Time-series ARIMA modelling of the Labeobarbus spp (Cyprinidae) fishery in water hyacinth-infested and non-infested sites in Lake Tana, Ethiopia

Abstract Ethiopia’s largest freshwater lake, Lake Tana, is home to 21 endemic fish species, the majority of which are cyprinids in the genus Labeobarbus. The lake is undergoing numerous ecological changes, including an invasion of water hyacinth (Eichhornia crassipes (Mart.) Solms (WH). The aim of this study was to predict the future fish productivity in Lake Tana by considering the increasing water hyacinth infestation. This was accomplished by analyzing a 12-year fishery-independent time-series data set of Labeobarbus, collected from sites infested and not infested by water hyacinth. The stationarity of the data was investigated using the Augmented Dickey-Fuller (ADF) unit root test. First-order data differencing was applied to solve the non-stationarity data to stationarity. With the same standardized fishing gillnet and the same sites, the CPUE during the dry season decreased from 3,502 grams (3.5 kg/day) in 2010 to 360 grams (0.36 kg) in 2020 at the water hyacinth-infested sites, demonstrating a 90% decrease in the daily catch. A reduction in catch per unit effort (CPUE) is evident at the WH sites in all seasons, and the rate of fall in there was faster than at the non-infested site. Box Jenkin’s auto-regressive integrated moving average models (ARIMA) modelling predicted that Labeobarbus CPUE will decline by a threefold % by 2032 compared to the current catch. Based on these results, the most suitable model for all the seasons and areas was confirmed to be ARIMA (0, 1, 0) using the lowest value of Akaike Information Criteria (AIC’s). The fish production declines with expanding WH infestation will necessitate integrated water hyacinth reduction strategies, closed area and closed spawning seasons, and mesh size regulations to conserve and utilize the resource sustainably. As Labeobarbus is the only examined species in this study, forecasts should be made for other commercially significant fish species in the Lake, such as Oreochromis niloticus and Clarias gariepinus.


