ABSTRACT. NASA’s Earth observing instruments have provided comprehensive observations of wildfires and aerosol plumes from wildfires. At present, JPL’s Segmentation, Instance Tracking, and Data Fusion Using multi-SEnsor imagery (SIT-FUSE), utilizes an unsupervised machine learning (ML) framework that allows users to segment instances of objects like wildfires and smoke plumes in single and multi-sensor scenes from NASA’s satellite instruments with minimal human intervention, in low and no label environments. The output of the ML framework also facilitates the tracking of smoke plumes, allowing users to more easily, but still manually, track plumes across multiple scenes over time. Here, we discuss the approaches and progress being made towards the automation of tracking instances across scenes from the same instrument sets as well as the exploration of techniques like contrastive learning (CL), enhanced by the topological features of the object instances detected, to augment SIT-FUSE with the capability to automatically track wildfire and smoke plume instances across datasets from like and disparate instrument sets.

ABSTRACT. Hidden Markov models (HMMs) provide a flexible framework to model time series data where the observation process, Y, is taken to be driven by an underlying latent state process, Z. HMMs can accommodate multivariate processes by (i) assuming that a single state governs the M observations at time t, (ii) assuming that each observation process is governed by its own HMM, or (iii) a balance between the two, as in the coupled HMM framework. Coupled HMMs assume that a collection of M observation processes is governed by its respective M state processes. However, the state process m at time t, Z[m,t] depends on the collection of state processes Z[-m,t-1]. We introduce spatially-coupled hidden Markov models where the state processes interact according to an imposed neighborhood structure with observations are collected across N spatial locations. We outline an application to short-term forecasting of wind speed using data collected across meteorological stations.

ABSTRACT. The increasing interest in studying long-term trends in the ionosphere can be traced back to the early 1990s, when the global warming observed in the troposphere raised concerns about climate change occurring also in the upper atmosphere. Investigating the ionosphere has since become a significant topic in global change research. Ionospheric trends are linked to anthropogenic sources, such as an increase in greenhouse gas concentrations, and to natural causes as well, such as long-term changes in solar and geomagnetic activity and secular variations in the Earth's magnetic field. However, before estimating ionospheric trends, it is crucial to filter out the effects of solar activity, which account for about 90% and more of ionospheric parameters' variance in most cases. The filtering process can lead to "spurious" trends and thus to incorrect conclusions. In this study, we analyze foF2 data series to examine the effect of different filtering procedures on the determination of long-term trends.

ABSTRACT. It is well established that ground-level ozone is a powerful oxidant that can be harmful to human health, plant health, and is a potent greenhouse gas. Ozone is formed from the photochemical reaction of precursor compounds, including nitrogen oxides, volatile organic compounds, carbon monoxide and methane. Ambient ozone concentrations are highly variable over space and time given the disparate sources of precursor gases, and non-linear creation and destruction processes. As a result, concentrations at a given location are difficult to predict when measurements are not available. Ireland is located on the Atlantic edge of Europe and has a limited network of ozone sampling stations. In the current study, ozone data from 15 active monitoring stations, supplemented with data from an additional 20 passive diffusion tubes deployed during summer 2022, were used in a land use regression analysis. Each sample location was characterised based on potential determinants for ozone concentrations, including land cover, climate, typography, and precursor emissions at a range of distances from each study site. Using a forward and backward stepwise procedure the model was built by including determinants that increased the adjusted R^2 value. By understanding the local drivers of ozone concentrations, we can produce more informed concentration maps for Ireland, and identify where exceedances of thresholds for the protection of human and plant health may occur.

