La historia de la enseñanza en Egipto

In this subsection, we present the ensemble results from running the 1-D MEDUSA in station PAP, using the original MEDUSA parameter which differ to that using the modified parameter, described in 2.4. From the results, the default overestimates the chlorophyll and show an overall poor match with the in situ when the model is run using MEDUSA parameters. The poor match may be due to the higher uptake and grazing parameter from HadOCC.

Running the ensemble using MEDUSA parameters produces similar seasonality, with the highest mean surface chlorophyll occurring in May. However, all the satellite-derived seasonal and inter-annual surface chlorophyll means are within the ensemble range, summarised in Fig 3.9(a) and (b), unlike that when modified parameters were used. This results in a broader ensemble spread, as indicated by the NRR values, summarised in Table 3.3, that are mostly closer to reliable spreads. The NRR value for the annual mean is broader than the modified parameters. The default run is also closer to the satellite- derived chlorophyll concentration, especially when it declines between July to October.

3.1. Abyssal Plain 85

Figure 3.4: 10-year monthly mean surface chlorophyll from all the study sites ((a)-(e)), showing the seasonal dynamics of surface chlorophyll (mg m−3). The boxplots show the ensemble seasonal means. Blue cross is the in situ observation, red open circle, black dot, and blue stars are the ensemble mean, median, and the default run respectively. The blue box is the 75th (top) and 25th (bottom) quartiles. The red line is the median. The whiskers are the ensemble minimum and maximum mean of surface chlorophyll. In station PAP, in situ data for December is not available due to low light and high cloud cover.

of DIN concentration occurring between February to August is not captured and therefore making the ensemble overestimates in situ monthly means. The DIN produced by ensemble only starts to decline between July to September, instead of February to March. This makes the NRR value for DIN profile higher (1.33) than that running the model using the parameters described in Table 1.

Table 3.3 summarised the statistical metrics when running the ensemble using MEDUSA parameters in station PAP. The default run produces better RMSE for DIN, although the correlation coefficient is higher in the ensemble median, similar to the modified parameters run. However, the RMSE and bias for the default run, ensemble mean, and median in DIN are generally higher compared to the modified parameters. Running the ensemble using MEDUSA parameters enhance the correlation coefficient for chlorophyll, both on the surface and over the depths, as well as reducing the RMSEs and bias for the ensemble mean, median, and default run. These improvements, therefore, making the NRR values for chlorophyll profile and surface closer to unity compared to using the modified parameter.

Figure 3.5: Inter-annual variability of DIN averaged over 200 m, from all the study sites ((a)-(e)), and the annual mean (f). Since the in situ data for PAP does not always cover the first 200m, the overall mean DIN concentration from all depth is used instead. For station L4, in situ DIN is only collected on the surface. Blue cross is the in situ observation, red open circle, black dot, and blue stars are the ensemble mean, median, and default run respectively. The blue box is the 75th(top) and 25th(bottom) quartiles. Red line is the median, and the whiskers are the ensmeble minimum and maximum of the averaged DIN. In station L4 and PAP data for DIN is only available from 2000-2007 and 2002-2004 respectively.

rate is twice higher (2.0 day−1), compared to that in the modified parameter. This will cause lower phy- toplankton abundance, and higher DIN concentrations as shown in Figure 3.9 and from the 10-year mean shown in Table 3.3, especially in surface chlorophyll, where the mean is reduced to 60%, resulting in smaller bias. However, the effect of structural sensitivity is still quite similar to when using the modified parameter, where higher chlorophyll is produced when G1is applied. In terms of DIN, apart from apply-

ing G1, low concentration is produced, when the functional form is combined with ρlζq, ρhζl, ρqζq, and

ρhζs. These functional form combinations coincide with low RMSEs. However, higher DIN concentra-

tion > 9.5 mmol m−3, is observed when the ensemble members contained G2, unlike that in the modified

parameter whereby high DIN is only produced when combining U4with G2. In terms of surface chloro-

phyll, lower concentrations when using functional forms that contain U4and U3have not been observed.

Instead, high chlorophyll concentrations and hence RMSE, both in the surface and profile, is produced when the functional forms that contain ρlζq and combining G1 with ρhζl. These are summarised in

Figure 3.10.

3.1. Abyssal Plain 87

Figure 3.6: 10-year monthly mean of DIN averaged over 200 m from all the study sites ((a)-(e)), showing the seasonal dynamics of DIN (mmol m−3). For station PAP, the DIN shown is the overall profile, and in L4, the in situ DIN concentration is only available at the surface. The boxplot shows the ensemble monthly means. Blue cross is the in situ observation, red open circle, black dot, and blue stars are the ensemble mean, median, and the default run respectively. The blue box is the 75th(top) and 25th(bottom) quartiles. The red line is the median. The whiskers are the ensemble minimum and maximum mean of averaged DIN. In station PAP, the in situ data is only collected from 2002-2004 and L4 from 2000-2007.

the default run, ensemble mean, and median underestimate the peak, shown in Figure 3.11(a). However for amplitude, the observation is within the interquartile range, although the ensemble mean, median, and the default run still underestimating the amplitude, unlike that in the original run. The observation of bloom initiation is still within the interquartile range, with the ensemble median showing similar timing as the observation (both in 20th April). The ensemble captures the observation on other metrics, such as the timing of the bloom peak, duration and termination, shown in Figure 3.11(b) and (c). This might be due to the broader range of ensemble that is produced from using the MEDUSA parameter instead of the modified parameter ensemble. For example, the initiation timing from the MEDUSA parameter ensemble occurs between March to early July, whereas the modified parameter range is only between March to early May. The most striking difference is in termination, where the MEDUSA parameter ensemble produces termination range between late May to November, and the modified parameter only covers termination time between June to October.

Indeed, using the MEDUSA parameters have improved the NRR values, RMSE, and correlation coef- ficients for surface chlorophyll and profile, but not DIN. The RMSE and bias for DIN are higher com- pared to using the modified parameters, as well as its NRR, indicating a narrow ensemble spread. Bias in

Figure 3.7: Annual mean of surface chlorophyll when changing only one process at a time (blue box), overlain with annual mean of all ensemble members (green box) at five oceanographic stations. Ensemble mean and median plotted in the figure (shown in red open circle and black closed circle), are the from the 128 ensemble members.

chlorophyll is also higher than that in the original run using modified parameters. Nonetheless, G2 still

produced lower chlorophyll concentrations. The mortality functions that produced high chlorophyll and low DIN are also similar to that in the modified parameter.