Introduction
Inland tropical fisheries are overexploited or mismanaged in many regions of the world and catches are declining (Watson and Pauly 2001).Annually, 24% of all fish produced globally from inland water bodies are from within Africa (Simon, 2012).Ethiopia is a landlocked country endowed with a variety of lakes and reservoirs, numerous small water bodies, and large floodplains that collectively make up about 13,637 km 2 (1.2% of the country's total land surface area) and are home to an enormous diversity of aquatic fauna (Tesfaye and Wolff 2014).Fisheries have long been a major source of protein for Ethiopians, especially for those who live adjacent to major inland water bodies such as Lakes Tana, Ziway, Hawassa, Chamo, and the Baro River (Tesfaye and Wolff 2014).
Lake Tana supports both commercial and artisanal fishing targeting three large species, including tilapia (Oreochromis niloticus, Cichlidae), catfish (Clarias gariepinus, Clariidae), and the endemic large barbs (Labeobarbus, Cyprinidae) (Anteneh et al. 2012).Due to the seventeen species flock of Labeobarbus, the most diverse cyprinid species flock still in existence in the world, the Lake is also known as a biodiversity hotspot (Stiassny and Getahun 2007).Despite the huge diversity (more than 2,400 species) of cyprinid fish's present in freshwater systems all over the world, the Labeobarbus species of Lake Tana are the only known intact species flock of large cyprinid fishes.One of the most vulnerable and overfished fish species in the lake is the Labeobarbus species, which move to rivers during the spawning season and are essential to the socioeconomic characteristics and ecosystem function of the lake.The Lake's overall fishing sector provides a living for more than 500,000 people (Gordon et al. 2007).Lake Tana's fisheries have been declining in recent years due to a variety of factors, including overfishing, illegal fishing methods, wetland-loss from floodplain agriculture, construction of dams, and riverine sand mining (Lindley and Techera 2017;Wondie et al. 2012).Catch per unit effort (CPUE) for commercially significant fish species in Lake Tana has declined from 177 kg/trip in 1993 to 56 kg/trip in 2010 (Wudneh 1998;De Graaf et al. 2006;Mohammed et al. 2013).However the endemic Labeobarbus spp.especially has decreased from 63 kg/trip in 1993 (Wudneh 1998 to 6 kg/trip in 2010 (Mohammed et al. 2013).In addition to the causes of the fish decline mentioned earlier, water hyacinth (Eichhornia crassippes) (hereafter, WH) has been rapidly expanding in the area along Lake Tana's northern and eastern shoreline since 2011 (Kibret and Worqlul 2018;Dersseh et al. 2020;Worqlul et al. 2020).
Several studies have estimated the extent and coverage of WH infestation in Lake Tana using various approaches and methodologies although the estimates differ greatly (Tewabe et al. 2017;Kibret and Worqlul 2018;Dersseh et al. 2020).Nevertheless, it is clear that the coverage has increased substantially.According to studies by Anteneh et al. (2012) and Tewabe et al. (2017), the original spread of the infestation was between 20,000 and 50,000 hectares within 3 years.Dersseh et al. (2020) detected 278.3 ha and 2,504.5 ha of WH infestation at the northeastern side of Lake Tana, which roughly covered 25 km 2 of the lake's surface area from 2015 to 2019, which is less than previously predicted.The water hyacinth population in Lake Tana normally reaches its peak during the rainy season and dips at the end of the dry season, which is consistent with seasonal trends seen in other tropical water bodies (Dube et al. 2017).
It has been hypothesized that the dense, interlocked WH mats have an impact on ecology and biodiversity in both beneficial and detrimental ways by reducing phytoplankton productivity and dissolved oxygen levels (Villamagna and Murphy 2010).It can also reduce fish catches by tangling fishing nets and smothering fish breeding grounds (Tewabe et al. 2017).Water hyacinth also has an adverse effect on native aquatic plant communities (Kateregga and Sterner 2007;Mironga et al. 2014).Furthermore, water hyacinth negatively impacts people in communities near Lake Tana by blocking water ways, obstructing irrigation channels, impeding travel and recreational activities, reducing tourism, and promoting the growth of mosquito larvae and leeches that pose a health risk to the general public (Tewabe et al. 2017;Gezie et al. 2018).
Although the aforementioned research determined that WH negatively affects aquatic biodiversity, other studies have noted that the plant's intricate root system fosters a favorable environment for the growth of epiphytic invertebrates.(Mereta et al. 2012;Toft et al. 2003).Gezie et al. (2018) reported that there were more macroinvertebrates in the locations where WH infestation had been observed at Lake Tana.In addition, in some tropical lakes, water hyacinth is also thought to have supported the recovery of native hypoxia tolerant fish species, such as Clarias gariepinus, Protopterus aethiopicus, and Oreochromis niloticus (Njiru et al. 2002).
Most earlier studies conducted at Lake Tana by fisheries scientists focused on effects of WH on the commercial CPUE and on socioeconomic impacts.The effects of WH on fish populations remain understood despite information from several studies on the impact of anthropogenic activities on the diversity and population processes of the fishes in the lake.Recent research in Lake Tana and other tropical lakes experiencing infestations of WH has centered primarily on how WH influenced fishing processes and coverage (Anteneh et al. 2012;Tewabe et al. 2017;Dersseh et al. 2020).
In this study, we focused on changes in the CPUE of Labeobarbus fishes in Lake Tana following the introduction and spread of WH.We assessed the impact of WH on CPUE using time series data of catch, effort, and locations in the wet and dry seasons at two WH-infested sites (Dirma and Gedamat) and two non-infested sites (Zegie and Gerima).We used different time series models with the hope that they will assist policymakers in sustainably managing the fishery through appropriate management practices and the integration of WH control options.
Time series analyses have been widely used in fisheries science to forecast catch and landings of various species, ranging from CPUE of rock lobster (Saila et al. 2010), monthly pilchard catches (Stergiou et al. 1996), and Indian mackerel catches (Noble and Sathianandan 1991).Anuja et al. (2017) applied Auto Regressive Integrated Moving Average (ARIMA) and Regression models to examine fish production trends and future forecasts, and suggested that ARIMA is better than regression analysis for evaluating the significance of any variable.Mah et al. (2018) used ARIMA and Auto Regressive Fractionally Integrated Moving Average (ARFIMA) models to forecast freshwater and marine fish production in Malaysia, while Raman et al. (2018) forecasted fish production in Chilika Lagoon in India.Despite their widespread use of time series analyses in the aforementioned several states, such modeling and forecasts of fish CPUE have not been previously applied to the Lake Tana fishery.Therefore, the primary goal of this research was to determine the impact of WH on Labeobarbus CPUE, forecast future fish catches and investigate seasonal fluctuations in the fishery.The findings will assist policymakers to sustainably manage the fishery by adapting the tools for fisheries management and integrating WH control options.

