High Temporal Correlation Research Articles

Monitoring of high-dimensional and high-frequency data streams: A nonparametric approach

ABSTRACT High-dimensional and high-frequency data streams (HHDS) have become widely available with the prevalence of data-acquisition technologies. Because HHDS have the characteristics of high dimensionality and temporal correlation, traditional statistical process control (SPC) techniques cannot be used directly to address the problem of monitoring HHDS. Most extant monitoring methods commonly assume that the process observations are independent or can be described by some parametric models and face the challenge posed by high-dimensional data. Little research has been conducted on the robust monitoring of HHDS that can simultaneously accommodate high dimensions and high frequencies. Therefore, in this paper, we propose a novel nonparametric approach for monitoring HHDS, based on data decorrelation, dimension reduction, and SPC. Specifically, process observations are first sequentially decorrelated and standardized using Cholesky decomposition, and then the k -nearest neighbors classification algorithm is applied to transform uncorrelated high-dimensional data into one-dimensional data. Finally, the traditional cumulative-sum procedure is used for online monitoring, based on an empirical log-likelihood ratio test. Numerical studies have shown that our monitoring scheme is sufficiently reliable and efficient for the online monitoring of HHDS. An illustrative example about the Dow Jones 30 industrial stock prices is presented to further validate the proposed approach.

Historical information fusion of dense multi-source satellite image time series for flood extent mapping

Existing lightweight flood extent mapping (FEM) methods, leveraging satellite imagery, offer respite from the burden of extensive data processing. However, these methods often grapple with suboptimal accuracy of the FEM results. To tackle this challenge, this paper introduces an innovative lightweight historical information fusion framework aimed at enhancing FEM results through the utilization of dense multi-source satellite image time series (MSSITS). Firstly, we propose an ultra-lightweight convolutional neural network comprising fundamental network modules, such as convolution layers, batch normalization layers, and parametric rectified linear unit layers. This network swiftly extracts initial FEM results from dense MSSITS. Secondly, we devise a historical information fusion strategy to refine the initial FEM results of the current image by fusing historical FEM results exhibiting high temporal correlation. During the fusion stage, we harness the complementary attributes of dense MSSITS. Specifically, the fusion of diverse satellite sensor data aids in mitigating false alarms in the initial FEM results, thereby facilitating the acquisition of refined FEM results. This approach effectively diminishes the impact of noise interferences, including terrain shadows and surging flood waves in synthetic aperture radar images, as well as cloud occlusions and cloud shadows in optical images. Experimental findings, based on the dense MSSITS of four authentic flood events, demonstrate that our proposed method yields relatively superior FEM performance while maintaining a low FEM complexity in comparison with commonly used semantic segmentation methods.

The 4DEnVar-based weakly coupled land data assimilation system for E3SM version 2

Abstract. A new weakly coupled land data assimilation (WCLDA) system based on the four-dimensional ensemble variational (4DEnVar) method is developed and applied to the fully coupled Energy Exascale Earth System Model version 2 (E3SMv2). The dimension-reduced projection four-dimensional variational (DRP-4DVar) method is employed to implement 4DVar using the ensemble technique instead of the adjoint technique. With an interest in providing initial conditions for decadal climate predictions, monthly mean anomalies of soil moisture and temperature from the Global Land Data Assimilation System (GLDAS) reanalysis from 1980 to 2016 are assimilated into the land component of E3SMv2 within the coupled modeling framework with a 1-month assimilation window. The coupled assimilation experiment is evaluated using multiple metrics, including the cost function, assimilation efficiency index, correlation, root-mean-square error (RMSE), and bias, and compared with a control simulation without land data assimilation. The WCLDA system yields improved simulation of soil moisture and temperature compared with the control simulation, with improvements found throughout the soil layers and in many regions of the global land. In terms of both soil moisture and temperature, the assimilation experiment outperforms the control simulation with reduced RMSE and higher temporal correlation in many regions, especially in South America, central Africa, Australia, and large parts of Eurasia. Furthermore, significant improvements are also found in reproducing the time evolution of the 2012 US Midwest drought, highlighting the crucial role of land surface in drought lifecycle. The WCLDA system is intended to be a foundational resource for research to investigate land-derived climate predictability.