ABSTRACT. Atmospheric nitrogen (N) deposition is one of the leading drivers of plant biodiversity loss globally. As such, there is a need to link atmospheric N deposition to ecological effects, such as biodiversity loss, in a quantitative way that can be used in policy applications related to emissions reductions. One tool used to make this connection is the critical load concept, defined as a quantitative estimate of pollutant deposition below which significant harmful effects on specified sensitive elements of an environment do not occur. Empirical critical loads are based on observed changes in the structure and function of a particular ecosystem and can be based on information gained from a number of different approaches. Several methods to estimate ecological change-points or thresholds have been used to infer empirical critical loads, but it is unclear how those approaches compare (i.e., whether there are systematic differences in the results), as they have not been applied to the same datasets. We aimed to explore those differences by taking four such statistical approaches (visual inspection of gradient categories, Threshold Indicator Taxa Analysis (TITAN), point at which significant reduction can be observed, and linear model with change-point) and applying each to two species abundance datasets (mountain hay meadows in Switzerland and Atlantic oak woodlands in Ireland). Our objectives were to 1) determine the vegetation community species abundance change-points for both habitats using the four approaches, 2) compare the results between methods to determine systematic differences, and 3) assess the advantages and shortcomings of each method and their suitability for informing critical loads. Of the four methods, visual inspection, TITAN, and linear model with change-point produced similar critical load estimates for each habitat, suggesting that there is no systematic difference in the results of those three models.

ABSTRACT. The rising atmospheric concentration of carbon dioxide (CO2) due to human emissions is the leading cause of climate change. Huge quantities of CO2 also cycle between the atmosphere and natural ecosystems, through which the land ecosystems and the ocean together absorb nearly half of what humans emit. However, the natural exchanges (or fluxes) of CO2 are changing, and quantifying the change is difficult since these fluxes occur over enormous geographical areas. Fluxes are transported from Earth's surface around the globe by winds, resulting in atmospheric CO2 concentrations that can be measured from the ground and in the air, including by space-based satellites covering large parts of the globe. This talk will present how we used these data to look for changes to ecosystems' CO2 fluxes using a method we developed called WOMBAT (WOllongong Methodology for Bayesian Assimilation of Trace-gases). We found rapid changes in many ecosystems, including the tropics, which transitioned from a net source of CO2 to a net sink between 2015 and 2020. We also found that the amplitude of the seasonal cycle of global natural flux increased over the study period by 8%, and that the seasonal cycle of natural flux in the northern temperate and northern boreal regions shifted earlier in the year, consistent with expectations for the carbon cycle under a warming climate.

ABSTRACT. Microplastics are plastic particles smaller than 1-5 mm formed by the fragmentation of larger plastics or created by the industry. Laboratory studies, flume experiments, and field sampling indicate that microplastic particles pollute rivers by being transported in river flow and deposited as layers of mixed plastics and sediments. Field sampling in rivers has identified a great dispersion and heterogeneity in the types and amounts of plastic transported and deposited by rivers, making it difficult to identify patterns in its dynamic. For this reason, a hydro-morphodynamical numerical model of a river was created to estimate a river bed where the sediments and microplastic particles interact as a function of the river flow.

The numerical model selected to simulate the transport and deposition of microplastic particles is Delft3D. The study case is a braided river with a computational grid of a width of 2.5 km and a length of 80 km. Three scenarios were used to recreate the microplastic-sediment interactions in the river bed to study the spatial and temporal distribution patterns, morphology changes, and load balances of plastic debris in the river.

The artificial river explains that the sediment bed acts as a source of storage of microplastic near the release point, limiting the availability to be resuspended and transported downstream. The high deposition of microplastic increases the capacity of the river flow to erode the bars and banks channels, resulting in deeper channels and increased river bars. The highest amounts of microplastics deposited are extended areas in the inner curve banks of the main channel, and the highly suspended microplastic load was transported in the thalweg of the main channel. The software simulated the sedimentation, erosion, resuspension, and transportation of microplastics together with sediment particles, predicting a more realistic dynamic, and creating a better method to understand the consequences of this pollutant.

ABSTRACT. Microplastic accumulation in the environment impacts ecosystems, and may adversely affect air quality and human health. Research into the behavior of microplastic particles (MPs) in the environment is growing. However, the entrainment and transport of these particles within turbulent atmospheric boundary-layer flows is highly stochastic and remains under-studied. First attempts to quantify the atmospheric limb of the global plastics cycle have highlighted numerous unknowns including the interplay between drifting mineral particles and the vertical displacement of low-density MPs, as well as influence of wind erosion on MP properties such as size, shape and surface chemistry.

