Temporal Aggregation Research Articles

Blockchain-based prosumer incentivization for peak mitigation through temporal aggregation and contextual clustering

Peak mitigation is of interest to power companies as peak periods may require the operator to over provision supply in order to meet the peak demand. Flattening the usage curve can result in cost savings, both for the power companies and the end users. Integration of renewable energy into the energy infrastructure presents an opportunity to use excess renewable generation to supplement supply and alleviate peaks. In addition, demand side management can shift the usage from peak to off-peak times and reduce the magnitude of peaks. In this work, we present a data driven approach for incentive-based peak mitigation. Understanding user energy profiles is an essential step in this process. We begin by analysing a popular energy research dataset published by the Ausgrid corporation. Extracting aggregated user energy behavior in temporal contexts and semantic linking and contextual clustering give us insight into consumption and rooftop solar generation patterns. We implement, and performance test a blockchain-based prosumer incentivization system. The smart contract logic is based on our analysis of the Ausgrid dataset. Our implementation is capable of supporting 792,540 customers with a reasonably low infrastructure footprint.

Identifying Temporal Aggregation Effect on Crash-Frequency Modeling

Aggregation of spatiotemporal data can encounter potential information loss or distort attributes via individual observation, which would influence modeling results and lead to an erroneous inference, named the ecological fallacy. Therefore, deciding spatial and temporal resolution is a fundamental consideration in a spatiotemporal analysis. The modifiable temporal unit problem (MTUP) occurs when using data that is temporally aggregated. While consideration of the spatial dimension has been increasingly studied, the counterpart, a temporal unit, is rarely considered, particularly in the traffic safety modeling field. The purpose of this research is to identify the MTUP effect in crash-frequency modeling using data with various temporal scales. A sensitivity analysis framework is adopted with four negative binomial regression models and four random effect negative binomial models having yearly, quarterly, monthly, and weekly temporal units. As the different temporal unit was applied, the result of the model estimation also changed in terms of the mean and significance of the parameter estimates. Increasing temporal correlation due to using the small temporal unit can be handled with the random effect models.

Benchmarking of drought and climate indices for agricultural drought monitoring in Argentina

Site-specific studies are required to identify suitable drought indices (DIs) for assessing and predicting drought-related impacts. This study presents a benchmark of eight DIs and 19 large-scale climate indices (CIs) to monitor agricultural drought in Argentina. First, the link between the CIs and DIs was investigated at the departmental-administrative level and at different temporal scales. Then, the effectiveness of the DIs in explaining the variability of crop yields, understood as impacts of agricultural droughts, was evaluated using statistical regression models. Soybeans were used as the reference crop. Additionally, the performances of DIs and CIs in explaining the variability of crop yields were compared. The CIs located in the Pacific Ocean (El Niño 3.4 and El Niño 4) were found to have the best correlations with the DIs (R values up to 0.49). These relationships were stronger with longer temporal aggregations and during the wet and hot seasons (summer), showing a significant role in the triggering of droughts in Argentina. The DIs that best corelated with CIs were those that included temperature in their calculations (STCI, SVHI, and SPEI). The impacts of droughts on soybean production were better explained using DIs than with CIs (up to 89% vs 8% of variability explained) as predictors of the statistical models. SVHI-6 and SPEI-6, depending on the area of interest, were, during the phenological period of crop growth (summer), the most effective DIs in explaining annual variations in soybean yields. The results may be of interest in water resource management, drought risk management, and the Argentinean soybean production sector. Furthermore, they provide a foundation for future studies aimed at forecasting agricultural droughts and their impacts.

Multi-Scale Spatial Temporal Graph Convolutional Network for Skeleton-Based Action Recognition

