Results
Our study represented 60 weeks of continuous survey of 132 bushmeat sellers. The largest market was Yopougon Siporex (26 sellers before the COVID lockdown), whereas the smallest market was Abobo (Grand Marché; four sellers). From the two-sample Kolmogorov-Smirnov tests, all curves differed significantly from each other for the entire study period (p < 0.05). Fréchet distances also showed a certain level of heterogeneity among bushmeat sites, with Yopougon Siporex and Abobo Grand Marché curves having the greatest distance between each other (distance = 625.85) across the whole study period (Fig. 2).
The number of sellers was stable before lockdown enforcement and started decreasing more or less abruptly from the first week of April in all the bushmeat sites, except in Toumodi restaurants. This was mirrored by the changepoint analysis, as the biggest change in mean for all trends was at the first week of April 2020, except the Toumodi restaurants that had their most important change at the third week of March 2020, i.e. right after the announcement of the ban on bushmeat consumption (Fig. 1, Appendix Table 1). Bushmeat site re-opening was first visible in Abobo Grand Marché and Toumodi (June 2020), before the end of state of health emergency (30 June 2020). Adjamé started reopening in July 2020 (concomitantly with end of lockdown in Abidjan), whereas Abobo Mairie (September 2020) and Yopougon Siporex (October 2020) reopened later, once sanitary measures were lifted. Growth curve analysis suggest that Toumodi restaurants saw the most rapid, but non-significant, growth rate (r = 0.927 ± 0.67, p = 0.2), followed by Yopougon Siporex market (r = 0.508 ± 0.143, p = 0.001). Adjamé market had the slowest growth rate (r = 0.105 ± 0.043, p = 0.02), followed by Abobo Grand Marché market (r = 0.128 ± 0.072, p = 0.08). Toumodi restaurants, Abobo Grand Marché and Adjamé markets had relatively high levels of standard error in the growth rate estimates.
During the period under governmental measures, two-sample Kolmogorov-Smirnov tests indicated that Toumodi restaurants significantly differed from the combined bushmeat markets (D = 0.57,p -value < 0.001), with the markets being completely closed following the first months of lockdown and the restaurants only closing for a period c. two months and a half. This period saw significant differences across most comparisons, except between the Toumodi restaurants and Adjamé market (D= 0.2, p -value= 0.379), as well as Yopougon Siporex and Abobo Mairie markets (D= 0.14,p -value= 0.524) trends. Fréchet distances suggested Yopougon Siporex market and Toumodi restaurants were the most divergent (distance = 130.88), while Yopougon Siporex and Abobo Mairie markets were the most similar (distance = 30.42; Fig 2).
In January 2021 (end of the survey), all the bushmeat sites exhibited lower numbers of sellers than before lockdown, whilst the control survey in September 2021 yielded the same result with the exception of Abobo (same as before lockdown) and Toumodi (higher than before lockdown). In all cases and models (with or without constraint), forecasting predictions after 76 weeks since the beginning of lockdown were inferior to the initial numbers of sellers (Fig. 3), with unconstrained models yielding lower estimations. Both models yielded predictions of number of sellers inferior to observed values after 76 weeks for Abobo Mairie and Grand Marché markets, and Toumodi restaurants. The number of sellers in Adjamé market was over-predicted by both models, while for Yopougon Siporex market the unconstrained model yielded inferior predictions and the constrained model higher predictions relative to the observed number. The two models yielded the most dissimilar predictions for Adjamé market (unconstrained model yielded simulated value closer to observed value) and Toumodi restaurants (constrained model yielded simulated value closer to observed value). Toumodi predictive models had the highest averaged RMSE values (0.47) and lowest proportion of variance explained (50.47%), while Yopougon Siporex had the lowest averaged RMSE (0.16) with high levels of variance explained (92.51%) (Table 1).