The local change in mean sea-level variation and water supply to the ocean by rivers were obtained by using daily sea-level anomalies and river discharges provided by the Copernicus Marine Environment Monitoring Service. The spatial horizontal grid of sea level anomaly and river discharge dataset are of 0.25˚ × 0.25˚ and 0.1˚ × 0.1˚ grids, respectively, and the coverage period is from 1993 to 2021 at daily timescale. These gridded mean sea level data are derived from satellite altimeters [22] . The daily river discharge dataset is produced by forcing a hydrological modelling chain with ERA5 meteorological reanalysis data, interpolated to the Global Flood Awareness System (GloFAS) resolution [23] .

The Ocean heat content dataset used in the study was provided by the Institute of Atmospheric Physics Ocean Gridded Product. This dataset is a combination of XBT measurements together with all available in situ observations from the World Ocean Database. It covers the global ocean at 1˚ × 1˚ horizontal resolution on 41 vertical levels from 0 - 2000 m, and monthly resolution from 1940 to the present [24] .

All these data were spatially averaged at the limit of each coastal section of the West African coastline defined previously in [25] to obtain a chronological series for each coastal section.

Coastal sea level anomaly, river discharge and ocean heat content at different depth time series for each coastal section were analyzed by several statistical methods including multiple change point detection, multiple linear regression and the Mann-Kendall non-parametric test for randomness against trend.

Detection of abrupt changes in the characteristics of some oceanographic processes such as sea level, river discharge and ocean heat content, highlights regime changes in these processes and trend analysis. In this study, the detection of the number of change points and their positions in ocean parameters time series are computed using the Pruned Exact Linear Time (PELT) method of [26] . This method can be orders of magnitude faster than Optimal Partitioning, particularly for long data sets. The PELT leads to a substantially more accurate segmentation of data. It combines optimal partitioning and pruning to achieve an exact and efficient computational cost which is linear in n (length of the time series). The optimal segmentation $F_n$ is defined as follows:

$F_n={}_{\tau }{}^{\min }\left[\underset{i=1}{\overset{m+1}{{\displaystyle \sum }}}\left[C\left(y_{{\tau }_{i-1}+1}+\cdots +y_{{\tau }_i}\right)+\beta \right]\right]$ (1)

where:

C is a cost function for the $i^{th}$ segment, $\beta$ is a penalty to guard against over fitting; m is the number of changepoints; $y_1,y_2,\cdots ,y_n$is a set of data and ${\tau }_m$ is a location of the $m^{th}$.

This method consists of by calculating $F_1$ and then recursively calculate $F_1,F_2,\cdots ,F_n$. At each step we store the optimal segmentation up to ${\tau }_{m+1}$. When $F_n$ is reached, the optimal segmentation for the entire data is identified and the number and location of changepoints were recorded.

At each step the minimisation over ${\tau }_m$ covers all previous values. The computational efficiency of the PELT method is achieved by removing candidate values of ${\tau }_m$ from the minimisation at each step [27] .

The increasing or decreasing of the ocean parameters trends was carried out by fitting Mann-Kendall test and Sen’s slope estimator [28] was carried out by fitting time series. This method is useful to detect and estimate the slope of possible trends in time series as decribed in [25] was used for this study.

The multiple linear regression is used to predict the response of predictor variables [29] . This method that was previously used by [30] to reconstruct sea level in some location sites in the south west Pacific region is adopted in this study to highlight ocean heat content and river discharge impact on sea level in West African basin. The prediction could be modeled in the case of two predictors and their combination is as follows:

$Y={\alpha }_0+{\alpha }_1{X}_1+{\alpha }_2{X}_2+{\alpha }_3{X}_1{X}_2$ (2)

where:

${\alpha }_0$, ${\alpha }_1$, ${\alpha }_2$ and ${\alpha }_3$ are model coefficients.

$Y$ is the response of predictors influences.

$X_1$ and $X_1$ are the predictors variables.

$X_1{X}_2$is a predictor variable resulting of the combination of the impact of $X_1$ and $X_2$ at the same time.

This method was used to carry out the possible inter-relationship between ocean parameters analyzed in this study.

The analysis of correlation between sea level and ocean heat content anomalies has shown a strong link ( $r>84\%$) between these two parameters at 2000 m depth. So, the ocean heat content data at this depth was be used to analyze its impact on sea level rise.

Figure 1 shows the temporal variability of ocean heat content at 2000 m depth along each coastal section of the study area.

The average ocean heat content in each coastal section of West African coastline shows that the minimum and maximum value are $5.1099\times {10}^{10}\text{J}\cdot {\text{m}}^{-2}$ (Togo) and $5.6851\times {10}^{10}\text{J}\cdot {\text{m}}^{-2}$ (Section 3 of Ghana coastline), respectively. The mean values are $5.4386\times {10}^{10}$ (Cote d’Ivoire), $5.4344\times {10}^{10}\text{J}\cdot {\text{m}}^{-2}$ (Ghana), $5.4177\times {10}^{10}\text{J}\cdot {\text{m}}^{-2}$ (Nigeria), $5.3495\times {10}^{10}\text{J}\cdot {\text{m}}^{-2}$ (Togo) and $5.3494\times {10}^{10}\text{J}\cdot {\text{m}}^{-2}$ (Benin). This result shows that ocean warming decreases from Cote d’Ivoire to Benin and, increases from Benin to Nigeria.

