Intervention analysis with autoregressive integrated moving average models for time series of urban air pollutants

The objective of this paper is to show an intervention analysis with autoregressive integrated moving average models for time series of air pollutants in a Latin American megacity. The interventions considered in this study correspond to public regulations for the control of urban air quality. The study period comprised 10 years. Information from 10 monitoring stations distributed throughout the megacity was used. Modelling showed that setting maximum emission limits for different pollution sources and improving fuel were the most appropriate regulatory interventions to reduce air pollutant concentrations. Modelling results also suggested that these interventions began to be effective between the first 4 days-15 days after their publication. The models developed on a monthly timescale had a short autoregressive memory. The air pollutant concentrations at a given time were influenced by the concentrations of up to three months immediately preceding. Moving average term of the models showed fluctuations in time of the air pollutant concentrations (3 months - 14 months). Within the framework of the applications of physics for the air pollution control, this study is relevant for the following findings: the usefulness of autoregressive integrated moving average models to temporal simulate air pollutants, and for its suitable performance to detect and quantify regulatory interventions.

Deterministic Chaos Detection and Simplicial Local Predictions Applied to Strawberry Production Time Series

In this work, we attempted to find a non-linear dependency in the time series of strawberry production in Huelva (Spain) using a procedure based on metric tests measuring chaos. This study aims to develop a novel method for yield prediction. To do this, we study the system’s sensitivity to initial conditions (exponential growth of the errors) using the maximal Lyapunov exponent. To check the soundness of its computation on non-stationary and not excessively long time series, we employed the method of over-embedding, apart from repeating the computation with parts of the transformed time series. We determine the existence of deterministic chaos, and we conclude that non-linear techniques from chaos theory are better suited to describe the data than linear techniques such as the ARIMA (autoregressive integrated moving average) or SARIMA (seasonal autoregressive moving average) models. We proceed to predict short-term strawberry production using Lorenz’s Analog Method.

Modelling Covid-19 Pandemic in Nigeria using Multivariate Autoregressive Distributed Lag-Moving Average Models

The aim of this paper was to study the trend of COVID-19 cases and fit appropriate multivariate time series models as research to complement the clinical and non-clinical measures against the menace. The cases of COVID-19, as reported by the National Centre for Disease Control (NCDC) on a daily and weekly basis, include Total Cases (TC), New Cases (NC), Active Cases (AC), Discharged Cases (DC) and Total Deaths (TD). The three waves of the COVID-19 pandemic are graphically represented in the various time plots, indicating the peaks as (June–August, 2020), (December–February, 2021), and (July–September, 2021). Multivariate Autoregressive Distributed Lag Models (MARDLM) and Multivariate Autoregressive Distributed Lag Moving Average (MARDL-MA) models have been found to be suitable for fitting different categories of the COVID-19 pandemic in Nigeria. The graphical representation and estimates have shown a gradual decline in the reported cases after the peak in September 2021. So far, the introduction of vaccines and non-pharmaceutical measures by relevant organisations are yielding plausible results, as evident in the recent decrease in New Cases, Active Cases and an increasing number of Discharged Cases, with fewer deaths. This paper advocates consistency in all clinical and non-clinical measures as a way towards the extinction of the dreaded COVID-19 pandemic in Nigeria and the world.

Models of predictive modeling on the example of a gas turbine

The ability to accurately predict the operation of a particular mechanism and, on the basis of this, estimate the equipment life is a very important task. The amount of losses of the enterprise can depend on such study, as well as the health of many people whose lives depend on the health of the working installation. As part of this work, the main time series models were considered and the most suitable for the study was selected. The operation of a gas turbine was studied and a forecast was made. On the basis of the study, linear regression, ARIMA and moving average models were built and evaluated.

Temporal analysis of the heavy metal concentration in road sediment and dust using statistical models

The objective of this paper is to show a temporal analysis using autoregressive integrated moving average models of the heavy metal concentration in road sediment and dust of Soacha, a Colombian locality. The representative size fractions in the road sediment and dust were <250 μm and ⩽10 μm, respectively. The results suggest that lead is the best metallic element to study the relationship between the heavy metal concentration in the road sediment and dust (r-Pearson = 0.90). Univariate models (R2 ⩾ 0.58) suggest that the time series of lead concentrations in road sediment and dust have the same temporal structure. Namely, because they are first-order autoregressive processes, concentrations at a given moment of time are influenced by the immediately preceding concentrations. The transfer function model (R2 = 0.91) suggests that there is no delay in impulse transfer from road dust concentration to lead concentration in the road sediment. The effect is immediate for a sampling interval of 3 days. The results show that modeling has a better fit during the rainy season compared to the dry season. In the context of the simulation of physical phenomena in engineering, this study is relevant to deepen knowledge in relation to the use of autoregressive integrated moving average models.

Analysis and forecasts for trends of COVID-19 in Pakistan using Bayesian models

Background COVID-19 is currently on full flow in Pakistan. Given the health facilities in the country, there are serious threats in the upcoming months which could be very testing for all the stakeholders. Therefore, there is a need to analyze and forecast the trends of COVID-19 in Pakistan. Methods We have analyzed and forecasted the patterns of this pandemic in the country, for next 30 days, using Bayesian structural time series models. The causal impacts of lifting lockdown have also been investigated using intervention analysis under Bayesian structural time series models. The forecasting accuracy of the proposed models has been compared with frequently used autoregressive integrated moving average models. The validity of the proposed model has been investigated using similar datasets from neighboring countries including Iran and India. Results We observed the improved forecasting accuracy of Bayesian structural time series models as compared to frequently used autoregressive integrated moving average models. As far as the forecasts are concerned, on August 10, 2020, the country is expected to have 333,308 positive cases with 95% prediction interval [275,034–391,077]. Similarly, the number of deaths in the country is expected to reach 7,187 [5,978–8,390] and recoveries may grow to 279,602 [208,420–295,740]. The lifting of lockdown has caused an absolute increase of 98,768 confirmed cases with 95% interval [85,544–111,018], during the post-lockdown period. The positive aspect of the forecasts is that the number of active cases is expected to decrease to 63,706 [18,614–95,337], on August 10, 2020. This is the time for the concerned authorities to further restrict the active cases so that the recession of the outbreak continues in the next month.