Investigating the spatio–temporal characteristics of lower atmospheric ducts across the China seas by performing a long–term simulation using the WRF model

In this work, a diagnostic scheme for lower atmospheric ducts was established based on the Weather Research and Forecasting (WRF) model. More specifically, a 10-year simulation test was conducted for the China seas to investigate the spatio-temporal characteristics of the lower atmospheric ducts phenomenon. Compared with the sounding data, the long-term simulations showed a high temporal correlation and the root mean square error of the modified atmospheric refractivity remained between 4 M and 7 M. Based on the simulations, significant regional differences in the occurrence probability of lower atmospheric ducts were detected from south to north. Among them, the surface ducts near the sea surface exhibited the highest occurrence probability, with higher probabilities being recorded in autumn and winter, and the probability gradually increased with the decreasing latitude. The spatio-temporal characteristics of duct height, thickness, and strength were generally consistent. In the seas at mid-latitudes, strong ducts mostly occurred in the spring and autumn, with the single-layer ducts being predominant and the first layer duct showing stronger characteristics than the second layer. In the lower latitude regions, the situation was exactly the opposite. The first duct layer, which existed throughout the year, exhibited weaker characteristics with less pronounced seasonal variations. On the other hand, the second duct layer demonstrated stronger features.

Revisiting the Link Between Thunderstorms and Upper Tropospheric Water Vapor

AbstractAs the Earth's temperatures continue to rise due to increasing greenhouse gases in the atmosphere, a large portion of the warming is due to positive feedbacks from increasing atmospheric water vapor or specific humidity (SH). Some of this water vapor in the boundary layer is transported via deep convection to the upper troposphere (as droplets and ice crystals), moistening the upper troposphere. These small changes in SH in the upper troposphere have a significant impact on the Earth's radiation balance. We compared global daily lightning from the WWLLN data set, and SH data from the ERA5 reanalysis product for 2019, at a spatial resolution of 5°. Our findings show high spatial and temporal correlations between the lightning activity and the SH concentrations in the upper troposphere. The best correlations (r2 = 0.72, p << 0.001) are between lightning activity and UTWV at the 200 mb level (∼12 km altitude), although the correlations with SH at 300 and 400 mb were only slightly lower. Lightning and SH migrate with the seasons north and south of the equator with both parameters showing maxima in the summer hemisphere around 10° latitude. Furthermore, the correlation increases slightly (r2 = 0.73) if a 1‐day lag is considered between the lightning activity and the UTWV. We find that the daily global SH in the upper troposphere can be best estimated using daily global lightning frequency and a power law having an exponent of 0.33.

A novel exergy-related fault detection and diagnosis framework with transformer-based conditional generative adversarial networks for hot strip mill process

Fault detection and diagnosis (FDD) plays a crucial role in the iron and steel industry. However, the iron and steel industry have unique complex characteristics such as high energy consumption, temporal and spatial correlation, and data imbalance, which makes the application of traditional FDD methods ineffective. With the development of FDD, deep learning-based methods have attracted more and more attention. To address the above issues, this work combines the advantages of transformer and conditional generative adversarial networks (CGAN) to propose a novel exergy-related FDD framework with transformer-based CGAN (TransCGAN). Firstly, we extract exergy-related variables and convert one-dimensional data into image data to preserve spatiotemporal information. Next, the transformer encoder is used as the basic block of CGAN to capture the relationship between normal and fault data, and generate fault samples. Then, the generated data and original data are mixed to train the proposed FDD model. Finally, the feasibility of TransCGAN is verified by the hot strip mill process (HSMP). Compared with other state-of-the-art (SOTA) methods, the proposed method has higher fault detection rate and better fault diagnostic performance.