Graph convolutional networks have been widely used for skeleton-based action recognition due to their excellent modeling ability of non-Euclidean data. As the graph convolution is a local operation, it can only utilize the short-range joint dependencies and short-term trajectory but fails to directly model the distant joints relations and long-range temporal information that are vital to distinguishing various actions. To solve this problem, we present a multi-scale spatial graph convolution (MS-GC) module and a multi-scale temporal graph convolution (MT-GC) module to enrich the receptive field of the model in spatial and temporal dimensions. Concretely, the MS-GC and MT-GC modules decompose the corresponding local graph convolution into a set of sub-graph convolution, forming a hierarchical residual architecture. Without introducing additional parameters, the features will be processed with a series of sub-graph convolutions, and each node could complete multiple spatial and temporal aggregations with its neighborhoods. The final equivalent receptive field is accordingly enlarged, which is capable of capturing both short- and long-range dependencies in spatial and temporal domains. By coupling these two modules as a basic block, we further propose a multi-scale spatial temporal graph convolutional network (MST-GCN), which stacks multiple blocks to learn effective motion representations for action recognition. The proposed MST-GCN achieves remarkable performance on three challenging benchmark datasets, NTU RGB+D, NTU-120 RGB+D and Kinetics-Skeleton, for skeleton-based action recognition.

Disentangled Multi-Relational Graph Convolutional Network for Pedestrian Trajectory Prediction

Pedestrian trajectory prediction is one of the important tasks required for autonomous navigation and social robots in human environments. Previous studies focused on estimating social forces among individual pedestrians. However, they did not consider the social forces of groups on pedestrians, which results in over-collision avoidance problems. To address this problem, we present a Disentangled Multi-Relational Graph Convolutional Network (DMRGCN) for socially entangled pedestrian trajectory prediction. We first introduce a novel disentangled multi-scale aggregation to better represent social interactions, among pedestrians on a weighted graph. For the aggregation, we construct the multi-relational weighted graphs based on distances and relative displacements among pedestrians. In the prediction step, we propose a global temporal aggregation to alleviate accumulated errors for pedestrians changing their directions. Finally, we apply DropEdge into our DMRGCN to avoid the over-fitting issue on relatively small pedestrian trajectory datasets. Through the effective incorporation of the three parts within an end-to-end framework, DMRGCN achieves state-of-the-art performances on a variety of challenging trajectory prediction benchmarks.

SSN3D: Self-Separated Network to Align Parts for 3D Convolution in Video Person Re-Identification

Temporal appearance misalignment is a crucial problem in video person re-identification. The same part of person (e.g. head or hand) appearing on different locations in video sequence weakens its discriminative ability, especially when we apply standard temporal aggregation such as 3D convolution or LSTM. To address this issue, we propose Self-Separated network (SSN) to seek out the same parts in different images. As the name implies, SSN, if trained in an unsupervised strategy, guarantees the selected parts distinct. With a few samples of labeled parts to guide SSN training, this semi-supervised trained SSN seeks out the parts that are human-understandable within a frame and stable across a video snippet. Given the distinct and stable person parts, rather than performing aggregation on features, we then apply 3D convolution across different frames for person re-identification. This SSN + 3D pipeline, dubbed SSN3D, is proved to be efficient through extensive experiments on both synthetic and real data.

The use of temporally aggregated data in modeling and testing a variance change in a time series

The purpose of this study is to investigate the effects of temporal aggregation on modeling and testing a single variance-change in a univariate time series process. We show that the aggregation effect on the variance-change ARMA model can be explained only with non-aggregate model parameters, based on functional links between a non-aggregate process and its m-aggregate process. In addition, through Monte Carlo simulations and a real data example, we compare empirical powers for the two CUSUMSQ tests—Lee and Park’s test and Inclán and Tiao’s test under given conditions of temporal aggregation and find out a better performance on the Inclán and Tiao CUSUMSQ test at moderate-level aggregation despite severe information loss on both tests at high-level aggregation.

Visualization of radar-observed rainfall for hydrological risk assessment

Abstract. Short-duration high-intensity rainfall constitutes a major hydro-meteorological hazard, with impacts such as pluvial (urban) flooding and debris flow. There is a great demand in society for improved information on small-scale rainfall extremes, both in real time (e.g. for early warning) and historically (e.g. for post-flood analysis). Observing this type of events is notoriously difficult, because of their extreme small-scale space-time variability. However, owing to recent advances in weather radar technology as well as integration with ground-based sensors, observational products potentially applicable in this context are now available. In this paper we present a visualization prototype tailored for hydrological risk assessment by using sub-basins as spatial units, by allowing temporal aggregation over different durations (i.e. accumulation periods) and by expressing high rainfall intensities in terms of return period exceedance. The radar-based data is evaluated by comparison with gauge observations and the quality is deemed sufficient for the intended applications. Different stakeholders have shown great interest in the prototype, which is openly accessible online.

