ABSTRACT. Incredible progress has been made in recent years with the vast quantity of accurate and reliable Earth Observations (EO) data and tools to drive innovation and inform complex environmental problems—from climate and socio-economic modeling for disaster risk reduction (DRR) assessments, to ecosystem accounting (EA) that provides a GDP value frame for environmental services. High resolution satellite data allows temporal tracking of environmental trends with multi-terabyte cloud catalogs and AI-supported statistical cloud computing, providing a greater granularity of natural assets for decision-making. Yet, “despite commitments to build resilience, tackle climate change and create sustainable development pathways, current societal, political and economic choices are doing the reverse” (UN DRR, 2022). Creating system of systems—Systems Thinking—is critical, as more data isn’t necessarily moving the lever sufficiently toward better decision-making, achieving international climate goals, nor toward justice and equity. Systems thinking re-frames the perspective: the biggest modeling unknown is the knowledge systems itself that scientists are working within, and how to constellate to other knowledge systems. Bridging barriers between more than just statistics and environmental sciences, but between western and Indigenous sciences, is proving a critical reorientation to action scientific approaches. Working with communities that experience disproportionate effects of climate change and environmental injustice challenges can offer a radically different perspective; when people’s relationship to the land is about more than an essential part of their survival, but as intrinsic to their identities, this offers a shift in worldview that can accommodate a new set point and flourishing innovation. We live in times that demand more than an EO data infrastructure, but a just and relevant knowledge infrastructure, where community-level knowledge guides research and informs policy and decision support tools. Dr. Caudill will talk about Systems Thinking frameworks for de-siloing western sciences, bridging across sectors in society, and co-design principles that constellate worldviews.

ABSTRACT. Ecosystems are governed by dynamic processes such as competition for resources, reproduction and dispersal. These shape their biodiversity and how the system responds to change. Current approaches to modelling ecosystems, especially plants, focus on either describing fine-scale processes for individual species or broad-scale patterns for limited groups of plant functional types.

We will describe a simulation system which models plant species across multiple ecosystem sizes, from patches and small islands to regions and entire continents. These simulated ecosystems support the ability to generate many different types of habitat, as well as reproducing different disturbance scenarios such as climate change, habitat loss and invasion. We can also reproduce examples of real-world species distributions by integrating plant occurrence records and global climate reconstructions to simulate plant species throughout the continent of Africa for the past century.

The system allows us to flexibly explore the dynamics of tens of thousands of species interacting across a continent. The code parallelises efficiently across multiple nodes on high performance computing platforms, and has been scaled up to run on over 1000 cores. It allows us to study the impact of changes to climate, resources and habitat and investigate real-life mechanisms surrounding climate change and biodiversity loss.

ABSTRACT. Digital twins are increasingly important in many domains, including for understanding and managing the natural environment. Digital twins of the natural environment are fueled by the unprecedented amounts of environmental data now available from a variety of sources from remote sensing to potentially dense deployment of earth-based sensors. Because of this, data science techniques inevitably have a crucial role to play in making sense of this complex, highly heterogeneous data. Here, we reflect on the role of data science in digital twins of the natural environment, with particular attention on how resultant data models can work alongside the rich legacy of process models that exist in this domain. We unpick the complex two-way relationship between data and process understanding. By focusing on the interactions, we end up with a template for digital twins that incorporates a rich, highly dynamic learning process with the potential to handle the complexities and emergent behaviors of this important area.

ABSTRACT. We are living at a time of unprecedented environmental change, which we are experiencing through hydrological extremes resulting more frequent floods seperated by longer periods drought. These extremes are usually also accompanied by deteriorations in water quality, a picture made more complex by by the increasing numbers of emerging contaminants of concern (e.g. pharmaceuticals, pesticides, micro plastics etc.). Collectively these impacts can have non-linear environmental, societal and economic impacts. However, knowledge gaps remain in understanding of these processes and their impacts. Water also provides many of the opportunities to develop solutions for the Green Recovery, including resource recovery and energy, delivering on societies net zero ambitions. However, the intelligence on water has become compartmentalised or even siloed across instututions and a complex patchwork of jurisdictions of responsibility.