In boundary-layer wind tunnel experiments carried out at Trent University, we have determined that microplastics are preferentially transported by wind compared to sand or soil. The effect of substrate on microplastic entrainment appears less important than MP shape; specifically, the enrichment of microplastic fibers within the entrained particulate matter was found to be one to two orders of magnitude higher than that of microplastic beads. Similarly, MP fibers were found in fall column experiments to settle much more slowly than beads; that is, once they are suspended it is likely that they will reside in the atmosphere for longer periods of time. Microplastic breakdown was further examined in an abrasion chamber by simulating the effects of inter-particle collision with quartz particles on microplastic bead size, shape and surface chemistry. Microbead diameters were reduced by 30−50% over 240−300 h of abrasion, which produced fine fragments, 95% of which were ≤10 μm. Dust particles detached from the quartz grains became embedded in the microbeads changing their surface chemistry.

Our results suggest that (i) microplastic shape needs to be carefully parameterized in models of atmospheric microplastic transport, and (ii) small fragments of microplastic particles produced during aeolian transport lie within the size range associated with human respiration and long-distance transport. As we enter the next phase of our work using high-speed photography and particle tracking velocimetry (PTV) to further refine our measurements of particle behavior within this highly stochastic geophysical system, many challenges lie ahead in developing protocols for processing the enormous volume of data generated for each second of the experiment.

ABSTRACT. The control of pollution received by the environment from natural and anthropogenic sources is of paramount importance for protecting human and ecosystem health. Examples include pollution load from a point source such as the smokestack from a factory and nonpoint sources such as sediments loaded with nutrients from agricultural activities. Two examples will be discussed. The first aims to control the impact of pollution emitted by a factory on the ability of an indicator organism to survive and to reproduce. Here the objective is to determine the level of dilution of the effluent which is safe for the organism. How to account for heterogeneity in count data when setting regulations will be discussed. The other case study deals with estimating the sediment pollution entering and leaving the Niagara River. The problem is estimating the yearly average daily load (ADL) from Lake Erie to Niagara River at Fort Erie (FE) and from the river to Lake Ontario at Niagara on the Lake (NOTL).

The difference between the loads at the two stations represents the differential load. Several ad hoc methods are often used to estimate ADL, without knowing the assumptions required for them to produce reliable estimators. Here these assumptions are specified. Water flow and sediment concentration are used to estimate ADL, where the flow is accurately available for every day of the year but the concentration is available for only a few days within the year. A finite population-based model is presented that predicts the concentrations for the missing days conditional on the available flow and concentration data. The model is used to estimate ADL sediment for both stations during 2010 in which the flow is available for 365 days but the sediment concentration is available for only 25 and 26 days for FE and NOTL respectively. This involves using the predicted sediment concentration for the missing 340 and 339 days to estimate ADL for each station and the differential load between the two stations.

ABSTRACT. The impacts of climate change on hydrological regimes are of great concern. Floods are among the most important hydrological events, in Canada and elsewhere, in terms of human loss and economic costs. In the future, climate change is one of the causes of increasing of flood events in terms of their frequency and intensity. On the other hand, floods are considered as multivariate events described through their peak and volume among other dependent variables/features. Therefore, a multivariate framework is required to deal with the different components of flood trends. In this study, the aim is to study trends in floods over Canada. Over 140 hydrometric stations with records from a joint period of 45 years are analyzed. The spatial trends of flood (peak, volume) series are studied using both univariate and multivariate statistical trend tests. In particular, recently developed multivariate trend tests are applied to better identify the affected components (either volume, peak, their dependence, or any combination of them). In terms of interpretation of the obtained results, meteorological variables, such as temperature and precipitation, are considered to determine the main causes of the detected trends. The results of this study are presented in different maps for a better visualisation. This study can be conducted in other regions and/or other hydrological events, such as droughts, where spatial data are available.