The ocean heat content regime has an uni-modal structure in seasonal cycle which is centred in June for Nigeria basin and July along the other coastline sections of West Africa. This result shows that the yearly cooling appears one month earlier in Nigeria basin than those of other basins. These minima appear during the major coastal upwelling period corresponding to the annual minimum value of sea level [3] . The minor upwelling is not well represented at this depth by heat content. The temporal variability exhibits five periods interupted by four breaks with unequal durations ( Figure 1 ). The trends observed between 2000 and 2010 from Côte d’Ivoire to Ghana (Section 1) are not significant. The same situation is observed after 2016 from Cote d’Ivoire to Ghana (sections 1 and 2) and along the coastal sections 3 and 4 of Nigeria. Positive trends are observed in each coastal section over the 1993 – 2000 period. This increasing trend continued until 2010 along the Nigerian coastline. The slopes of these trends ranged between 0.12 $\text{J}\cdot {\text{m}}^{-2}\cdot {\text{yr}}^{-1}$ (Ghana) and $22.86\times {10}^6\text{J}\cdot {\text{m}}^{-2}\cdot {\text{yr}}^{-1}$ (Ghana). A decreasing of

thermal expansion is noted during 2000 to 2021 along the coastal section 1 of Côte d’Ivoire and from coastal section 2 of Ghana to Benin. This trend is observed over the 2016-2021 period through Nigeria basin. The associated coastal slopes do not exceed $-0.007\times {10}^6\text{J}\cdot {\text{m}}^{-2}\cdot {\text{yr}}^{-1}$. Long-term trends indicate that there is a spatial and temporal increasing of heat content from Cote d’Ivoire to Nigeria. These upward trends have slopes between $0.97\times {10}^6\text{J}\cdot {\text{m}}^{-2}\cdot {\text{yr}}^{-1}$ (Nigeria) and $1.14\times {10}^6\text{J}\cdot {\text{m}}^{-2}\cdot {\text{yr}}^{-1}$ (Cote d’Ivoire). The spatial and temporal variability of thermal expansion along this coastline could be linked to the El-Niño, ocean current and over climate index such as Atlantic Multidecadal oscillation that influence Gulf of Guinea ocean basin [31] .

The coastal river discharge is another parameter which could play an important role in the mean sea level change at spatial and temporal scales.

Figure 2 shows the monthly cumulative river discharge climatology calculated for each country coastline in the study area between 1993 and 2021.

From Cote d’Ivoire to Togo, a bi-modal structure is observed in the seasonal cycle of the river discharge regime. The peaks are centered in June and October. Figure 2 illustrates the weak impact of the major rainy season on river discharge. The high values of river discharge observed in October could be linked to both minor and major rainy seasons along the coastal area and/or in the sahel region.

The river discharge regime has an uni-modal structure in the seasonal cycle along the Benin and Nigeria coastlines. The peak of this seasonal cycle is centered in September and coincides with the major rainy season. One can conclude that the river discharge regime alone is governed mostly by the coastal precipitation regime of Nigeria. The different modes observed in the seasonal cycle of river discharge along the different coastal sections are governed by the latitudinal displacement of the Inter-Tropical Convergence Zone (ITCZ) [25] [32] [33] .

Figure 3 illustrates the temporal evolution of the cumulative river discharge in

each coastal section of the northern coast of the Gulf of Guinea over the 1993-2021 period. The temporal average of the cumulative river discharge in each country coastal basin is characterized by the values ranging between 1.119 m³∙s⁻¹ (Togo) and 3463.8 m³∙s⁻¹ (Nigeria). The mean values are 56.365 m³∙s⁻¹ (Cote d’Ivoire), 96.001 m³∙s⁻¹ (Ghana), 16.156 m³∙s⁻¹ (Togo), 107.995 m³∙s⁻¹ (Benin), 461.776 m³∙s⁻¹ (Nigeria).

The analysis of change points by linear segmentation method [26] [34] in each coastal subsection time series exhibits three periods along the coastal subsections 1 and 3 of Cote d’Ivoire, 2 of Ghana and 1 of Nigeria. The other coastal subsections time series are characterized by four breakpoints. The slopes of associated trends ranged between $-127.93$ and $129.83{\text{m}}^3\cdot {\text{s}}^{-1}\cdot {\text{yr}}^{-1}$ (Coastal section 3 of Nigeria) over 1994-1998 and 2000-2007, respectively ( Figure 3 ). The temporal variability of river discharge is characterized by an increasing or a decreasing with unequal duration of the cumulative river discharge along each coastal section on the northern coast of the Gulf of Guinea. Negative trends followed by positive trends are observed along the coastal sections 1 (i.e., Cote d’Ivoire), 3 (Ghana) and Togo. This trend is reversed in the third section. Negative and positive trends alternate along the sections 2 (Côte d’Ivoire) and 1 - 3 (Nigeria). Only one trend has a weak positive slope value along the Benin coast and the coastal section 4 of Nigeria. Negative trend dominates the variability of river discharge over the 1993-2021 period. These trends are governed by precipitation dynamics [35] and human activities through dam building for agriculture and hydropower generation [36] [37] .