Performance of Multi-Radar Multi-Sensor (MRMS) product in monitoring precipitation under extreme events in Harris County, Texas

With an increase in intensity and frequency of extreme precipitations as a result of climate change, it is necessary to develop effective strategies for emergency flood management plan. It is critical to integrate the use of high-resolution hydro-meteorological data (e.g., precipitation) as input for hydrologic modeling to accurately predict and reduce the negative impact of floods. The Multi-Radar Multi-Sensor (MRMS) system has been developed by the National Severe Storms Laboratory (NSSL)to produce high-resolution spatio-temporal precipitation data. While the MRMS data are available at relatively high spatial (1 km) and temporal (2 min) resolutions across the continental United States (CONUS), MRMS’s accuracy in measuring actual precipitation needs to be investigated across some urban areas such as Harris County, TX. Therefore, the objectives of this study are to evaluate i) the performance of the MRMS system compared to other precipitation products (rain gauge network, Multisensor Precipitation Estimator (MPE)) at different spatial (5, 10, 15, 30 km) and temporal aggregations (5, 10,15, 30, 60 min) during four major flooding events of May 2015 (Memorial Day flood), April 2016 (Tax Day Flood), August 2017 (Hurricane Harvey), and September 2019 (Tropical Storm Imelda) in Harris County, Texas; and ii) the effects of temporal and spatial aggregation scales on the performance of the MRMS system using a suite of statistical parameters. Point-to-grid comparisons were conducted between 142 rain gauges and MRMS system data during four extreme flood events. Overall, the MRMS system captured precipitation reasonably well with a coefficient of determination (R2) of 0.78, correlation coefficient (CC) of 0.88, root mean square error (RMSE) of 1.21 mm, critical success index (CSI) of 0.65, probability of detection (POD) of 0.98, and false alarm ratio (FAR) of 0.34 over Harris County at 15 min and 15 km temporal and spatial resolutions. The results indicate that MRMS product tends to underestimate higher precipitation rates and overestimate light precipitation. Coarser temporal resolutions from 5 min to 1 h resolved some of the overestimation issues. Temporal aggregation increased R2, CC, CSI, and error variances and decreased FAR. However, increasing spatial resolution from 1 to 30 km increased R2, CC, and CSI and reduced RMSE and FAR. A comparison of MPE QPE and MRMS products at hourly temporal resolution with gauge observations showed that both products estimate rainfall accurately for the four events. Still, on average, MRMS product has a slightly better agreement with rain gauge observations at 1-hr temporal resolution.

Comparison of nurse effect zones of shrubs versus trees for dryland rehabilitation in Northeast China

Nurse plants can form islands of fertility to rehabilitate degraded land. Areas outside plant canopies may not recover as much as areas under canopies, but the spatial extent of nurse plants' beneficial impacts is not fully understood. This study aimed to quantify the spatial and temporal aggregate nurse effects (zones of influence) of shrubs compared with trees planted for land rehabilitation in a desertified region of Inner Mongolia, China. We conducted vegetation surveys in pairs of replicate 20 × 20 m plots, each subdivided into 100 subplots (each 2 × 2 m), on four types of rehabilitation sites: shrub (Caragana microphylla) plantation sites managed for 6 or 24 years and tree (Populus simonii) plantation sites managed for 5 or 24 years. Surface soil physiochemical properties and litter amount were also measured inside and outside of nine canopies on each site. The aggregate nurse effects of improved understory vegetation and surface soil conditions extended beyond the canopies of individual plants in C. microphylla and P. simonii plantations. However, these effects were more concentrated under the canopies of C. microphylla, whereas the effects under P. simonii were spread evenly over the plantation site. This difference resulted in faster rehabilitation at the site scale in P. simonii plantations. A major factor contributing to the higher rehabilitation performance of P. simonii appeared to be larger plant height, which provides greater windbreak and shading effects compared to shrubs. Xerophytic native trees that require less water than P. simonii may be a better option for rehabilitation.