Description of the study area
Lake Tana is the largest freshwater lake in Ethiopia and third largest in the Nile Basin, with a surface area of 3,150 − 3,500 km 2 , a seasonal water level fluctuation of roughly 1.6 m, and a storage volume of 28.4 km 3 (Duan and Bastiaanssen 2013).It is a relatively shallow lake, with an average depth of 8 m and maximum depth of 14 m (Dejen et al. 2004a;Vijverberg et al. 2009).Twenty one of Lake Tana's 28 fish species are endemic to the lake (Getahun and Dejen 2012;Nagelkerke et al. 1994).Seven large permanent rivers, Gilgel Abay, Gelda, Gumara, Rib, Arno-Garno, Megech, and Dirma, and 40 seasonal rivers, all flow into the lake (Anteneh et al. 2012).The four most important in-flowing rivers are the Gilgel Abay, Rib, Gumara, and Megech which, in total, contribute nearly 93% of the inflow (Kebede et al. 2006) and 58% of the sediment load to Lake Tana (Lemma et al. 2018).These rivers serve as refuges and spawning grounds for 17 Labeobarbus and eleven other fish species and are essential for the growth of agriculture, hydropower, and fisheries (Nagelkerke et al. 1994).The Blue Nile River and Tana-Beles diversion tunnels are the lake's only outflows.
During the wet seasons, the aforementioned in-flowing rivers carry 9.8 kg m −2 yr −1 (Lemma et al. 2020) of suspended sediment into Lake Tana, raising the lake's turbidity and nutrient concentrations (Wondie et al. 2007;Goshu and Aynalem 2017).Short-term studies have shown that the trophic state of the lake has changed from oligo-mesotrophic to seasonally meso-eutrophic, especially near inflowing river mouths (Fetahi 2019;Wondie et al. 2007;Dejen et al. 2017;Dersseh et al. 2022).
The climate around Lake Tana is classified as tropical highland monsoon, with one main rainy season (MRS = WET season) between July and September that contributes more than 60% of the region's annual precipitation.The remaining months of the year can be broken down into three categories: pre-rainy season (PrRS, May-June), dry season (DS -December to April), and post-rainy season (PoRS-October to November) (Wondie et al. 2007).

Fish sampling
This research utilized Labeobarbus CPUE data from two sites infested by WH areas in the north side of the Lake (Dirma, Gedamat) and two non-infested sites (Zegie, Gerima) sites (Figure 1).Dirma is near the Dirma River mouth, Gedamat and Gerima are littoral habitats and Zegie (> 10 m depth) is an open-water habitat which is found in the southern part of the Lake.The first reports of this weed invading Lake Tana were made in September 2011.The time series data for this study were collected in the spot where Bahirdar Fisheries and other Aquatic Life Research Center (hereafter, BFALRC) had been collecting for the past ten years.The northern part of Lake Tana, which encompasses Dirma and Gedamat, has been afflicted with WH since the beginning of the invasion.We used these sites as the WH-infested representative sites for comparison with the non-infested sites, which are represented by Zegie and Gerima in the southern side of the lake.BFALRC was also sampling fish using the same standardized gillnet sampling as the current study's sampling time in 2019-2020.Due to the BFALRC's lack of data collection in other areas of the lake, we chose the sites that have the research center's time series data.
Fishery-independent time-series data for this study were collected in the wet and dry season by BFALRC from 2009 to 2018.Data for 2019 and 2020 were collected by this study's authors and applied to the same species in a particular geographical area using the same standardized data collection procedure as BFALRC.Samples were collected on the overnight catch of the Labeobarbus fish production in Lake Tana during the main rainy season, which is referred to as the wet season for the current analysis, and during the dry season.
The unit of measurement for the sampled fishes was their wet weight at the time of removal from the water in grams per single sampling day.Multi-panel gill nets measuring 100 m in length and with mesh sizes of 6, 8, 10, 12, and 14 cm were always used.These were set at 0400 and 0600 h and pulled out the next morning between 0600 and 0800.The daily catch was standardized for the typical number of gillnets per trip in order to minimize bias.We combined the catch of all Labeobarbus species at the genus level to simplify the analysis.

