Congo Basin Hydrology, Climate, and Biogeochemistry. Группа авторов

Чтение книги онлайн.

СКАЧАТЬ upper M a x left-parenthesis left-parenthesis upper X Subscript upper G upper C upper M Baseline minus upper X Subscript upper O upper B upper S Baseline right-parenthesis squared comma left-parenthesis upper X Subscript upper R upper C upper M Baseline minus upper X Subscript upper O upper B upper S Baseline right-parenthesis squared right-parenthesis EndFraction"/>

      The AV calculation is done between the general circulation model (GCM) used as a boundary condition of RegCM4, a reference observation (OBS), and the regional climate model (RCM), where X denotes the spatial distribution of the data set under consideration. Defined in this way, the experiment with the highest AV relative to the GCM will perform best.

      In this study, the analysis of spatial distributions was carried out during four seasons of the year defined as follows: December–January–February (DJF), March–April–May (MAM), June–July–August (JJA), and September–October–November (SON). In terms of Taylor diagram and seasonal cycle analyses, the focus was on four sub‐regions of the Congo Basin identified by Balas et al. (2007) as being relatively homogeneous in terms of interannual rainfall variability caused by SST fluctuations. These sub‐regions are North Central Africa (NCA) and South‐Central Africa (SCA) with a monomodal rainfall regime; and West Equatorial (WE) and East Equatorial (EE) with a bimodal rainfall regime.

4.3. RESULTS AND DISCUSSIONS

      4.3.1. Average Seasonal Climatology of Temperature and Rainfall

       Average Seasonal Temperature Climatology

Figures 4.2 and 4.3 show, respectively, the spatial distribution and biases of the surface temperature for the reanalysis data (ERA 5 and ERA‐Interim) and the two experiments carried out with the RegCM model over the period from 2001 to 2006. Reanalysis data (Figure 4.2 a–h) show that the spatial distribution of temperature is strongly influenced by relief (Kamga, 2001). Thus, the highest temperatures (> 32 °C) are located in the northern part of the domain while the lowest temperatures (< 18 °C) are observed along the coasts of the Atlantic and Indian Oceans, and also around mountainous regions. Biases of about 3 °C are reported between the two reanalysis products at high relief areas and land‐ocean boundaries (Figure 4.3 a–d), highlighting the presence of uncertainties between the two data sets. In comparison with the reanalysis data, RegCM’s simulations satisfactorily locate the zones of temperature maxima and minima both on the continent and in the oceans. However, as the panels in Figure 4.3 e–l show, the maxima (minima) are overestimated (underestimated) by the model in both experiments. Nevertheless, we note that experiments with slab‐ocean present more extensive biases over the Atlantic Ocean, but these are less extended over the continent and almost zero over the Indian Ocean. The reduction or cancellation of the range of biases noted above can be attributed to the presence of the slab‐ocean, which interacts with the atmosphere, forcing it to calculate realistic SST. We also note that the experience with slab‐ocean has great added value compared to that without slab‐ocean.

Figure 4.2 Seasonal means of the surface air temperature (T2m °C) for the period 2001–2006: ERA 5 (a–d), ERA‐Interim (e–g), GFC_SLAB (i–l), GFC_CTR (m–p).

Figure 4.3 Bias of mean seasonal surface air temperature (T2m °C) for the period 2001–2006: ERA‐Interim ‐ ERA 5 (a–d), GFC_SLAB ‐ ERA5 (e–g), GFC_CTR ‐ ERA 5 (i–l), GFC_SLAB ‐ GFC_CTR (m–p).

       Average Seasonal Rainfall Climatology

In Figure 4.4, we make an intercomparison of the climatology of seasonal precipitation from the different experiments carried out and the observational data (ARC2 and GPCP) for the period from 2001 to 2006. Observations (Figure 4.4 a–h) show that bands of high rainfall (11–12 mm/day) are visible in all seasons. The alternation of dry and wet seasons is well represented. We observe a spatial difference between the observed data (Figure 4.5 a–d); this difference can be interpreted as the uncertainties (errors) in the production of these data (Nikulin et al., 2012). Although some biases persist, the RegCM experiments (Figure 4.5 i–p) reproduce the rainfall patterns in the study area fairly well according to observations during all seasons. The experiments carried out successfully capture the positions of the observed rainfall maxima, but fail to represent their amplitudes. We note that both experiments tend to underestimate (up to 5 mm/day) the rainfall band observed on the continent, except in mountainous regions and across the oceans where the opposite effect is observed; this difference could be attributed to a thermal breeze, surface parameterization, and the Foehn effect. Despite the large biases obtained in the two RegCM simulation experiments, those of the slab‐ocean experiments seem to be significantly reduced or cancelled in the Indian Ocean compared to the Atlantic Ocean (Figure 4.5 e–p). This implies that contrary to the majority of the RCMs used in Nikulin et al. (2012), which systematically overestimate rainfall over the Indian Ocean, the atmosphere–ocean interactions integrated into this study take into account the dynamics that govern the climate in the Indian Ocean.

Figure 4.4 Seasonal means of rainfall in mm/day for the period 2001–2006: GPCP (a–d), ARC2 (e–h), GFC_CTR (i–l), GFC_SLAB (m–p).

Figure 4.5 Seasonal mean biases of precipitation in mm/day for the period 2001–2006: ARC2 ‐ GPCP (a–d), GFC_SLAB ‐ GPCP (e–h), GFC_CTR ‐ GPCP (i–l), GFC_SLAB ‐ GFC_CTR (m–p).

      4.3.2. Taylor Diagram

We assessed the interannual variability and degree of similarity between the observed and model‐simulated temperature/rainfall using the Taylor diagram (Figure 4.6). It emerges that the experiments reproduce СКАЧАТЬ