peak between September and November and disappearance in May. The favorable season for the rapid growth of water hyacinth in Lake Tana was winter consistent with the study of Dersseh et al. (2020). The study made by Ouma et al. (2005), reported that the highest quantity of water hyacinth in Lake Victoria was observed in November, more than twice that of March. However, this study contradicted the study of Mironga et al. (2014), and Thamaga and Dube (2018b), high water hyacinth coverage  was observed in the summer compared with the dry season in Lake Naivasha, Kenya, and Greater Letaba River System, South Africa, respectively. The maximum and minimum mean seasonal water hyacinth area was observed in the winter and summer seasons, respectively. This could be due to the highest values of phosphorus concentration in the dry season and the lowest in the rainy season (Dersseh et al. 2019). Another reason may be that the Lake area has decreased gradually after the rainy season allowing water hyacinth to have stagnant water and internal surface as it spreads rapidly in shallow bays and inlets with muddy surfaces (Mitchell 1976). The infestation of water hyacinth in the whole study period was observed in the north and east shorelines of Lake Tana. This can be explained by the geographical setting of the Lake in the northern and eastern parts seems to favor the expansion and distribution of water hyacinth since the area is flat, the Lake depth is shallow, a large portion of the Basin and four major rivers (Megech, Rib, Gumara, and Dirma) are found in this direction of the Lake. These rivers carry suspended sediments from upstream that favor the proliferation of water hyacinths during flood events. The severity of water hyacinth infestation, which is confined to this part of the Lake, may be related to the water flows and wind direction, which may have pushed water hyacinth to the north and east as the monthly and seasonal wind speed, showed a negative correction with the area of water hyacinth and a similar finding was reported by (Ouma et al. 2005).
The negative correlation between rainfall and the area of water hyacinth in seasonal and monthly periods could be due to heavy rains that manifest higher rates of run-off (Barbosa and Lakshmi Kumar 2012), and rainfall in the form of runoff is a major cause for population size reduction in aquatic plants (Kitamura et al. 2016). Besides, stagnant water and the shallow depth of the Lake are the causes of weeds' expansion (Dersseh et al. 2019). Moderate rainfall amounts are optimal for water hyacinth propagation, very high and very low rainfall slows the proliferation of the weed (Ouma et al. 2005). Based on the results of the study, in Lake Tana, the rainfall has indirectly contributed to the proliferation of water hyacinth by transporting nutrients during the rainy season from the Basin to the Lake.

Conclusions
The spatial coverage of water hyacinth increased steadily from year to year while manual and mechanical removal techniques of the weed have been intensively operational since 2012. The weeds resulted in a reduction of the surface area of water, and increased water loss due to evapotranspiration, For instance, about 1603 ha of the surface area of the Lake (0.5% of the average surface area of the Lake) was lost from 2011 to 2019, about 60 million m 3 volume of water was also lost in 2.4 years (September 2016 to December 2018) due to water hyacinth.
The weed control and management methods that have been used so far should be evaluated critically and thoroughly, and important mitigation approaches should be taken to curb the rapid spread of weeds in the Lake Tana Basin. Based on the findings of the study, the following recommendations are suggested for future research. The land-based water hyacinth and other water bodies (rivers) system require more involvement of a ground survey-based approach, the long-term spatiotemporal dynamics of Lake Tana before and after the existence of water hyacinth to understand whether other factors contribute to the reduction of the Lake size, the impact of evapotranspiration of water hyacinth on the water balance, the influence of monthly and seasonal Lake wind and current directions on the distribution of weed need to be studied. Also, to precisely compute the evapotranspiration of the weeds, the monthly crop coefficient (Kc) value for water hyacinth based on climate variables in the Lake Tana Basin should be estimated.