Air Quality Time Series Research Articles

MAML-Enhanced LSTM for Air Quality Time Series Forecasting

MAML-Enhanced LSTM for Air Quality Time Series Forecasting

Simulated Surface‐Column NO2 Connections for Satellite Applications

AbstractObservations of near‐surface NO2 show a diurnal pattern with midday minima and daily maxima in the morning and evening. These surface cycles are dependent on chemical processing, transport, and emissions. We evaluate these cycles with the United States Environmental Protection Agency (EPA) community multiscale air quality (CMAQ) model data from the EPA Air Quality Time Series (EQUATES) project, compared with ground‐based measurements from the EPA air quality system, two Pandora measurement sites, and satellite data from TROPOMI. We find that the morning vertical column density (VCD) lags surface concentrations by 1 hr on average, where this lag varies with location and day. The peak VCD can also lead the surface maximum concentration, especially in the evening, responding to transport and afternoon compression of the boundary layer. Modeled NO2 VCD is sensitive to column calculation technique. With hourly daytime satellite‐based NO2 observations newly available from the TEMPO instrument, the timing and magnitude of cycles in near‐surface NO2 versus column NO2 will help inform the utilization of hourly satellite data. This work will help inform the timing of surface‐column connections to better interpret new hourly satellite observations for health and air quality applications, including emissions characterization.

Short-term air quality prediction based on EMD-transformer-BiLSTM

Actual acquired air quality time series data are highly volatile and nonstationary, and accurately predicting nonlinear time series data containing complex noise is an ongoing challenge. This paper proposes an air quality prediction method based on empirical mode decomposition (EMD), a transformer and a bidirectional long short-term memory neural network (BiLSTM), which is good at addressing the ultrashort-term prediction of nonlinear time-series data and shows good performance for application to the air quality dataset of Patna, India (6:00 am on October 3, 2015–0:00 pm on July 1, 2020). The AQI sequence is first decomposed into intrinsic mode functions (IMFs) via EMD and subsequently predicted separately via the improved transformer algorithm based on BiLSTM, where linear prediction is performed for IMFs with simple trends. Finally, the predicted values of each IMF are integrated using BiLSTM to obtain the predicted AQI values. This paper predicts the AQI in Patna with a time window of 5 h, and the RMSE, MAE and MAPE are as low as 5.6853, 2.8230 and 2.23%, respectively. Moreover, the scalability of the proposed model is validated on air quality datasets from several other cities, and the results prove that the proposed hybrid model has high performance and broad application prospects in real-time air quality prediction.

Impact of Heatwaves and Declining NOx on Nocturnal Monoterpene Oxidation in the Urban Southeastern United States.

Nighttime oxidation of monoterpenes (MT) via the nitrate radical (NO3) and ozone (O3) contributes to the formation of secondary organic aerosol (SOA). This study uses observations in Atlanta, Georgia from 2011-2022 to quantify trends in nighttime production of NO3 (PNO3) and O3 concentrations and compare to model outputs from the EPA's Air QUAlity TimE Series Project (EQUATES). We present urban-suburban gradients in nighttime NO3 and O3 concentrations and quantify their fractional importance (F) for MT oxidation. Both observations and EQUATES show a decline in PNO3, with modeled PNO3 declining faster than observations. Despite decreasing PNO3, we find that NO3 continues to dominate nocturnal boundary layer (NBL) MT oxidation (FNO3 = 60%) in 2017, 2021, and 2022, which is consistent with EQUATES (FNO3 = 80%) from 2013-2019. This contrasts an anticipated decline in FNO3 based on prior observations in the nighttime residual layer, where O3 is the dominant oxidant. Using two case studies of heatwaves in summer 2022, we show that extreme heat events can increase NO3 concentrations and FNO3, leading to short MT lifetimes (<1 h) and high gas-phase organic nitrate production. Regardless of the presence of heatwaves, our findings suggest sustained organic nitrate aerosol formation in the urban SE US under declining NOx emissions, and highlight the need for improved representation of extreme heat events in chemistry-transport models and additional observations along urban to rural gradients.