ABSTRACT. In this talk, we present some inference methods in generalized mean-reverting processes with multiple unknown change-points and unknown number of change-points. We also consider the scenario where the drift parameter may satisfy some restrictions. We generalize some recent findings in five ways. First, the established method incorporates the uncertain prior knowledge. Second, we derive the unrestricted estimator (UE) and the restricted estimator (RE) as well as their asymptotic properties. Third, we establish a test for testing the hypothesized constraint and we derive its asymptotic power. Fourth, we propose a class of shrinkage estimators (SEs) which includes as special cases the UE, RE, and classical SEs. Fifth, we study the relative risk dominance of the proposed estimators, and we establish that SEs dominate the UE and the RE performs very well near the hypothesized restriction, but this performs poorly when the restriction is seriously violated. On the top of these interesting contributions, the additional novelty of the established results consists in the fact that the dimensions of the proposed estimators are random. Because of that, the asymptotic power of the proposed test and the asymptotic risk analysis do not follow from classical results in statistical literature. To overcome this problem, we establish an asymptotic result which is useful in its own.

ABSTRACT. Introduction: Lake Erie, the smallest and shallowest of the Laurentian Great Lakes, has experienced an increase in harmful algae blooms (HABs) in recent decades, with significant economic and public health costs. Anthropogenic eutrophication through conversion of land for agriculture, increased pollution, and altered nutrient and carbon loadings is believed to be responsible for this trend. HABs are consequences of excessive and unrestricted autotrophic growth and, thereby, are inherently related to metabolic processes.

Methods: To estimate daily metabolism parameters (GPP, ER, NEP) in the western basin of Lake Erie (WBLE), we utilized continuous limnological and environmental data collected from the Real-time Aquatic Ecosystem Observation Network (RAEON) buoys deployed during May-October 2022. Our study aimed to investigate temporal patterns in daily metabolism parameters and explore possible drivers in nutrients (TDN, NO₃, NH₄, TDP, SRP, DOC) and environmental parameters (light, water temperature, wind speed, total weekly precipitation) using partial least squares regression (PLS).

Results: Our analysis revealed metabolic peaks in July and late August. Cumulative NEP between May and October equates to 0.117 Tg carbon evaded by WBLE. GPP had statistically significant (VIP > 1) negative relationships with TDN, NO₃, light, and positive relationships with NH₄ and water temperature, while ER was not significantly related to any predictor variable.

Conclusion: Our results suggest nutrients are not the sole drivers of metabolic processes in WBLE, indicating the need to consider other factors in developing management strategies. Further research is required to better understand the mechanisms driving metabolism in the lake and inform the development of more effective management approaches.

ABSTRACT. Health risks of air pollution are particularly severe as WHO shows: air pollution is a risk for all-cause mortality as well as for specific diseases. Even if the EEA's Air Quality Index allows people to know the air quality of individual countries, regions and cities in Europe, the index reflects only the potential impact on health of the single pollutant for which concentrations are poorest due to associated health impacts. We know that correlations could represent critical information for determining the appropriate air quality value when more pollutants show poor values, because their interaction could make air quality worse than predicted by the single pollutant, and therefore, a riskier health condition. Moreover, observed volatilities could represent important information given that high variability in the observed values contribute to uncertainty in the determination of the air quality index. Borrowing from financial literature, we focus on a class of Multivariate Generalized AutoRegressive Conditional Heteroskedasticity models originally proposed by Engle (2002) for hedging financial risks. In particular, we consider the extension of time varying volatility models to dynamic multivariate regression models, in which the diagonal elements of the conditional covariance matrix of the errors are modelled as univariate GARCH models, whereas the off-diagonal elements are modelled as nonlinear dynamic functions of the diagonal terms and of the conditional quasi-correlations. As we can observe from data analysis, pollutants concentrations show the presence of significant and different GARCH effects. The objective of the paper is to explore whether the use of a multivariate asymmetric GARCH-DCC model can lead to a more accurate risk prediction for air pollution. In particular, we aim to determine how positive shocks to the observed pollutants can increase health risk. Interesting results emerge for particulate matter and ozone, both of which have great effects on human health.