Temporal stability and correlation of EEG markers and depression questionnaires scores in healthy people

Mental disorders, especially depression, have become a rising problem in modern society. The development of methods and markers for the early detection of mental disorders is an actual problem. Psychological questionnaires are the only tools for evaluating the symptoms of mental disorders in clinical practice today. The electroencephalography (EEG) based non-invasive and cost-effective method seems feasible for the early detection of depression in occupational and family medicine centers and personal monitoring. The reliability of the EEG markers in the early detection of depression assumes their high temporal stability and correlation with the scores of depression questionnaires. The study was been performed on 17 healthy people over three years. Two hypotheses have been evaluated in the current study: first, the temporal stability of EEG markers is close to the stability of the scores of depression questionnaires, and second, EEG markers and depression questionnaires’ scores are not correlated in healthy people. The results of the performed study support both hypotheses: the temporal stability of EEG markers is high and close to the stability of depression questionnaires scores and the correlation between the EEG markers and depression questionnaires scores is not detected in healthy people. The results of the current study contribute to the interpretation of results in depression EEG studies and to the feasibility of EEG markers in the detection of depression.

HD‐Net: A hybrid dynamic spatio‐temporal network for traffic flow prediction

AbstractAccurately predicting traffic flow is crucial for intelligent transportation systems (ITS). In recent years, many deep learning‐based prediction models have been widely applied in traffic flow prediction, and various spatio‐temporal networks have been proposed. However, most of the existing models follow a general technical route to extract the spatio‐temporal features, which lack the capacity of extracting the important historical information with the high spatial and temporal correlations dynamically and deeply. How to develop a well‐performance traffic flow prediction model for a complex transportation network is still facing some challenges. In this paper, a hybrid dynamic spatio‐temporal network (HD‐Net) for traffic flow prediction is proposed. In HD‐Net, the authors first extract the dynamic spatio‐temporal features using dynamic graph convolution and bidirectional gate recurrent uni (BiGRU). Subsequently, the authors extract the important features with high spatial and temporal correlations from the obtained dynamic spatio‐temporal features using an auto‐correlation mechanism from a local perspective, and self‐attention mechanism from a global perspective, respectively. Extensive experiments have been conducted on two real‐world traffic datasets. The experimental results demonstrate that the proposed HD‐Net outperforms the baselines in the field of capturing the dynamic and important spatio‐temporal features with high correlations.

Observations of near-inertial internal wave amplification and enhanced mixing after surface reflection

The overreflection process of near-inertial internal waves (NIWs) has been theoretically predicted for several decades; however, to the best of our knowledge, this phenomenon has never been comprehensively investigated in real ocean scenarios. Based on the buoy observations collected several days after the passage of Typhoon Lekima in the Yellow Sea, a NIW surface overreflection event is clearly captured. The observed NIWs undergo nearly total reflection meridionally but are amplified zonally after reflection by approximately 20% in amplitude and 56% in vertically integrated horizontal kinetic energy. Ray tracing analysis indicates that the NIW was generated in the wake of Typhoon Lekima in the area north of the Shandong Peninsula and may propagated to the buoy station as coastal-trapped internal Kelvin waves. A simulation using a slab mixed layer model suggests that local wind work was insufficient to generate the amplified NIWs. The temporal evolution of near-inertial energy also implies that the intensified near-inertial waves cannot be attributed to the spontaneous generation resulting from unbalanced flows or the parametric subharmonic instability of M2 internal tides during the reflection period. We found a high temporal correlation between the zonal NIW enhancement and the duration of a meridional lens-type shear flow after reflection, which is consistent with the Stern’s overreflection theory (Stern, 1977) that perpendicular background shear flow can feed energy to the incident NIWs. This indicates that the enhanced NIW may be stimulated by the near-surface reflection and the rotation effect plays a crucial role in the NIWs overreflection process in the real ocean. Furthermore, enhanced instability are found between the ocean surface and the upper thermocline after reflection. This study provides observational evidence that the background field could inject energy into the near-inertial band through NIW overreflection process, and may shed some light on understanding upper ocean mixing caused by NIW reflection.