MGSFformer: A Multi-Granularity Spatiotemporal Fusion Transformer for air quality prediction

Air quality spatiotemporal prediction can provide technical support for environmental governance and sustainable city development. As a classic multi-source spatiotemporal data, effective multi-source information fusion is key to achieving accurate air quality predictions. However, due to not fully fusing two pieces of information, classical deep learning models struggle to achieve satisfactory prediction results: (1) Multi-granularity: each air monitoring station collects air quality data at different sampling intervals, which show distinct time series patterns. (2) Spatiotemporal correlation: due to human activities and atmospheric diffusion, there exist correlations between air quality data from different air monitoring stations, necessitating the consideration of other air monitoring stations' influences when modeling each air quality time series. In this study, to achieve satisfactory prediction results, we propose the Multi-Granularity Spatiotemporal Fusion Transformer, comprised of the residual de-redundant block, spatiotemporal attention block, and dynamic fusion block. Specifically, the residual de-redundant block eliminates information redundancy between data with different granularities and prevents the model from being misled by redundant information. The spatiotemporal attention block captures the spatiotemporal correlation of air quality data and facilitates prediction modeling. The dynamic fusion block evaluates the importance of data with different granularities and integrates the prediction results. Experimental results demonstrate that the proposed model surpasses 11 baselines by 5% in performance on three real-world datasets.

Diverging trends in aerosol sulfate and nitrate measured in the remote North Atlantic in Barbados are attributed to clean air policies, African smoke, and anthropogenic emissions.

Sulfate and nitrate aerosols degrade air quality, modulate radiative forcing and the hydrological cycle, and affect biogeochemical cycles, yet their global cycles are poorly understood. Here, we examined trends in 21 years of aerosol measurements made at Ragged Point, Barbados, the easternmost promontory on the island located in the eastern Caribbean Basin. Though the site has historically been used to characterize African dust transport, here we focused on changes in nitrate and non-sea-salt (nss) sulfate aerosols from 1990-2011. Nitrate aerosol concentrations averaged over the entire period were stable at 0.59 μg m-3 ± 0.04 μg m-3, except for elevated nitrate concentrations in the spring of 2010 and during the summer and fall of 2008 due to the transport of biomass burning emissions from both northern and southern Africa to our site. In contrast, from 1990 to 2000, nss-sulfate decreased 30% at a rate of 0.023 μg m-3 yr-1, a trend which we attribute to air quality policies enacted in the United States (US) and Europe. From 2000-2011, sulfate gradually increased at a rate of 0.021 μg m-3 yr-1 to pre-1990s levels of 0.90 μg m-3. We used the Community Multiscale Air Quality (CMAQ) model simulations from the EPA's Air QUAlity TimE Series (EQUATES) to better understand the changes in nss-sulfate after 2000. The model simulations estimate that increases in anthropogenic emissions from Africa explain the increase in nss-sulfate observed in Barbados. Our results highlight the need to better constrain emissions from developing countries and to assess their impact on aerosol burdens in remote source regions.

Enhancing Air Quality Forecasting: A Novel Spatio-Temporal Model Integrating Graph Convolution and Multi-Head Attention Mechanism

Forecasting air quality plays a crucial role in preventing and controlling air pollution. It is particularly significant for improving preparedness for heavily polluted weather conditions and ensuring the health and safety of the population. In this study, a novel deep learning model for predicting air quality spatio-temporal variations is introduced. The model, named graph long short-term memory with multi-head attention (GLSTMMA), is designed to capture the temporal patterns and spatial relationships within multivariate time series data related to air quality. The GLSTMMA model utilizes a hybrid neural network architecture to effectively learn the complex dependencies and correlations present in the data. The extraction of spatial features related to air quality involves the utilization of a graph convolutional network (GCN) to collect air quality data based on the geographical distribution of monitoring sites. The resulting graph structure is imported into a long short-term memory (LSTM) network to establish a Graph LSTM unit, facilitating the extraction of temporal dependencies in air quality. Leveraging a Graph LSTM unit, an encoder-multiple-attention decoder framework is formulated to enable a more profound and efficient exploration of spatio-temporal correlation features within air quality time series data. The research utilizes the 2019–2021 multi-source air quality dataset of Qinghai Province for experimental assessment. The results indicate that the model effectively leverages the impact of multi-source data, resulting in optimal accuracy in predicting six air pollutants.

