In the literature, several epidemiological studies have already associated respiratory and cardiovascular diseases to acute exposure of mineral dust. However, frail people are also sensitive to chronic exposure to particulate matter with an aerodynamic diameter $10\mathrm{\mu }\mathrm{m}$ or less ($PM10$). Consequently, it is crucial to better understand $PM10$ fluctuations at all scales. This study investigates $PM10$ background atmosphere in the Caribbean area according to African dust seasonality with complex network framework. For that purpose, the regular Visibility Graph (VG) and the new Upside-Down Visibility Graph (UDVG) are used for a multifractal analysis. Firstly, concentration vs degree (v-k) plots highlighted that high degree values (hubs behavior) are related to the highest $PM10$ concentrations in VG while hubs is associated to the lowest concentrations in UDVG, i.e. probably the background atmosphere. Then, the degree distribution analysis showed that VG and UDVG difference is reduced for high dust season contrary to the low one. As regards the multifractal analysis, the multifractal degree is higher for the low season in VG while it is higher for the high season in UDVG. The degree distribution behavior and the opposite trend in multifractal degree for UDVG are due to the increase of $PM10$ background atmosphere during the high season, i.e. from May to September. To sum up, UDGV is an efficient tool to perform noise fluctuations analysis in environmental time series where low concentrations play an important role as well.

在文献中，一些流行病学研究已经将呼吸道和心血管疾病与矿物粉尘的急性暴露相关联。但是，体弱的人也对长期暴露于空气动力学直径的颗粒物敏感。$10\mathrm{\mu }\mathrm{米}$ 或更少 （$P\mathrm{中号}10$）。因此，更好地理解至关重要$P\mathrm{中号}10$各种规模的波动。这项研究调查$P\mathrm{中号}10$根据非洲沙尘季节和复杂网络框架，加勒比地区的背景大气。为此，常规可见性图（VG）和新的上下可见性图（UDVG）用于多重分形分析。首先，浓度与度数（vk）曲线图突出表明，高度值（轮毂行为）与最高$P\mathrm{中号}10$VG中的浓度较高，而枢纽与UDVG中的最低浓度相关，即可能是背景大气。然后，度数分布分析表明，高沙尘季节的VG和UDVG差异减小，而低沙尘季节的VG和UDVG差异减小。关于多重分形分析，VG的淡季多重分形度较高，而UDVG的旺季则较高。UDVG的度分布行为和多重分形度的相反趋势是由于$P\mathrm{中号}10$旺季（即五月至九月）的背景气氛。综上所述，UDGV是在低浓度也起着重要作用的环境时间序列中执行噪声波动分析的有效工具。