ABSTRACT. Inland waters are increasingly being recognized as a significant source of carbon (C) to the atmosphere. However, quantifying the fate of C in large lake ecosystems is inherently difficult, limiting our understanding of their role in the global C cycle. Autonomous underwater vehicles (AUVs) that collect real-time high-resolution vertical and lateral water quality data across large spatial scales have been used in marine metabolic studies. Over the last decade, AUVs have been used to conduct observation research through The Consortium of Great Lakes Gliders in the Laurentian Great Lakes. To assess the epilimnetic metabolic rates, we used 7 years of archived high-resolution AUV data from Lake Superior (LS), the world’s largest freshwater lake by surface area. The AUV missions differed in location and span but were biased towards southern Lake Superior and summer. Continuous limnological data collected by AUVs (i.e., dissolved oxygen and temperature) and meteorological data (i.e., wind and solar radiation) were used to estimate daily epilimnetic gross primary production (0.32 ± 0.49 g O2 m-3 day-1) and ecosystem respiration (-0.50 ± 0.60 g O2 m-3 day-1) using established methods. Compiling all these data, LS was net heterotrophic (net ecosystem metabolism -0.17 ± 0.53 g O2 m-3 day-1) while exhibiting some spatial and temporal variation. Although metabolic rates for LS are low compared to many other large lake ecosystems, given the size of LS, the present study reveals that it can have a substantial contribution to the global C cycle.

ABSTRACT. Climate data are essential for analysing and modelling climate variability and trends and their impacts on different health and socio-economic activities. Though, Africa is one of the most vulnerable regions to climate change, such climate studies and applications are very scarce in Africa due to the limited availability and access to climate data. The weather stations are sparse and unevenly distributed across many parts of Africa and suffer from large proportions of missingness over space and time. Thus, the in-situ geostatistical climate data measured directly from monitoring weather stations are assumed to be accurate within its measurement error, but sparse in space and time. Alternatively, physical climate model output provides another source of climate gridded data that cover large and dense spatial and temporal domains at a certain resolution but not at smaller scales. Physical climate model data do not account for the uncertainty in the data and hence tend to exhibit bias compared to the in-situ observations. To enhance the accuracy of climate model outputs and the spatial and temporal coverage of in-situ data, the simulated climate model output data can be calibrated against the real measurements from monitoring weather stations. However, integrating such climate model output data with in-situ observed data for improved interpolation is not trivial as it involves misalignment in space and time which can lead to biases in predictions. In this study, our aim is to present a statistical downscaling framework for combining the monthly maximum temperature observations from 52 monitoring weather stations across the Nile Basin countries throughout the 10-year period (2011 – 2020) with the 44 x 44 km gridded data simulated from a regional climate model (RCM) across the same study region and study period. To accurately account for uncertainty from the different data sources and propagate it to predictions, a spatio-temporal coregionalization model that assumes a joint distribution between the covariate (simulated model output) and the response (in-situ observations) is employed. The proposed spatio-temporal coregionalization model is fitted under a hierarchical Bayesian framework using integrated nested Laplace approximation (INLA) coupled with stochastic partial differential equation (SPDE) approach. This spatio-temporal model assumes that the true underlying process for both sources of data is a Gaussian process. This allows us to flexibly adjust the spatio-temporal latent field of the simulated climate model data, improve the prediction of the monthly maximum temperature data along the entire spatial domain at finer resolution and forecast the monthly maximum temperature for the future. The model predictions are compared against the results of a spatio-temporal generalized additive model (GAM) fitted using only one source of data (in-situ observations) and proved to be relatively better by lowering the root mean square error.