Hi-EADN: Hierarchical Excitation Aggregation and Disentanglement Frameworks for Action Recognition Based on Videos

Most existing video action recognition methods mainly rely on high-level semantic information from convolutional neural networks (CNNs) but ignore the discrepancies of different information streams. However, it does not normally consider both long-distance aggregations and short-range motions. Thus, to solve these problems, we propose hierarchical excitation aggregation and disentanglement networks (Hi-EADNs), which include multiple frame excitation aggregation (MFEA) and a feature squeeze-and-excitation hierarchical disentanglement (SEHD) module. MFEA specifically uses long-short range motion modelling and calculates the feature-level temporal difference. The SEHD module utilizes these differences to optimize the weights of each spatiotemporal feature and excite motion-sensitive channels. Moreover, without introducing additional parameters, this feature information is processed with a series of squeezes and excitations, and multiple temporal aggregations with neighbourhoods can enhance the interaction of different motion frames. Extensive experimental results confirm our proposed Hi-EADN method effectiveness on the UCF101 and HMDB51 benchmark datasets, where the top-5 accuracy is 93.5% and 76.96%.

Heat load forecasting using adaptive temporal hierarchies

Heat load forecasts are crucial for energy operators in order to optimize the energy production at district heating plants for the coming hours. Furthermore, forecasts of heat load are needed for optimized control of the district heating network since a lower temperature reduces the heat loss, but the required heat supply at the end-users puts a lower limit on the temperature level. Consequently, improving the accuracy of heat load forecasts leads to savings and reduced heat loss by enabling improved control of the network and an optimized production schedule at the plant. This paper proposes the use of temporal hierarchies to enhance the accuracy of heat load forecasts in district heating. Usually, forecasts are only made at the temporal aggregation level that is the most important for the system. However, forecasts for multiple aggregation levels can be reconciled and lead to more accurate forecasts at essentially all aggregation levels. Here it is important that the auto- and cross-covariance between forecast errors at the different aggregation levels are taken into account. This paper suggests a novel framework using temporal hierarchies and adaptive estimation to improve heat load forecast accuracy by optimally combining forecasts from multiple aggregation levels using a reconciliation process. The weights for the reconciliation are computed using an adaptively estimated covariance matrix with a full structure, enabling the process to share time-varying information both within and between aggregation levels. The case study shows that the proposed framework improves the heat load forecast accuracy by 15% compared to commercial state-of-the-art operational forecasts.

Quantitative Analysis of the Performance of Spatial Interpolation Methods for Rainfall Estimation Using Commercial Microwave Links

AbstractUsing signal level measurements from commercial microwave links (CMLs) has proven to be a valuable tool for near-ground 2D rain mapping. Such mapping is commonly based on spatial interpolation methods, where each CML is considered as a point measurement instrument located at its center. The validity of the resulted maps is tested against radar observations. However, since radar has limitations, accuracy of CML-based reconstructed rain maps remains unclear. Here we provide a quantitative comparison of the performance of CML-based spatial interpolation methods for rain mapping by conducting a systematic analysis: first by quantifying the performance of maps generated from semisynthetic CML data, and thereafter turning to real-data analysis of the same rain events. A radar product of the German Weather Service serves as ground truth for generating semisynthetic data, in which several temporal aggregations of the radar rainfall fields are used to create different decorrelation distances. The study was done over an area of 225 × 245 km2 in southern Germany, with 808 CMLs. We compare the performance of two spatial interpolation methods—inverse distance weighting and ordinary kriging—in two cases: where each CML is represented as a single point, and where three points are used. The points’ measurements values in the latter are determined using an iterative algorithm. The analysis of both cases is based on a 48-h rain event. The results reconfirm the validity of CML-based rain retrieval, showing a slight systematic performance improvement when an iterative algorithm is applied so each CML is represented by more than a single point, independent of the interpolation method.

Temporal attention learning for action quality assessment in sports video

This paper proposes an end-to-end temporal attention learning method to improve the performance of action quality assessment in sports video. For temporal weighted training, an attention-learning module is built to simulate the attention mechanism and judgement preference of human perception on action quality assessment. The weights are learned based on the loss of the segmented prediction errors and used to balance the significance of segmented features. We evaluate the proposed method on diving and gym-vault action of the benchmark AQA-7 dataset. The experimental results show that the proposed attention-aware feature training method is more effective than temporal aggregation and existing temporal relationship learning methods. Furthermore, only using the distance loss between the predicated score and the ground-truth score, without considering the ranking loss of different videos for training, this paper has achieved the state-of-the-art performance on both of the spearman rank correlation and mean Euclidean distance of the predicted scores against the judge’s scores.