Revisiting day-of-week ozone patterns in an era of evolving US air quality

Abstract. Past work has shown that traffic patterns in the USA and resulting NOx emissions vary by day of week, with NOx emissions typically being higher on weekdays than weekends. This pattern of emissions leads to different levels of ozone on weekends versus weekdays and can be leveraged to understand how local ozone formation changes in response to NOx emission perturbations in different urban areas. Specifically, areas with lower NOx but higher ozone on the weekends (the weekend effect) can be characterized as NOx-saturated and areas with both lower NOx and ozone on weekends (the weekday effect) can be characterized as NOx-limited. In this analysis, we assess maximum daily 8 h average (MDA8) ozone weekend–weekday differences across 51 USA nonattainment areas using 18 years of observed and modeled data from 2002–2019, using the following two metrics: mean MDA8 ozone and percentage of days with MDA8 ozone &gt; 70 ppb (parts per billion). In addition, we quantify the modeled and observed trends in these weekend–weekday differences across this period of substantial NOx emission reductions in the USA. The model assessment is carried out using U.S. Environmental Protection Agency (EPA)'s Air QUAlity TimE Series Project (EQUATES) Community Multiscale Air Quality (CMAQ) model dataset. We identify three types of MDA8 ozone trends occurring across the USA, namely transitioning chemical regime, disappearing weekday effect, and no trend. The transitioning chemical regime trend occurs in a subset of large urban areas that were NOx-saturated (i.e., volatile organic compound (VOC)-limited) at the beginning of the analysis period but transitioned to mixed chemical regimes or NOx-limited conditions by the end of the analysis period. Nine areas have strong transitioning chemical regime trends using both modeled and observed data and with both metrics indicating strong agreement that they are shifting to more NOx-limited conditions: Milwaukee, Houston, Phoenix, Denver, the Northern Wasatch Front, the Southern Wasatch Front, Las Vegas, Los Angeles – San Bernardino County, Los Angeles – South Coast, and San Diego. The disappearing weekday effect was identified for multiple rural and agricultural areas of California which were NOx-limited for the entire analysis period but appear to become less influenced by local day-of-week emission patterns in more recent years. Finally, we discuss a variety of reasons why there are no trends in certain areas including complex impacts of heterogeneous source mixes and stochastic impacts of meteorology. Overall, this assessment finds that the EQUATES modeling simulations indicate more NOx-saturated conditions than the observations but do a good job of capturing year-to-year changes in weekend–weekday MDA8 ozone patterns.

Unusual response of O3 and CH4 to NO2 emissions reduction in Japan during the COVID-19 pandemic

ABSTRACT We investigate the impact of NO2 emissions reduction on O3 and CH4 levels in 14 metropolitan areas of Japan in 2020. To account for meteorological variations, we employ business-as-usual air quality time series generated by machine learning models. Additionally, we use satellite observations and biogeochemical model simulations to analyse air quality changes. During the lockdown period from April 7 to May 25 in 2020, we observed a NO2 reduction that equated to a decrease equivalent to 3.4 and 5 years of the corresponding trends in roadside and ambient air quality recorded from 2010 to 2019. After meteorological normalization, NO2 decreased by 14.5% at ambient air stations and 19.1% at roadside stations. Surprisingly, the NO2 reduction did not immediately lead to increased O3. Instead, O3 levels rose after the lockdown, specifically in August due to favorable sunny conditions. This finding is important for Japan and has not been reported in previous studies. We found that changes in NO2 and CO marginally contributed to variations in CH4 levels across the study areas. It is recommended to simultaneously reduce NOx as well as non-methane volatile organic compounds to mitigate their adverse effects on future policies.