ABSTRACT. Using biomass for bio energy is becoming an increasingly popular solution for reducing carbon intensity in fuels to combat climate change. Canada is aiming to achieve an annual reduction of 30 million tonnes of greenhouse gas emissions by 2030 that includes a Clean Fuel Standard. Here we explore the optimal location for biorefineries in relation to regional biomass supply, using a multi-integer linear programming (MILP) approach. We identify optimal locations for Canadian biorefineries that minimize environmental and socio-economic costs under different supply scenarios to ensure environmental sustainability and economic viability. The scenarios simulate price competition, biomass type specific spatial competition, biomass supply competition, biomass supply demand compatibility, extreme weather, different environmental and social costs, biomass price switching, and finally, an integration of all scenarios. Using available data, we find that that social considerations are crucial and it is challenging to monetize social values and to design constraints that protect communities, because building a biorefinery in close proximity to highly populated communities has the potential to pose health risks due to pollution, and reduce housing and green areas due to spatial needs of the facility. Our modeling provides a way to assess environmental, economic, and societal impacts associated with bioenergy production in a spatially-explicit way.

ABSTRACT. Article presents study of controlling the contents of wastewater by means of automation of the wastewater supply with machine for biological treatment of urban wastewater of Kamianske city - aerated pond. We have investigated the automation of the aeration and wastewater treatment system in an aerated pond by means of equipping the system with a water and silt mixture depth sensing element and an electric motor, which will provide the ability to automatically control the supply of air to the aerator, coordinated with the drainage system to maintain an optimal cleaning mode. Such techniques contribute to an increased reliability of the treatment facilities operation under the conditions of quick fluctuations in the volume of sewage water entering the aerated ponds. The problem was that due to the fluctuations, workers had to manually adjust the sludge flow and the intensity of air supply to the aeration tank every hour. It is presented a daily dynamics of the wastewater flow into the treatment facilities of the left bank of Kamianske city and the hourly change in dissolved oxygen concentration. Also the average of nitrogen and phosphates concentration in the source wastewater from December 2020 to May 2022 is shown. It is found that contents of saline ammonium is decreased from 7.50 mg/dm3 in untreated wastewater to 2.00 mg/dm3 in treated one. The process of wastewater treatment in the aerated pond using a three-dimensional mathematical model was simulated. Thanks to the research the efficiency of wastewater treatment increases, and there is an urgent issue of further research to reduce the concentration of phosphates in such machines to the norms of discharge into the river Dnipro.

ABSTRACT. According to the IPCC and other leading scientific organizations, “it is extremely likely that human influence has been the dominant cause of the observed increase in global temperatures since the mid-20th century.” One gap in the research underlying this assessment is that the statistical relationship between CO2 concentrations and the hourly temperature has not been rigorously investigated. Addressing this gap in the research is challenging because the hourly temperature data are noisy, which makes it difficult to extract the CO2 signal. Yet, this challenge needs to be resolved to advance climate science (including the emerging science of climate attribution) and public policy.

This paper examines the statistical relationship between CO2 concentrations and hourly temperature using data from the Barrow Atmospheric Observatory in Alaska, USA. It is first noted that the average annual temperature at Barrow over the 2015-2020 period was about 3.37 oC higher than in the 1985-1990 period. The analysis employs solar irradiance (a key driver of the weather and climate system), CO2, and temperature data. The data are analyzed using a time-series econometric approach that addresses the data’s heteroskedastic and autoregressive nature, which would otherwise “mask” CO2’s “signal” in the noisy data. The model is estimated using hourly data from 1985 through 2015. The results are consistent with the hypothesis that increases in CO2 concentration levels have consequences for hourly temperature. The model is evaluated using hourly data from January 1, 2016, through December 31, 2021. The model’s out-of-sample hourly temperature predictions are highly accurate. However, this accuracy is significantly degraded if the estimated effects of CO2 on temperature are ignored. The implications for selected global locations, such as Ottawa, Canada, and Beijing, China, are assessed using time-series models that capture the spatial nature of hourly temperatures.