Detecting drought stress occurrence using synergies between Sun induced fluorescence and vegetation surface temperature spatial records

Drought stress occurrence and recovery from drought can be detected using a single spatial set of simultaneous observations of SIF and canopy temperature records. Temporal and spatial responses to drought and heat stresses by plant stands of a drought-adapted diverse grassland ecosystem were studied using sun induced fluorescence (SIF,O2A and O2B bands) and further ecophysiological (canopy temperature (Tsurf), spatially modeled evapotranspiration, vegetation reflectance spectra) variables collected along spatial sampling grids while also utilizing eddy covariance measured carbon dioxide (net ecosystem exchange: NEE, gross primary production: GPP) and water flux (evapotranspiration: ET) data. The grids were of 0.5 and 5 ha spatial extents and contained 78 sampling points. Data were collected in four spatial sampling campaigns, two under drought (early summer) and another two during and after recovery (midsummer) at both spatial resolutions.Small values of spatial SIF_A averages (around 0.5 mW m−2 nm−1 sr−1) under strong early summer drought increased (to around 2 mW m−2 nm−1 sr−1) due recovery upon rain arrivals, showing high (R2: 0.8–0.88) positive temporal correlations to eddy covariance measured carbon (GPP, NEE) and water (ET) fluxes. Spatial averages of LAI, vegetation indices (NDVI, NIRv) and modeled ET followed similar temporal patterns.While SIF was depressed by drought, it showed higher values in high canopy temperature vegetation patches than in vegetation patches with lower Tsurf. The spatial pattern of higher SIF in higher Tsurf patches was persistent (2 weeks) under drought. The positive SIF_A-Tsurf spatial correlation turned into negative/not significant after recovery of the grassland from the drought, while hot summer weather persisted. It is proposed that, by using a single set of simultaneously measured spatial SIF and Tsurf data it is possible to infer whether the studied vegetation is under drought (and heat) stress while it could not be decided on the base of SIF data alone. Evaluation of the slope of the above relationship seems therefore beneficial before e.g. starting the (stress) classification procedure based on SIF.

A Flexible Spatiotemporal Thick Cloud Removal Method with Low Requirements for Reference Images

Thick cloud and shadows have a significant impact on the availability of optical remote sensing data. Although various methods have been proposed to address this issue, they still have some limitations. First, most approaches rely on a single clear reference image as complementary information, which becomes challenging when the target image has large missing areas. Secondly, the existing methods that can utilize multiple reference images require the complementary data to have high temporal correlation, which is not suitable for situations where the difference between the reference image and the target image is large. To overcome these limitations, a flexible spatiotemporal deep learning framework based on generative adversarial networks is proposed for thick cloud removal, which allows for the use of three arbitrary temporal images as references. The framework incorporates a three-step encoder that can leverage the uncontaminated information from the target image to assimilate the reference images, enhancing the model’s ability to handle reference images with diverse temporal differences. A series of simulated and real experiments on Landsat 8 and Sentinel 2 data is performed to demonstrate the effectiveness of the proposed method. The proposed method is especially applicable to small/large-scale regions with reference images that are significantly different from the target image.

An ERA5 based local modelling of weighted mean temperature over hilly region in India for improved spatiotemporal analysis of extreme weather event using GNSS PWV