The digital revolution now provides the opportunity to drive a new paradigm of understanding of water and wastewater, enable smart solutions to reduce greenhouse gas emissions and find within catchment solutions for water and wastewater management, and resource recovery. This presentation will chart the development of the Forth Environmental Resilience Array (Forth-ERA), a digital observatory operating at the catchment scale. Forth-ERA brings together data from state-of-the-art and next generation sensors and satellites to deliver new, rich and highligh heterogeneous data streams. These data streams are being combined with AI and process based modelling that collectively offer the potential to transform our understanding of the management and impact of hydroclimatic extrements, the complex mixture of pollutant cocktails impacting on our environment and opportunities for climatre adaptation. This presentation will provide examples of hopw this approach is transforming our understanding hydroclimatic impacts and the stepping stones to building a digital twin at the catchment scale.

ABSTRACT. How we generate environmental data is changing and continuing to evolve, which means also that the complexity of environmental systems can be studied in greater depth, and hidden connections can be explored. It also means that statistical methods need to evolve to deal with new data streams. It is in this landscape, that we often see, terms such as digital environment, digital twin, and digital earth describing the system we are studying. Focusing on digital twins, a digital twin is defined as a virtual representation of real-world objects or systems; in the environmental context, it can also be phrased as a Digital Earth which is an interactive digital replica of the entire planet that can facilitate a shared understanding of the multiple relationships between the physical and natural environments and society. There are at least three parts in the digital twin: • a model or nested models of the system. We might ask what type of model, how are they being coupled, how is uncertainty represented and quantified? • an evolving set of data relating to the object. There are many questions we might imagine, including what is the sampling frame for the data collection or how do we integrate real time and historic data and • a means of dynamically updating or adjusting the model in accordance with the data A digital twin needs and consumes large volumes of rich and highly heterogeneous data often from sensors. It needs to bring together data driven and process modelling, and it takes a whole system approach. Ultimately it enables a decision pipeline that runs in real- or near real-time ( data acquisition and analysis, model execution, uncertainty quantification and visualisation) that supports the decision-making process. We present a discussion of some of the statistical challenges in developing a digital twin.

ABSTRACT. Time series data is ubiquitous in environmental and general scientific analyses. However, the power of a time series data set is proportional to its length, and its total number of missing observations. Most observational time series gathered for scientific purposes contain missing observations, due to instrumental or environmental issues. Reconstruction (often referred to as imputation in the statistics literature, or interpolation in the signal processing) is a powerful tool to allow for longer spans of available data.

In this talk we will explore recent developments in time series interpolation algorithms, centered around the Hybrid Wiener Filter. A testing framework for interpolators has been developed, which allows the comparison of the performance of different interpolators under provided constraints. Some recent work incorporating convolutional neural network predictors will also be discussed.

ABSTRACT. Missing data is common in environmental studies, especially in water level data, and poses a serious problem for data analysis. Past studies have tried to address the challenges of data gaps using various approaches, notably data imputation, but in the last decade some single imputation methods for handling univariate time series missing values such as water levels were proposed. These imputation methods improve data quality and provide a basis for smooth statistical analysis. In this work, we compare the performance of three single imputation approaches, comprising seasonal decomposition, Kalman smoothing and random methods, on monthly univariate water level data between years 2010-2016 from the Kainji water station on river Niger, Nigeria, using two approaches. The first approach compares the actual and imputed data, while the second approach evaluates the imputation methods using Autoregressive Integrated Moving Averages (ARIMA) models fitted from the dataset with imputed values and compares the imputed values with the fitted values from these models. The best imputation method is determined using the root mean square error (RMSE) and mean absolute percentage error (MAPE) metrics, for which lower values are better. For both approaches, missing data at six levels of missingness 5%, 10%, 20%, 30%, 40% and 50%, were created in the complete data using MCAR, MAR and MNAR missing mechanisms. The results from the two approaches clearly revealed that, irrespective of the missing data mechanism, the best method to impute the missing values is the seasonal decomposition method. The Kalman Smoothing method is also suitable for imputing missing values under the three mechanisms, but only does well for lower amounts of missing data. The random method was worse than the other two methods, hence is not recommended for imputation of missing values in data with similar characteristics to our water level data.

ABSTRACT. Vitousek et al. (1997), examining multiple facets of the domination of Earth’s ecosystems by humans, had pointed out that large scale anthropogenic land use land cover (LULC) changes, could be raising a serious risk that goes beyond climate change, by their impact on the structure and function of ecosystems. Here, a depth and category resolved statistical comparative study of available landscape scale soil data, on moisture (WC), organic matter (SOC) and clay content and the correlations between them, between rubber growing soil–FS (hundred rubber plantations) and forest soil–FS (twenty-one contiguous forests/sacred groves), from a ~700 km long and < 150 km wide strip of land on the south west coast of India is presented.  Rubber (Hevea Brasiliensis) plantations are considered as forests by the FAO in forest resource assessments.