Catch per unit effort (CPUE) analysis
Catch per unit effort (CPUE) is a measure of species abundance used in the evaluation of fisheries resources.It can also provide information on the sustainability of fishing operations in a particular geographic area (Dunn et al. 2000).CPUE was calculated for the months of May to September (wet season) and December to March (dry season).In both seasons, the average CPUE (g) was calculated.
The total catch for the day divided by the number of days of fishing effort is represented by the average CPUE from the experimental sample per day for the dry and wet seasons.

Average daily CPUE in grams
Catch (C) (usually expressed in weight), f i is its respective fishing effort.

Model identification
Autocorrelation function (ACF) and partial autocorrelation function (PACF) were established after the non-stationary data from the daily Labeobarbus CPUE were transformed to stationary data using ADF and the 'auto.arima()' command in R software.Optimal values of p and q for an ARIMA (p, d, q) model were estimated by analyzing the ACF and PACF plots of the stationary time series.The autocorrelation function ρ(k) at lag k was denoted by: Where y(K) is the auto-covariance function at lag K of a stationary random function {Y(t)} The PACF produced an autoregressive (AR) of order p when the ACF tails off and has a cut-off at p.In cases when the ACF additionally had a non-zero lag at q, a moving-average (MA) of order q is produced.
An autoregressive model (AR) of a time series X t is a regression model of that time series on its previous history.The next model was used to discover the autoregressive process of order (p).
In an effort to smooth the process or make the time series stationary, a moving average (MA) model of a time series tried to average out past error steps of a time series (Xt).The next model was used to discover the Moving Average process of order (q); X t j q j t j t The Autoregressive Moving Average (ARMA) is created by combining the features of linear autoregressive and moving average: ARMA of the (p, q) order is, According to Judge et al. (1988), the general form of the ARIMA model of order (p, d, and q) is as follows: where X t is the original data series or difference of degree d of the original data at time t, a t the random error, innovation or shock at time t, Ø 1 , Ø 2 , …, Ø p the autoregressive parameters, p the autoregressive order, θq a constant term, θ 1 , θ 2 , …, θ q the moving average parameter, q the moving average order and w t is the white noise at time t (Box and Jenkins 1976).

Stationarity analysis, model parameter estimation and validation
The stationarity of the data was examined using the Augmented Dickey-Fuller (ADF) unit root test (1979).In order for the ARIMA to function, the time series must be stationary, which necessitates that its mean, variance, and autocorrelation be constant.First-order data differencing was also carried out, the non-stationarity issue was solved by using the Box-Cox method (Box and Tiao 1976) by the forecast package and 'auto.arima' command in R software version 4.2.3 and stationary tests were performed on the newly created series of data.In order to determine the values of p and q in the ARIMA models, the Autocorrelation Function (ACF) (Figure 3) and Partial auto-correlation Function (PACF) plots (Figure 4) were plotted after confirming that the new data series was stationary.The best-fit model was based on the minimal Akaike Information Criteria (AIC) (Pal et al. 2007).The Akaike Information Criteria AIC's (Akaike 1973) minimum value was used as the basis for model selection to select the best-fitting model out of all those that could have been fitted after the Labeobarbus daily CPUE data sets became stationary (Table 2).

Diagnostic checking
Diagnostic checking of the recommended best model was done by examining the model's residuals to see if any systematic structure was present that might be eliminated to enhance the chosen ARIMA.Diagnostic tests were carried out to check to what extent the forecast could be trusted by plotting the residual errors ACF and PACF and the

Forecasting
For the daily Labeobarbus CPUE time series data, the best candidate ARIMA (p, d, q) was selected and then the parameters of the best model were estimated.The future daily  Labeobarbus CPUE in Lake Tana was then forecasted using the fitted ARIMA models as a predictive model for up until 2032.