The long-term variability of river discharge from Cote d’Ivoire (except section 2) to Ghana is characterized by a positive slope ranging from $0.023{\text{m}}^3\cdot {\text{s}}^{-1}\cdot {\text{yr}}^{-1}$ (section 1 of Ghana coast) to $0.95{\text{m}}^3\cdot {\text{s}}^{-1}\cdot {\text{yr}}^{-1}$ (Togo). This means that river discharge increases over the period 1993-2021. A negative trend with slopes between $-2.96{\text{m}}^3\cdot {\text{s}}^{-1}\cdot {\text{yr}}^{-1}$ (section 3 of Nigeria coast) to $-0.021{\text{m}}^3\cdot {\text{s}}^{-1}\cdot {\text{yr}}^{-1}$ (Togo) characterized the long-term variability of the river discharge from Togo to Nigeria, and the coastal section 2 of Cote d’Ivoire.

Figure 4 represents the temporal and spatial change in sea level trend along the West African coastline over the study period. These are obtained by removing seasonal cycles and residuals in the original signal of the sea level anomaly time series. Different trends separated by breaks are observed in each coastal section. Five trends were noted in each coastal section from Cote d’Ivoire, Ghana, to Nigeria, except section 1 of Nigeria coast where four trends are recorded. All these trends are statistically significant at a 5% level according to the Mann-Kendall test and Sen’s slope estimation method. The minimum slope of −0.53 mm∙yr⁻¹ was observed from 2003 to 2009 in coastal section 3 of Nigeria and the maximum slope of 2.29 mm∙yr⁻¹ obtained in the coastal section 3 of Cote d’Ivoire is consistent with regional trend [3] .

The first breakpoint appeared in 1997 along all coastal sections. The associated trends increase from 1993 to 1997 with slopes ranging between 0.23 and 0.79 mm∙yr⁻¹ corresponding to the coastal sections 4 (Nigeria) and 1 (Cote d’Ivoire).

The second breakpoint was observed between 2000 and 2005 in all coastal sections except the coastal sections 2 - 3 of Cote d’Ivoire and the coastal section 1 of Ghana where a change in sea level regime occurred in 1998. An acceleration of the increasing trend with slopes between 0.52 and 2.06 mm∙yr⁻¹ was noted.

The third breakpoint occurred in 2015 from Cote d’Ivoire to Benin coastline, in 2016 and 2009 along the coastal section 1 and coastal section 2 - 4 of Nigeria, respectively. A weak positive trend with a slope not exceeding 0.14 mm∙yr⁻¹ characterize the temporal evolution of sea level trend in most of the coastal sections of West African coastline over the 1998-2015 period.

The temporal change is expressed by a negative trend with a slope of ~ −0.5 mm∙yr⁻¹ along the Nigeria coast.

The fourth breakpoint occurs mostly in 2016 along the coast, except in the coastal section 1 of Cote d’Ivoire, section 3 of Ghana and section 2 - 4 of Nigeria coast. Positive trends are noted over this fourth period and their slope can reach 2.29 mm∙yr⁻¹ as the case of coastal section 3 of Cote d’Ivoire.

During the 2018-2021 period, negative trends are observed in costal sections 1(Cote d’Ivoire), 3 (Ghana) and 2 - 4 (Nigeria). The slopes of these trends ranged between −0.45 and −0.35 mm∙yr⁻¹. One observes a weak and gradual increasing trend with slope reaching 0.08 mm∙yr⁻¹ along the other coastal sections.

Figure 5 shows the spatial and temporal evolution of normalized trends of ocean heat content (red line), river discharge (blue line) and sea level (black line) along the West Africa coastline. These were obtained by removing annual cycle and residual in original time series. Some similar changes are observed between ocean heat content and sea level in each coastal area. The temporal change between river

discharge and sea level is similar during the study period. During certain period, one can see a decreasing of river discharge and an increasing of sea level. This observation shows that ocean heat content has a strong impact on sea level change along the West Africa coastline. In this case, the possible contribution of river discharge to sea level rise could be assigned to the warming through coastal rivers discharge as the case of Niger river along the West Africa coastline [38] .

Previous results have shown the existing dependance between sea level change, river discharge and ocean heat content at 2000 m depth. This dependance could be expressed by the following equation :

$T_{SL}=0.73+0.02T_{RD}+0.62T_{OHC}+0.05T_{RD}{T}_{OHC}$

where:

$T_0=0.73$is a constant of the model;

$T_{SL}$: trend of normalized sea level anomaly data;

$T_{RD}$: trend normalized river discharge data;

$T_{OHC}$: trend of normalized ocean heat content data at 0 - 2000 m depth.

The parameters of this model are statistically significant because $\text{p}-\text{value}=3.19\times {10}^{-161}\ll 5\text{\%}$. This model could be used for sea level monitoring trend along the West African coastline.