Precipitable water vapor (PWV) from Global Navigation Satellite System (GNSS) offers the benefits of high spatiotemporal resolution logging-consistency and all-weather operability. Estimation of water vapor weighted mean temperature (Tm) is a crucial part and a major error source in extracting GNSS PWV from atmospheric delay. Empirical modelling of Tm is a convenient approach, especially in cases where, the meteorology community suffers from inefficient and inconsistent logging of temperature and humidity profiles etc., in the existing meteorological observing systems, particularly for harsh weather conditions. In a first, a novel site specific model (Tm-SSM) is developed for the atmospheric weighted mean temperature (Tm) for 13 CORS (continuously operating reference stations) in the Survey of India (SOI) network, recently established in the hilly state of Uttarakhand (UK) using the ERA5 reanalysis data from 2016 to 2018. GNSS observations at 10 CORS are then used to observe the spatiotemporal characteristics of PWV from Tm-SSM during the heavy rainfall event in October 2021. The newly developed Tm-SSM performs better with the mean bias error (MBE) and root mean square error (RMSE) of 0.498 and 2.9 K, respectively, when compared with other globally accepted models. The Tm-SSM PWV estimates are well correlated with the precipitation data from AWS stations. The newly developed Tm-SSM is well conditioned and highly stable, with the mean stability indicator value as low as 1.043%. Further, a close relationship can be observed from the spatiotemporal characteristics of GNSS PWV during heavy rainfall event, with the in-situ precipitation measurements. It is apparent from the results that PWV steadily builds up before the arrival of rain, with sharp peaks in PWV observed just before rainfall event. It then, follows a downward slope through which it decreases to a stable value of around 30 mm when the rainfall ceases. The fluctuations in the PWV from Tm-SSM and the passage of the severe mesoscale convective system in a complex terrain such as UK were shown to have high spatial and temporal correlation, which is significant for the monitoring and forecasting such extreme weather. The findings imply that GNSS PWV represents the compendious signature of the water vapor dynamics during severe weather, demonstrating the potential to improve the early detection and sensing of severe weather.

Adaptive Feature Fusion Networks for Origin-Destination Passenger Flow Prediction in Metro Systems

Accurately predicting Origin-Destination (OD) passenger flow can help metro service quality and efficiency. Existing works have focused on predicting incoming and outgoing flows for individual stations, while little attention was paid to OD prediction in metro systems. The challenges are that OD flows 1) have high temporal dynamics and complex spatial correlations, 2) are affected by external factors, and 3) have sparse and incomplete data slices. In this paper, we propose an Adaptive Feature Fusion Network (AFFN) to a) adaptively fuse spatial dependencies from multiple knowledge-based graphs and even hidden correlations between stations and b) accurately capture the periodic patterns of passenger flows based on the auto-learned impact from external factors. To deal with the incompleteness and sparsity of OD matrices, we extend AFFN to multi-task AFFN to predict the inflow and outflow of each station as a side-task to further improve OD prediction accuracy. We conducted extensive experiments on two real-world metro trip datasets collected in Nanjing and Xi’an, China. Evaluation results show that our AFFN and multi-task AFFN outperform the state-of-the-art baseline techniques and AFFN variants in various accuracy metrics, demonstrating the effectiveness of AFFN and each of its key components in OD prediction.

A twenty-year dataset of soil moisture and vegetation optical depth from AMSR-E/2 measurements using the multi-channel collaborative algorithm

Soil moisture (SM) and vegetation optical depth (VOD) are essential variables in the terrestrial ecosystem. The multi-frequency radiometers AMSR-E and AMSR2 provide >20 years of data records, enabling the development of long-term SM and VOD products. Most of the current retrieval algorithms either only focus on SM or VOD, and generally ignore the polarization or simplify the frequency dependence of vegetation effects for reducing the unknowns and facilitating the retrieval process, limiting the synergic applicability of VOD and SM products in the soil-plant-atmosphere continuum. In this study, a new global SM and frequency- and polarization-dependent VOD product from 2002 to 2021 was developed using the multi-channel collaborative algorithm (MCCA), based on the inter-calibrated AMSR-E/2 multi-frequency passive microwave measurements. The MCCA algorithm comprehensively considers the physical relationship between multiple microwave channels and could retrieve frequency- and polarization-dependent VOD while considering the accuracy of the SM retrievals. In the overall comparison with other SM products (AMSR-ANN, CCI-passive v07.1, LPRM-C/X, JAXA) over 25 dense SM networks, MCCA achieved the best scores in terms of root mean square error (RMSE = 0.074 m3/m3), unbiased root mean square error (ubRMSE = 0.073 m3/m3) and bias (0.007 m3/m3), and presented slightly lower value of Pearson's correlation coefficient (R = 0.709) than LPRM-X (R = 0.735). For the indirect evaluation of VOD with aboveground biomass (AGB) and MODIS NDVI, the MCCA product showed the performance comparable to other products (LPRM-C/X, VODCA-C/X/Ku). MCCA-derived VODs, especially for the H-polarized VODs, exhibited smooth non-linear density distribution with AGB and high temporal correlations with MODIS NDVI over most regions of the globe. In particular, MCCA-derived VODs can physically present reasonable variations across the microwave spectrum (values of VOD increase with microwave frequency), which is superior to the LPRM and VODCA products. It is expected that the MCCA algorithm can be extended to the observations of the ongoing AMSR2 or other similar satellite missions with multi-frequency capability, such as FY-3B/C/D/F/G or the upcoming AMSR3 and CMIR missions.

