Statistical modeling approach for PM₁₀ prediction before and during confinement by COVID-19 in South Lima, Perú

A total of 188,859 meteorological-PM₁₀ data validated before (2019) and during the COVID-19 pandemic (2020) were used. In order to predict PM₁₀ in two districts of South Lima in Peru, hourly, daily, monthly and seasonal variations of the data were analyzed. Principal Component Analysis (PCA) and linear/nonlinear modeling were applied. The results showed the highest annual average PM₁₀ for San Juan de Miraflores (SJM) (PM₁₀-SJM: 78.7 μ g/m³) and the lowest in Santiago de Surco (SS) (PM₁₀-SS: 40.2 μ g/m³). The PCA showed the influence of relative humidity (RH)-atmospheric pressure (AP)-temperature (T)/dew point (DP)-wind speed (WS)-wind direction (WD) combinations. Cool months with higher humidity and atmospheric instability decreased PM₁₀ values in SJM and warm months increased it, favored by thermal inversion (TI). Dust resuspension, vehicular transport and stationary sources contributed more PM₁₀ at peak times in the morning and evening. The Multiple linear regression (MLR) showed the best correlation (r = 0.6166), followed by the three-dimensional model LogAP-LogWD-LogPM₁₀ (r = 0.5753); the RMSE-MLR (12.92) exceeded that found in the 3D models (RMSE < 0.3) and the NSE-MLR criterion (0.3804) was acceptable. PM₁₀ prediction was modeled using the algorithmic approach in any scenario to optimize urban management decisions in times of pandemic.