The impact of privacy protection measures on the utility of crowdsourced cycling data

The use of new forms of data in the transport research domain is rapidly gaining popularity. However, these data come with specific challenges and one of the major concerns is maintaining the privacy of data subjects. One widely used approach to anonymise the data is to apply binning. Recently, data from activity-tracking applications like Strava has been utilised to study and analyse active travel. Due to privacy concerns, Strava has started providing data in a discretised format from July 2018. In this study, we aim to analyse the impact of the binning criteria on the utility of the crowdsourced data by using Strava data from 2013 to 2016 for the city of Glasgow. We applied the Strava binning criteria on the original dataset at three different temporal aggregations (i.e., Hourly, Daily and Monthly) and conducted different analyses to examine its impacts. First, we compared manual cycling counts with original and binned cycling counts from Strava data. Second, net-errors were calculated by comparing original and binned cycling counts from Strava data. Third, we estimated spatial autocorrelation statistics based on original and binned Strava counts and investigated the extent to which research outcomes change because of the binning approach. Our results confirmed significant amount of information loss. Worryingly, we also show that conclusions reached by previous studies could have been reversed if the new specification of the data had been used. We outline here what precautions researchers and planners should take when working with the binned data.

Functional deep echo state network improved by a bi-level optimization approach for multivariate time series classification

The multivariate time series (MTS) classification is one of the major tasks of time series data mining. Many methods have been proposed to investigate the MTS classification. Among them, the method based on feature representation is the most popular and widely used one. However, there exist some shortcomings for this method, such as unsatisfactory accuracy, being sensitive to noise and not able to fully make use of time series data attributes. In order to overcome these disadvantages, we propose a new method called functional deep echo state network (FDESN) for MTS classification that utilizes two special operators: temporal aggregation and spatial aggregation. In general, the parameters of the FDESN are determined by random selection, human experience or trial and error. This may increase the complexity of the FDESN or reduce the accuracy of the FDESN. In this study, a novel bi-level optimization approach is proposed to optimize the parameters of the FDESN. The parameter selection problem in the FDESN is transformed into the bi-level optimization problem. The state transition algorithm (STA) is used to solve the bi-level optimization problem. Finally, the experimental results show that the proposed method is superior to other methods. In addition, the proposed method is successfully applied to anode condition identification in aluminum electrolysis. For the aluminum electrolysis datasets, the proposed method improved the average classification accuracy by about 3.5% compared with the other methods. For a specific aluminum electrolysis dataset ACS2504, the classification accuracy significantly increased from 77.92% to 82.69% by using the proposed method.

Toward an Interoperable National Hazards Events Database for South Africa

South Africa’s National Disaster Policy and international reporting requirements including the Sendai Framework for Disaster Risk Reduction 2015-2030 and the Sustainable Development Goals require reporting databases on disaster losses in order to facilitate evidence-based decisions and actions. Data fragmentation, gaps in data availability, spatial and temporal aggregations, poor data quality, and lack of metadata are South Africa’s most significant challenges to improved risk management planning, mitigation, and disaster loss reporting. Researchers need to discover and access scale- and task-appropriate data in order to integrate into their models or analyses. Although there is a growing number of international and national initiatives aimed at publishing disaster loss data, open access, integration and reuse remains difficult due to interoperability barriers to working with datasets. The current COVID-19 Pandemic has placed a number of these challenges at the forefront with the issues of access to trustworthy data coming under scrutiny. We consider the current sources of disaster data for South Africa and present our recommended solutions built into an online disaster reporting prototype system.

How much information on precipitation is contained in satellite infrared imagery?