Dynamics of social media behavior before and after SARS-CoV-2 infection.

Online social media have been both a field of research and a source of data for research since the beginning of the COVID-19 pandemic. In this study, we aimed to determine how and whether the content of tweets by Twitter users reporting SARS-CoV-2 infections changed over time. We built a regular expression to detect users reporting being infected, and we applied several Natural Language Processing methods to assess the emotions, topics, and self-reports of symptoms present in the timelines of the users. Twelve thousand one hundred and twenty-one twitter users matched the regular expression and were considered in the study. We found that the proportions of health-related, symptom-containing, and emotionally non-neutral tweets increased after users had reported their SARS-CoV-2 infection on Twitter. Our results also show that the number of weeks accounting for the increased proportion of symptoms was consistent with the duration of the symptoms in clinically confirmed COVID-19 cases. Furthermore, we observed a high temporal correlation between self-reports of SARS-CoV-2 infection and officially reported cases of the disease in the largest English-speaking countries. This study confirms that automated methods can be used to find digital users publicly sharing information about their health status on social media, and that the associated data analysis may supplement clinical assessments made in the early phases of the spread of emerging diseases. Such automated methods may prove particularly useful for newly emerging health conditions that are not rapidly captured in the traditional health systems, such as the long term sequalae of SARS-CoV-2 infections.

Variability in Antarctic surface climatology across regional climate models and reanalysis datasets

Abstract. Regional climate models (RCMs) and reanalysis datasets provide valuable information for assessing the vulnerability of ice shelves to collapse over Antarctica, which is important for future global sea level rise estimates. Within this context, this paper examines variability in snowfall, near-surface air temperature and melt across products from the Met Office Unified Model (MetUM), Regional Atmospheric Climate Model (RACMO) and Modèle Atmosphérique Régional (MAR) RCMs, as well as the ERA-Interim and ERA5 reanalysis datasets. Seasonal and trend decomposition using LOESS (STL) is applied to split the monthly time series at each model grid cell into trend, seasonal and residual components. Significant systematic differences between outputs are shown for all variables in the mean and in the seasonal and residual standard deviations, occurring at both large and fine spatial scales across Antarctica. Results imply that differences in the atmospheric dynamics, parametrisation, tuning and surface schemes between models together contribute more significantly to large-scale variability than differences in the driving data, resolution, domain specification, ice sheet mask, digital elevation model and boundary conditions. Despite significant systematic differences, high temporal correlations are found for snowfall and near-surface air temperature across all products at fine spatial scales. For melt, only moderate correlation exists at fine spatial scales between different RCMs and low correlation between RCM and reanalysis outputs. Root mean square deviations (RMSDs) between all outputs in the monthly time series for each variable are shown to be significant at fine spatial scales relative to the magnitude of annual deviations. Correcting for systematic differences results in significant reductions in RMSDs, suggesting the importance of observations and further development of bias-correction techniques.