Stationarity analysis
The initial phase in the ARIMA forecasting process began with the identification of the model through the stationarity test, followed by the transformation of the non-stationary data to stationarity and the plotting of the ACF and PACF plots.The stationarity of the daily Labeobarbus CPUE time series data was tested using several tests.Visualizing the data using a graphical analysis method (Figure 2) is the technique that is most utilized to prove stationarity.Since there are trends in some parts from 2009 to 2020, it was obvious that the data were not stationary and had a constant variance.When the stationarity is examined, except for the dry season non-infested sites (Dry non-infested), the Augmented Dickey-Fuller (ADF) p-value in this study was greater than 0.05, indicating that the data were not stationary across all sites and seasons (Table 1).
Labeobarbus CPUE at WH-infested sites showed a significant falling trend with an accelerating rate of reduction over the dry and wet seasons (Figure 2).CPUE increased from 2009 to 2010 (Figure 2).However, the CPUE in the Dry season declined from 3,502 grams (3.5 kg/day) in 2010 to 360 grams (0.36 kg) in 2020 at the WH-infested sites, showing a 90% decline in the daily catch using the same standardized fishing gillnet and the same sites (Table 1).However, this drastic decline typically refers to recent years, specifically from 2013 onward.Box Jenkin's auto-regressive integrated moving average models (ARIMA) predicted that the subsequent CPUE from the time series data will show a threefold reduction in 2032.In light of this, it is possible that the trend of the forecasted catch about the future drop in CPUE output will change for the worse or stay the same (Figure 7).
Plots of the partial autocorrelation function (PACF) and autocorrelation (ACF) showed some lags to have high autocorrelation.As a result, the ARIMA model was employed to model and predict the time series data of CPUE of Labeobarbus species from Lake Tana.

Model estimation
It was discovered that the model's coefficient parameters were statistically significant, which is necessary for predicting models.The best-fitting model using lowest AIC was subsequently put through a diagnosis test to see how well it matched the data.Forecasts were made if the model succeeded in the test.Based on the findings, the most suitable model for all the seasons and areas was confirmed to be ARIMA (0, 1, 0) (Table 2).

Diagnostic checks
The plots of the ACF and PACF residuals showed that none of the autocorrelations were statistically different from zero at any reasonable threshold.This indicated that the selected ARIMA model was suitable for forecasting the Labeobarbus CPUE.Figures 5  and 6 shows the auto-correlation function and partial auto-correlation function of the residuals from both data sets.The plots showed that the autocorrelation coefficients of the residuals are within the 95% confidence interval, and that the model chosen was the most effective one for predicting Labeobarbus CPUE across all sites and seasons.

Forecasting fish production (CPUE)
The fitted ARIMA model was used to forecast the Labeobarbus CPUE for the period beyond 2020 (Figure 7).The species indicated that their daily CPUE were correlated with the CPUE from the prior year, as shown by the Auto Correlation function (ACF) plots and the Partial Autocorrelation Function (PACF) plots in Figures 3 and 4, respectively.