Detection and Predicting Air Pollution Level in a Specific City using Deep Learning

Air pollution affects millions of people worldwide, making it a growing issue. Deep learning can identify and forecast metropolitan air pollution. Deep learning needs a massive dataset of air quality measurements and meteorological factors to predict city air pollution levels. Government monitoring stations and citizen scientific programs collect this data. Once we have our dataset, we can apply deep learning to develop a model that predicts air pollution levels. Temperature, humidity, wind speed, and air quality data will be used to predict future air pollution levels. Predicting air pollution using the LSTM network is popular. This neural network works well with air quality time-series data. The LSTM network's long-term data learning is essential for accurate air pollution predictions. We would pre-process our data to prepare it for an LSTM network to predict air pollution. Scaling, splitting, and encoding data may be needed. Train the LSTM network using backpropagation and gradient descent on our dataset. Adjusting the network's weights and biases would lessen the air pollution gap. After training, the network can predict city air quality. Inputting current meteorological and environmental factors may help accomplish this aim and deliver timely predictions. Deep learning can detect and predict urban air pollution. LSTM neural network algorithms may accurately forecast complex air quality data patterns, providing vital information about our planet's health.

Evaluating Data Product Exposure Metrics for Use in Epidemiologic Studies of Dust Storms.

Dust storms are increasing in frequency and correlate with adverse health outcomes but remain understudied in the United States (U.S.), partially due to the limited spatio-temporal coverage, resolution, and accuracy of current data sets. In this work, dust-related metrics from four public areal data products were compared to a monitor-based "gold standard" dust data set. The data products included the National Weather Service (NWS) storm event database, the Modern-Era Retrospective analysis for Research and Applications-Version 2, the EPA's Air QUAlity TimE Series (EQUATES) Project using the Community Multiscale Air Quality Modeling System (CMAQ), and the Copernicus Atmosphere Monitoring Service global reanalysis product. California, Nevada, Utah, and Arizona, which account for most dust storms reported in the U.S., were examined. Dichotomous and continuous metrics based on reported dust storms, particulate matter concentrations (PM10 and PM2.5), and aerosol-type variables were extracted or derived from the data products. Associations between these metrics and a validated dust storm detection method utilizing Interagency Monitoring of Protected Visual Environments monitors were estimated via quasi-binomial regression. In general, metrics from CAMS yielded the strongest associations with the "gold standard," followed by the NWS storm database metric. Dust aerosol (0.9-20μm) mixing ratio, vertically integrated mass of dust aerosol (9-20μm), and dust aerosol optical depth at 550nm from CAMS generated the highest standardized odds ratios among all metrics. Future work will apply machine-learning methods to the best-performing metrics to create a public dust storm database suitable for long-term epidemiologic studies.

Convolutional neural network-based deep learning model for air quality prediction in October city of Egypt