Satellite infrared (IR) brightness temperature providing information on the characteristics of cloud tops is one of the primary input data used for estimating precipitation from satellites. Thus, the natural questions to ask are: Does IR imagery contain more information about precipitation in some regions more than others? What are the factors that lead to such differences? Should we expect the amount of information provided by IR imagery to change with respect to temporal or spatial aggregation? Comprehensive analysis of the accuracy of satellite-based precipitation estimates obtained from any given algorithm may provide qualitative answers to the above questions. However, such an approach only offers conclusions within the limited context of the algorithm used to obtain precipitation estimates; that is, patterns caused by the inherent information content of IR imagery are inextricable from those induced by assumptions embedded in the precipitation estimation algorithm. Here, we address the abovementioned questions from a different perspective using an information-theoretic measure, free of structural assumptions and general to a wide range of relationships, to characterize the average and seasonal dependence between IR imagery provided by satellite sensors in the spectral range (10.7–11.5 μm) and precipitation rates in the contiguous United States (CONUS) across distinct temporal and spatial scales. We analyze a total of more than 1.3 billion pairs of IR and precipitation observations over CONUS, and we observe interesting patterns. First, we show that there is a strong inverse relationship (ρ = − 0.73) between the information content of IR data and the number of no-rain observations; however, its robustness varies regionally and seasonally with less significant correlation during the warm season. Second, we demonstrate that the intuitive relationship of increasing dependence between IR and precipitation as a result of temporal or spatial aggregation exhibits a diminishing returns behavior. For instance, temporal aggregation from 1 to 3 h increases the dependence approximately 7 times as much as temporal aggregation from 12 to 24 h. Furthermore, we quantitatively examine and re-confirm several statements previously reported in the literature such as the strong association between IR and precipitation in convective storms and the low association in orographic rainfall. Finally, we conclude by illustrating the potential of the analysis in diagnosis of operational algorithms for estimating precipitation from IR brightness temperature. The findings of this study pinpoint spatial domains and time scales at which IR is not an adequate proxy for estimating precipitation. Thus, they can potentially guide the development of operational algorithms that utilize satellite Infrared imagery for estimating, downscaling and data fusion of precipitation. Furthermore, the analysis presented here opens up the possibility of developing new methodologies for diagnosis of satellite-based precipitation estimation algorithms.

Incorporating Information on Control Diseases Across Space and Time to Improve Estimation of the Population-level Impact of Vaccines.

The synthetic control method evaluates the impact of vaccines while adjusting for a set of control time series representing diseases that are unaffected by the vaccine. However, noise in control time series, particularly in areas with small counts, can obscure the association with the outcome, preventing proper adjustments. To overcome this issue, we investigated the use of temporal and spatial aggregation methods to smooth the controls and allow for adjustment of underlying trends. We evaluated the impact of pneumococcal conjugate vaccine on all-cause pneumonia hospitalizations among adults ≥80 years of age in 25 states in Brazil from 2005 to 2015. Pneumonia hospitalizations in this group indicated a strong increasing secular trend over time that may influence estimation of the vaccine impact. First, we aggregated control time series separately by time or space before incorporation into the synthetic control model. Next, we developed distributed lags models (DLMs) to automatically determine what level of aggregation was most appropriate for each control. The aggregation of control time series enabled the synthetic control model to identify stronger associations between outcome and controls. As a result, the aggregation models and DLMs succeeded in adjusting for long-term trends even in smaller states with sparse data, leading to more reliable estimates of vaccine impact. When synthetic control struggles to identify important prevaccine associations due to noise in control time series, users can aggregate controls over time or space to generate more robust estimates of the vaccine impact. DLMs automate this process without requiring prespecification of the aggregation level.

TSMAA‐TRI: A temporal multi‐criteria sorting approach under uncertainty

AbstractIn recent years, the Quebec government has highlighted the importance of making decisions that are both sustainable and robust under climate change uncertainties. This paper aims to answer the following question: How to sort the alternatives according to their degree of sustainability achievement while evaluations are uncertain and temporal? The general objective of the paper is to propose a first temporal sorting method under stochastic uncertainty. The proposed method, called TSMAA‐Tri, will assign each alternative to a predefined category based on a generalization of SMAA‐Tri to a temporal context (multi‐period evaluation of alternatives) where alternative evaluations are stochastic. The method starts performing Monte Carlo simulations to generate stochastic evaluation values. Each simulation (scenario of uncertainty) will generate a specific value for each criterion using the corresponding probability distribution. Then, aggregation consists in applying SMAA Tri at each period and performing a triple aggregation: (a) uncertainty aggregation; (b) multi‐criteria aggregation; and (c) temporal aggregation. Multi‐criteria aggregation consists in applying the SMAA‐TRI method at each period. Then, two ways of temporal aggregation are proposed, based either on acceptability index or on outranking index of boundary profile. The proposed method is illustrated for sustainable forest management to show its applicability.