Discussion
The findings of this study's time series data non stationarity is commensurate with Zuur and Pierce (2004) research finding that most fisheries data are non-stationary.The Labeobarbus CPUE forecast shown in Figure 7 indicates that the rate of CPUE decline will continue in the coming 10 years.The findings show that the mean Labeobarbus CPUE is drastically falling across all sites and seasons, with infested sites declining more rapidly than non-infested sites.This trend is expected to continue for the next 8 to 10 years, based on the forecast ARIMA model (Figure 7).The probable cause for the decline in the CPUE is mainly the WH invasion as the infested site is showing higher and rapid decline compared to the non-infested site.Our study's findings concur with those of Tewabe et al. (2017) who found that the WH invasion of Lake Tana had an impact on people's livelihoods as well as the production of fish, crops, and livestock.In agreement with the results of this study, Mailu (2001) also found that this weed affected the amount and quality of water in Lake Victoria, as well as the flora and fauna resources.In addition to the infestation of water hyacinth, Lake Tana's Labeobarbus populations are currently dealing with several problems, such as overfishing, illegal fishing, the building of dams for irrigation and hydropower, and sand mining, all of which have impacted their habitat and spawning grounds.Both point and nonpoint sources are likely responsible sources that may also have contributed to declining water quality and the observed declines in Labeobarbus CPUE over the past ten years.
We observed a sharp decline in mean Labeobarbus CPUE across all sites and seasons, but with infested sites declining more abruptly than non-infested sites.Even without a WH infestation at the Zegie site, other variables such as illegal fishing sedimentation and nutrient loading might be responsible for the CPUE declines.
From the late 1990s to 2011, the Labeobarbus Spp CPUE in the commercial catch has decreased from 63 kg/trip in 1993 (Wudneh 1998 to 6 kg/trip in 2010 (Mohammed et al. 2013).This is consistent with observations from the current study.According to earlier research, causes like overfishing, illegal fishing, habitat destruction, construction of dams for irrigation and hydropower and sedimentation, were to blame for the decline of the lake's commercial fishery (Dejen et al. 2017;Gebremedhin et al. 2018).The typical fishing gears used throughout the Lake Tana fisheries, which are intended to catch all the small fishes before they reproduce themselves, include illegal undersized monofilament nets, beach seines, and long-lines with various hook sizes.Currently, however, illegal fishing activity is increasing because fishermen are switching to trawling nets, which are much more harmful because they catch all fish sizes, including the larger and juvenile fishes (personal observation).
Most studies on WH expansion in the tropics found that it caused biodiversity loss, fish habitat degradation, dysfunctional food webs, sedimentation, increased turbidity, eutrophication, a reduction in the amount of dissolved oxygen, and a loss of phytoplankton, all of which led to a decline in water quality (Kateregga and Sterner 2007;Villamagna and Murphy 2010;Tewabe et al. 2017;Dersseh et al. 2020).In the current study, we noticed that WH growth at Dirma makes it difficult for fishermen to set their gillnets and it has also entangled our gillnets during our sampling time, which is consistent with the majority of the research findings stated above, especially those carried out in Lake Tana.
Although the water hyacinth's life history in Lake Tana is not yet fully known, blossoming, or the occurrence of sexual reproduction, has frequently been noticed year long.Research by Cronk and Fenness,(2016) found that each inflorescence can yield thousands of seeds, that are enclosed in a capsule and finally sink into the sediments.However, we have observed that the greatest WH bloom occurred in December.We observed that the invasive weed's coverage on the lake's northeast and eastern shores had significantly increased, and the eastern shore had recently been invaded.Thus, it follows that this rise in the infestation will surely have an effect on CPUE of the Labeobarbus species.
A recent decline in water quality in Lake Tana has been noted, particularly near river mouths in the north and east (such as Dirma, Megech, Rib, and Gumara) (Goshu and Aynalem 2017), resulting in high phosphorus concentrations in sediments and lake water in the northeastern part of the lake (Dersseh et al. 2022).Water hyacinth densities in Lake Tana are particularly high along the northern and eastern shorelines.Given the rising water level and increased eutrophication in the watershed area, water hyacinth infestation may still spread to nearby floodplains even if it has become stable in the north and northeastern corner of Lake Tana (Dersseh et al. 2022).Currently there is also a water hyacinth infestation in some parts of the southern area of the lake, especially in the vicinity of the 'Woramit' area.
Water hyacinth mainly concentrates along the shore and moves to the outflows of tributary rivers or streams during the dry season, especially between the months of February and April.This is a result of the wind and wave direction, as well as the lake's hydrology, which is also supported by Dersseh et al. (2020) who asserted that the increased evapotranspiration brought on by the weed mat will have a significant impact on the hydrology.Water hyacinth in Lake Tana exhibits more vigorous growth during the dry season than during the wet season.Considering that Dirma River flows into the lake and serves as breeding habitat for the endangered Labeobarbus species, the declining CPUE may be a response to just the spread of WH, but also the fishing pressure to which riverine spawners, sedimentation, and other factors that are the general plight of the other part of the Lake are particularly sensitive.Ogutu-Ohwayo (1990) asserts that most of large African cyprinids that have adapted to lakes concentrate around river mouths to spawn and are especially vulnerable to different challenges, which may also be the reason for the decline in the Dirma River.
The collapse of the Labeo mesops fisheries in Lake Malawi (Skelton et al. 1991), the decrease in the number of Labeo victorianus (Cadwalladr 1965), and the drastic decline in the population of the endemic species flock in Lake Tana in previous studies are all examples of the African cyprinids' susceptibility to overexploitation (De Graaf et al. 2006;Mohammed et al. 2013;Dejen et al. 2017).The current study's findings regarding the sharp decline of the Labeobarbus CPUE and the forecast that the decline will continue for the next 8-10 years call for enhanced fisheries management and WH control before it spreads to the entire lake.
The newly established Lake Tana and Other Water bodies Protection and Development Agency (LTOWPDA) and the community itself has made significant efforts to eradicate water hyacinth in Lake Tana, including physical removal (using mechanical and human harvesting methods) and mobilizing concerned Ethiopians to lessen the WH coverage.According to the LTOWDPA, its WH eradication efforts were 90% successful in 2020 (personal communication and media).However, the water hyacinth's biological characteristics-rapid growth rate, ease of propagation, excellent resilience to harsh conditions-as well as the fact that the harvested or removed water hyacinth is primarily dumped near the lake's shore-another favorable condition for regeneration allow the population to quickly regenerate after removal.Furthermore, the long-lasting seeds, which may persist in sediments for up to 20 years, make weed eradication considerably more difficult (Gopal 1987).Due to the weed's sexual reproduction through seeds and unchecked vegetative growth through budding production eradicating water hyacinth has been difficult in different water bodies in the world (Jawed et al. 2022).
Water hyacinth first significantly declined in Lake Victoria in the late 1990s as a result of the El Nino weather trend and release of a herbivorous weevil for biocontrol (Albright et al. 2004;Wilson et al. 2007).Once established, water hyacinth is exceedingly difficult to eliminate if the management effort is not sustained, and it still occupies a 32 km 2 region in the Winam Gulf (Nyawacha et al. 2021).Harvesting water hyacinth during the dry season may be useful for reducing negative effects (Eid and Shaltout 2017) considerations should be made to avoid potential seed germination.If the existing removal strategy is maintained, WH infestation will be a usual concern in Lake Tana.Solutions to minimize the negative impacts of WH include long-term lake restoration and integrated water hyacinth control strategies, implementation of the fisheries management tools like closing the breeding season, closed spawning area, mesh size regulation, taxing of fishers and licensing to protect the decline in the Labeobarbus CPUE.