Purpose Modern human society has continuous advancements that have a negative impact on the quality of the air. Daily transportation, industrial and residential operations churn up dangerous contaminants in our surroundings. Addressing air pollution issues is critical for human health and ecosystems, particularly in developing countries such as Egypt. Excessive levels of pollutants have been linked to a variety of circulatory, respiratory and nervous illnesses. To this end, the purpose of this research paper is to forecast air pollution concentrations in Egypt based on time series analysis. Design/methodology/approach Deep learning models are leveraged to analyze air quality time series in the 6th of October City, Egypt. In this regard, convolutional neural network (CNN), long short-term memory network and multilayer perceptron neural network models are used to forecast the overall concentrations of sulfur dioxide (SO2) and particulate matter 10 µm in diameter (PM10). The models are trained and validated by using monthly data available from the Egyptian Environmental Affairs Agency between December 2014 and July 2020. The performance measures such as determination coefficient, root mean square error and mean absolute error are used to evaluate the outcomes of models. Findings The CNN model exhibits the best performance in terms of forecasting pollutant concentrations 3, 6, 9 and 12 months ahead. Finally, using data from December 2014 to July 2021, the CNN model is used to anticipate the pollutant concentrations 12 months ahead. In July 2022, the overall concentrations of SO2 and PM10 are expected to reach 10 and 127 µg/m3, respectively. The developed model could aid decision-makers, practitioners and local authorities in planning and implementing various interventions to mitigate their negative influences on the population and environment. Originality/value This research introduces the development of an efficient time-series model that can project the future concentrations of particulate and gaseous air pollutants in Egypt. This research study offers the first time application of deep learning models to forecast the air quality in Egypt. This research study examines the performance of machine learning approaches and deep learning techniques to forecast sulfur dioxide and particular matter concentrations using standard performance metrics.

LSTM-Autoencoder-Based Anomaly Detection for Indoor Air Quality Time-Series Data

Anomaly detection for indoor air quality (IAQ) data has become an important area of research as the quality of air is closely related to human health and well-being. However, traditional statistics and shallow machine-learning (ML)-based approaches in anomaly detection in the IAQ area could not detect anomalies involving the observation of correlations across several data points (i.e., often referred to as long-term dependencies). We propose a hybrid deep-learning model that combines long short-term memory (LSTM) with an autoencoder (AE) for anomaly detection tasks in IAQ to address this issue. In our approach, the LSTM network is comprised of multiple LSTM cells that work with each other to learn the long-term dependencies of the data in a time-series sequence. The AE identifies the optimal threshold based on the reconstruction loss rates evaluated on every data across all time-series sequences. Our experimental results, based on the Dunedin carbon dioxide (CO2) time-series dataset obtained through a real-world deployment of the schools in New Zealand, demonstrate a very high and robust accuracy rate (99.50%) that outperforms other similar models.

Multiway clustering with time-varying parameters

This paper proposes a clustering approach for multivariate time series with time-varying parameters in a multiway framework. Although clustering techniques based on time series distribution characteristics have been extensively studied, methods based on time-varying parameters have only recently been explored and are missing for multivariate time series. This paper fills the gap by proposing a multiway approach for distribution-based clustering of multivariate time series. To show the validity of the proposed clustering procedure, we provide both a simulation study and an application to real air quality time series data.

Long-term regional trends of nitrogen and sulfur deposition in the United States from 2002 to 2017