Downscaling MODIS nighttime land surface temperatures in urban areas using ASTER thermal data through local linear forest

• A new nighttime LST downscaling method using local linear forest (LLF) was proposed. • Feature-selected ASTER nighttime LSTs were identified as important input kernels. • The LLF-based model predicted LST with high linearity, especially for both extremes. • DLST revealed more distinct characteristics of urban climate than MODIS LST. Spatial downscaling effectively produces high spatiotemporal resolution land surface temperature (LST) in urban areas. Although nighttime LST is an essential indicator in urban thermal research, few LST downscaling studies have focused on nighttime in fine resolution. This study proposed a novel approach using local linear forest (LLF) to downscale 1 km Moderate Resolution Imaging Spectroradiometer (MODIS) nighttime LSTs to 250 m spatial resolution in three cities: Rome, Madrid, and Seoul. First, we used Least Absolute Shrinkage and Selection Operator (LASSO) to select a set of past clear-sky ASTER LSTs (ALST) which showed a high spatial correlation with the target MODIS LST. Downscaling models were then developed using input kernels of the selected ALSTs and eight auxiliary variables: normalized difference vegetation index (NDVI), elevation, slope, built-up area percentage, road density, population density, wind speed, and distance from the built-up weighted center of the study area. Three schemes were evaluated: scheme 1 (S1) using only auxiliary variables as input kernels with a random forest (RF) model; scheme 2 (S2) using selected ALSTs and auxiliary variables as input kernels with an RF model; and scheme 3 (S3) using input kernels as in S2 but with the LLF model. Validation was performed using bias-corrected ALSTs for seven reference dates in the three cities. LLF-based S3 showed the highest accuracy with an average correlation coefficient (R) ∼ 0.94 and Root Mean Square Error (RMSE) ∼ 0.64 K while maintaining the dynamic range of the original LST at the finer resolution. The downscaled LST (DLST) based on S3 effectively depicted the nocturnal thermal spatial pattern in greater detail than the other two schemes did. The S3-based DLST also showed a relatively high spatial correlation with the in-situ nighttime air temperature within the cities. When compared to the original 1 km LST, S3-based DLST showed larger surface urban heat island intensity for the urban-type surfaces and a higher temporal correlation with nighttime air temperature.

Simulation of Isoprene Emission with Satellite Microwave Emissivity Difference Vegetation Index as Water Stress Factor in Southeastern China during 2008

Isoprene is one of the most important biogenic volatile organic compounds (BVOCs) emitted by vegetation. The biogenic isoprene emissions are widely estimated by the Model of Emission of Gases and Aerosols from Nature (MEGAN) considering different environmental stresses. The response of isoprene emission to the water stress is usually parameterized using soil moisture in previous studies. In this study, we designed a new parameterization scheme of water stress in MEGAN as a function of a novel, satellite, passive microwave-based vegetation index, Emissivity Difference Vegetation Index (EDVI), which indicates the vegetation inner water content. The isoprene emission rates in southeastern China were simulated with different water stress indicators including soil moisture, EDVI, Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI). Then the simulated isoprene emission rates were compared to associated satellite top-down estimations. The results showed that in southeastern China, the spatiotemporal correlations between those simulations and top-down retrieval are all high with different biases. The simulated isoprene emission rates with EDVI-based water stress factor are most consistent with top-down estimation with higher temporal correlation, lower bias and lower RMSE, while soil moisture alters the emission rates little, and optical vegetation indices (NDVI and EVI) slightly increase the correlation with top-down. The temporal correlation coefficients are increased after applied with EDVI water stress factor in most areas; especially in the Yunnan-Guizhou Plateau and Yangtze River Delta (>0.12). Overall, higher consistency of simulation and top-down estimation is shown when EDVI is applied, which indicates the possibility of estimating the effect of vegetation water stress on biogenic isoprene emission using microwave observations.