Conclusions and recommendations
In conclusion, our research result revealed that the CPUE of the endemic Labeobarbus in Lake Tana drastically decreased and will likely continue to be so at the current rate.As a result, various forms of mismanagement and the difficulties the Lake is experiencing are the causes of the expected threefold percent dropping tendency of the Labeobarbus CPUE in the following 10 years.If the current fishing pressure continues to focus on the spawning grounds and spawning seasons of Labeobarbus species, and if this situation is not controlled and prohibited as well as if the WH coverage is not well managed, the future of Lake Tana cyprinids and their fisheries may follow a similar path to that of other African cyprinid fisheries.The forecast based on previous data on the condition and growth could be further studied as the current research is focused solely on the CPUE.As Labeobarbus is the only species being examined in this study, forecasts should be made for other commercially significant fish species in the Lake, such as Oreochromis niloticus and Clarias gariepinus.

Figure 1 .
Figure 1.lake tana in north-western ethiopia, and the location of the four study sites.

Figure 2 .
Figure 2. average daily Labeobarbus cPue (gram/hour) from lake tana for period 2009-2020 in the wet and dry seasons at the infested and non-infested sites.
Ljung-Box test, as these were widely used and efficient in model validation(Dare  et al. 2022).

Figure 3 .
Figure 3. auto correlation plot for the wet and dry infested and non-infested sites.

Figure 4 .
Figure 4. Partial auto correlation plot for the wet and dry infested and non-infested sites.

Figure 5 .
Figure 5. acf residuals plots for all the seasons and sites.

Figure 7 .
Figure 7. arIMa forecasts of the Labeobarbus daily cPue for 2022 to 2032.

Table 1 .
augmented Dickley-fuller test for all the sites and seasons.
Figure 6.Pacf residuals plots for all the seasons and sites.

Table 2 .
aIc and log likelihood of the fitted arIMa models of all the sites in the two seasons.