Abstract. Atmospheric deposition of nitrogen (N) and sulfur (S) compounds from human activity has greatly declined in the United States (US) over the past several decades in response to emission controls set by the Clean Air Act. While many observational studies have investigated spatial and temporal trends of atmospheric deposition, modeling assessments can provide useful information over areas with sparse measurements, although they usually have larger horizontal resolutions and are limited by input data availability. In this analysis, we evaluate wet, dry, and total N and S deposition from multiyear simulations within the contiguous US (CONUS). Community Multiscale Air Quality (CMAQ) model estimates from the EPA's (Environmental Protection Agency) Air QUAlity TimE Series (EQUATES) project contain important model updates to atmospheric deposition algorithms compared to previous model data, including the new Surface Tiled Aerosol and Gaseous Exchange (STAGE) bidirectional deposition model which contains land-use-specific resistance parameterization and land-use-specific deposition estimates needed to estimate the differential impacts of N deposition to different land use types. First, we evaluate model estimates of wet deposition and ambient concentrations, finding underestimates of SO4, NO3, and NH4 wet deposition compared to National Atmospheric Deposition Program observations and underestimates of NH4 and SO4 and overestimates of SO2 and TNO3 (HNO3+NO3) compared to the Clean Air Status and Trends Network (CASTNET) ambient concentrations. Second, a measurement–model fusion approach employing a precipitation and bias correction to wet-deposition estimates is found to reduce model bias and improve correlations compared to the unadjusted model values. Model agreement of wet deposition is poor over parts of the West and Northern Rockies, due to errors in precipitation estimates caused by complex terrain and uncertainty in emissions at the relatively coarse 12 km grid resolution used in this study. Next, we assess modeled N and S deposition trends across climatologically consistent regions in the CONUS. Total deposition of N and S in the eastern US is larger than the western US with a steeper decreasing trend from 2002–2017; i.e., total N declined at a rate of approximately −0.30 kg N ha−1 yr−1 in the Northeast and Southeast and by −0.02 kg N ha−1 yr−1 in the Northwest and Southwest. Widespread increases in reduced N deposition across the Upper Midwest, Northern Rockies, and West indicate evolving atmospheric composition due to increased precipitation amounts over some areas, growing agricultural emissions, and regional NOx/SOx emission reductions shifting gas–aerosol partitioning; these increases in reduced N deposition are generally masked by the larger decreasing oxidized N trend. We find larger average declining trends of total N and S deposition between 2002–2009 than 2010–2017, suggesting a slowdown of the rate of decline likely in response to smaller emission reductions. Finally, we document changes in the modeled total N and S deposition budgets. The average annual total N deposition budget over the CONUS decreases from 7.8 in 2002 to 6.3 kg N ha−1 yr−1 in 2017 due to declines in oxidized N deposition from NOx emission controls. Across the CONUS during the 2002–2017 time period, the average contribution of dry deposition to the total N deposition budget drops from 60 % to 52 %, whereas wet deposition dominates the S budget rising from 45 % to 68 %. Our analysis extends upon the literature documenting the growing contribution of reduced N to the total deposition budget, particularly in the Upper Midwest and Northern Rockies, and documents a slowdown of the declining oxidized N deposition trend, which may have consequences on vegetation diversity and productivity.

Marginal reduction in surface NO2 attributable to airport shutdown: A machine learning regression-based approach

Emissions from aviation and airport-related activities degrade surface air quality but received limited attention relative to regular transportation sectors like road traffic and waterborne vessels. Statistically, assessing the impact of airport-related emissions remains a challenge due to the fact that its signal in the air quality time series data is largely dwarfed by meteorology and other emissions. Flight-ban policy has been implemented in a number of cities in response to the COVID-19 spread since early 2020, which provides an unprecedented opportunity to examine the changes in air quality attributable to airport closure. It would also be interesting to know whether such an intervention produces extra marginal air quality benefits, in addition to road traffic. Here we investigated the impact of airport-related emissions from a civil airport on nearby NO2 air quality by applying machine learning predictive model to observational data collected from this unique quasi-natural experiment. The whole lockdown-attributable change in NO2 was 16.7 μg/m3, equals to a drop of 73% in NO2 with respect to the business-as-usual level. Meanwhile, the airport flight-ban aviation-attributable NO2 was 3.1 μg/m3, accounting for a marginal reduction of 18.6% of the overall NO2 change that driven by the whole lockdown effect. The airport-related emissions contributed up to 24% of the local ambient NO2 under normal conditions. Additionally, the average impact of airport-related emissions on the nearby air quality was ∼0.01 ± 0.001 μg/m3 NO2 per air-flight. Our results highlight that attention needs to be paid to such a considerable emission source in many places where regular air quality regulatory measures were insufficient to bring NO2 concentration into compliance with the health-based limit.