A Reconstructed Global Daily Seamless SIF Product at 0.05 Degree Resolution Based on TROPOMI, MODIS and ERA5 Data

Satellite-derived solar-induced chlorophyll fluorescence (SIF) has been proven to be a valuable tool for monitoring vegetation’s photosynthetic activity at regional or global scales. However, the coarse spatiotemporal resolution or discrete space coverage of most satellite SIF datasets hinders their full potential for studying carbon cycle and ecological processes at finer scales. Although the recent TROPOspheric Monitoring Instrument (TROPOMI) partially addresses this issue, the SIF still has drawbacks in spatial insufficiency and spatiotemporal discontinuities when gridded at high spatiotemporal resolutions (e.g., 0.05°, 1-day or 2-day) due to its nonuniform sampling sizes, swath gaps, and clouds contaminations. Here, we generated a new global SIF product with Seamless spatiotemporal coverage at Daily and 0.05° resolutions (SDSIF) during 2018–2020, using the random forest (RF) approach together with TROPOMI SIF, MODIS reflectance and meteorological datasets. We investigated how the model accuracy was affected by selection of explanatory variables and model constraints. Eventually, models were trained and applied for specific continents and months given the similar response of SIF to environmental variables within closer space and time. This strategy achieved better accuracy (R2 = 0.928, RMSE = 0.0597 mW/m2/nm/sr) than one universal model (R2 = 0.913, RMSE = 0.0653 mW/m2/nm/sr) for testing samples. The SDSIF product can well preserve the temporal and spatial characteristics in original TROPOMI SIF with high temporal correlations (mean R2 around 0.750) and low spatial residuals (less than ±0.081 mW/m2/nm/sr) between them two at most regions (80% of global pixels). Compared with the original SIF at five flux sites, SDSIF filled the temporal gaps and was better consistent with tower-based SIF at the daily scale (the mean R2 increased from 0.467 to 0.744. Consequently, it provided more reliable 4-day SIF averages than the original ones from sparse daily observations (e.g., the R2 at Daman site was raised from 0.614 to 0.837), which resulted in a better correlation with 4-day tower-based GPP. Additionally, the global coverage ratio and local spatial details had also been improved by the reconstructed seamless SIF. Our product has advantages in spatiotemporal continuities and details over the original TROPOMI SIF, which will benefit the application of satellite SIF for understanding carbon cycle and ecological processes at finer spatial and temporal scales.

SoK] The Great GAN Bake Off, An Extensive Systematic Evaluation of Generative Adversarial Network Architectures for Time Series Synthesis

There is no standard approach to compare the success ofdifferent neural network architectures utilized for time seriessynthesis. This hinders the evaluation and decision process,as to which architecture should be leveraged for an unknowndata set. We propose a combination of metrics, which empiri-cally evaluate the performance of neural network architecturestrained for time series synthesis. With these measurementswe are able to account for temporal correlations, spatial cor-relations and mode collapse issues within the generated timeseries.We further investigate the interaction of different genera-tor and discriminator architectures between each other. Theconsidered architectures include recurrent neural networks,temporal convolutional networks and transformer-based net-works. So far, the application of transformer-based models islimited for time series synthesis. Hence, we propose a newtransformer-based architecture, which is able to synthesisetime series. We evaluate the proposed architectures and theircombinations in over 500 experiments, amounting to over2500 computing hours. We provide results for four data sets,one univariate and three multivariate. The data sets vary withregard to length, as well as patterns in temporal and spatialcorrelations.We use our metrics to compare the performance of genera-tive adversarial network architectures for time series synthesis.To verify our findings we utilize quantitative and qualitativeevaluations. Our results indicate that temporal convolutionalnetworks currently outperform recurrent neural network andtransformer based approaches with regard to fidelity and flex-ibility of the generated time series data. Temporal convolu-tional network architecture are the most stable architecture fora mode collapse prone data set. The performance of the trans-former models strongly depends on the data set characteristics,it struggled to synthesise data sets with high temporal andspatial correlations. Discriminators with recurrent networkarchitectures suffer from vanishing gradients. We also show,that the performance of the generative adversarial networksdepends more on the discriminator rather than the generator.