A Method of Pruning and Random Replacing of Known Values for Comparing Missing Data Imputation Models for Incomplete Air Quality Time Series

The data obtained from air quality monitoring stations, which are used to carry out studies using data mining techniques, present the problem of missing values. This paper describes a research work on missing data imputation. Among the most common methods, the method that best imputes values to the available data set is analysed. It uses an algorithm that randomly replaces all known values in a dataset once with imputed values and compares them with the actual known values, forming several subsets. Data from seven stations in the Silesian region (Poland) were analyzed for hourly concentrations of four pollutants: nitrogen dioxide (NO2), nitrogen oxides (NOx), particles of 10 μm or less (PM10) and sulphur dioxide (SO2) for five years. Imputations were performed using linear imputation (LI), predictive mean matching (PMM), random forest (RF), k-nearest neighbours (k-NN) and imputation by Kalman smoothing on structural time series (Kalman) methods and performance evaluations were performed. Once the comparison method was validated, it was determine that, in general, Kalman structural smoothing and the linear imputation methods best fitted the imputed values to the data pattern. It was observed that each imputation method behaves in an analogous way for the different stations The variables with the best results are NO2 and SO2. The UMI method is the worst imputer for missing values in the data sets.

Time series clustering of Malaysia Air quality time series data

Air quality is often associated with the area location and activities where air quality in cities is usually more polluted than in rural areas. This study aims to study the pattern of time series data from air quality stations by performing cluster analysis of air quality station based on the particulate matter 10 micrometres or less in diameter (PM10) and particulate matter 2.5 micrometres or less in diameter (PM2.5) time series data. The clusters obtained from the cluster analysis were compared with the station area category and station location. This study which uses air quality data obtained from the Department of Environment, Malaysia from 5 July 2017 until 30 June 2019, shows five types of air quality patterns in Malaysia. The results also show that none of the clusters is dominated by any station's category. Therefore, it is less appropriate to relate the air quality patterns and the station area category. However, the results show that air quality patterns were related to the station's location, where nearby stations have similar air quality patterns.

Mutual Information Variational Autoencoders and Its Application to Feature Extraction of Multivariate Time Series

The application of deep learning in time-series prediction has developed gradually. In this paper, we propose a deep generative network model for feature extraction of multivariate time series, namely, mutual information variational autoencoders (MI-VAE). In the architecture of the proposed model, we use the latent space of VAE for feature learning, which can extract the essential features of multivariate time-series data effectively. The latent space employed directly as a feature extractor can avoid poor interpretability of model. In addition, we introduce a mutual information term into the loss function, which improves the expression capability and accuracy of model. The proposed model, combining the merits of VAE and mutual information, extracts features for multivariate time-series data from a new perspective. The Lorenz system and Beijing air quality time series are used to test performance of the proposed model and comparative models. Results show that the proposed model is superior to other similar models in terms of accuracy and expression capability of latent space.

Imputing missing indoor air quality data with inverse mapping generative adversarial network

Sensors deployed all over the buildings are nowadays collecting a large amount of data, such as the Indoor Air Quality (IAQ) data which can provide valuable suggestions on improving indoor environments and energy consumption strategies. However, as treated as Multivariate Time Series (MTS), IAQ data often contain missing values that severely limit further analysis on them. Unfortunately, most of the existing methods fail to handle a couple of technical issues due to the complexity of MTS data, such as data distribution approximation, removing the redundancy, and so on. In this paper, we formulate the IAQ missing data imputation problem and propose an Inverse Mapping Generative Adversarial Network (IM-GAN) to tackle that problem. IM-GAN takes advantage of Bi-directional Recurrent Neural Network (BRNN), Denoising Auto-Encoder (DAE), and Generative Adversarial Network (GAN) to overcome the aforementioned technical issues. To validate the effectiveness of our proposed IM-GAN, we conduct comprehensive experiments on two public IAQ datasets GAMS and Gainesville. Results show that our IM-GAN achieves the new state-of-the-art performance in accurately estimating missing values in indoor air quality time series data, with the average performance of 0.1566 and 0.0789 in terms of Mean Relative Error, and 17.2884 and 2.7434 in terms of Mean Absolute Error on GAMS and Gainesville respectively at different missing rates. Our ablation study and visualization also validate that IM-GAN effectively overcomes the aforementioned technical issues by capturing data distribution, eliminating network saturation, and so on for IAQ data imputation.