Taking over-the-landscape averages at various depth ranges down to ≳ 1.5 m it was seen that in general RS is drier while being more carbon rich compared to FS; the effect sizes are not very large (Kavila I, Hari B V 2022; see also Singh et al. 2021).  The slope for the WC vs SOC relation, is seen to be positive for FS and negative RS at most depth ranges.  The slope is close to zero at all depths for clay-WC and clay-SOC data.  Fisher's exact test shows that the difference in the distribution across soil taxa, of FS and the contiguous RS is significant (p < 0.01).

Further categorizing RS into plantations set up on forest land (FRS) and cropped land (CRS), CRS is seen to be comparatively more dry and carbon rich than FRS.  Considering the extra dryness of CRS and the possibility that this might be inherited from land cover change from forest to crops, these phenomena and their implications merit more detailed, large scale study. 

References

Kavila, I., Hari, B.V. (2022). A Study of the Impact of Some Land Use Land Cover Changes on Watershed Hydrology. In: Chembolu, V., Dutta, S. (eds) Recent Trends in River Corridor Management. Lecture Notes in Civil Engineering, vol 229. Springer, Singapore. https://doi.org/10.1007/978-981-16-9933-7_13

Singh A. K., Liu W., Zakari S., Wu J., Yang B., Jiang X. J., Zhu X., Zou X., Zhang W., Chen C., Singh R., Nath A. J. A global review of rubber plantations: Impacts on ecosystem functions, mitigations, future directions, and policies for sustainable cultivation. Sci Total Environ. 2021 Nov 20;796:148948

Vitousek, P. M., Mooney H. A., Lubchenco J., and Melillo J. M. (1997), Human domination of Earth's ecosystems, Science, 277, 494–499

ABSTRACT. Atmospheric Nitrogen (N) has been recognized as one of the main influential drivers of biodiversity loss on a global level. In this context it is important to understand to what extent N deposition is impacting Canadian forest habitats and ecosystems. The objective of this study is to assess the impact of N deposition to forest understorey taxa distribution changes and identify change thresholds across N deposition gradients in Canadian forest ecosystems. A method called Threshold Indicator Taxa Analysis (TITAN) was applied to assess taxa that significantly change in abundance along N deposition gradients, identifying the points along each gradient at which the most significant individual changes occurred. Indicator species scores were used to integrate occurrence, abundance and directionality of taxa responses. Community-level thresholds were determined based on the convergence of individual taxa abundance change points, using selected data from Saskatchewan's Provincial Forest Ecosites and presented by forest habitat. Identifying these thresholds will assist to assess N critical loads for each of the habitats, which in turn can be helpful in establishing emissions reduction policies.

ABSTRACT. Models of forest dynamics are an essential tool to predict forest ecosystem responses to global change. These predictions are used to make management and policy decisions to maintain forest health, function, and ecosystem services. Despite myriad existing models of forest dynamics, none were developed to address the specific challenges of modeling forest ecosystems at a range of spatio-temporal scales under global change. Here, we define several objectives related to a model’s ability to generate meaningful predictions of forest dynamics and associated ecological outcomes under global change including: assimilation of and statistical learning from real-time monitoring data; computationally feasible scaling of complex tree-level processes to regional and continental levels; approximation of the outcomes of adaptive forest management; and, the quantification of uncertainty in ecological processes and multifarious forest observations. We then define a dynamical statistical framework that synthesizes observations of forest composition, structure, size, and demography to predict the outcomes of forest dynamics. The framework integrates the well-known McKendrick-von Foerster equation of population dynamics within a broader Bayesian hierarchical model that allows for demographic parameter estimation based on forest inventory observations and accounts for uncertainty in both the applied model and data. The model integrates across individual variability to approximate the growth, mortality, and regeneration of physiologically-structured cohorts of trees---the inferential scale of interest for management. We apply the dynamical framework to predict well-known forest development patterns in both single and multi-species forests using both simulated and real-world data from spruce plantations in Quebec, Canada. The model provides probabilistic predictions of species-specific demographic rates and changes in the size distribution of populations over time. We demonstrate the synthesis of common forest data types to reduce uncertainty in model predictions. We conclude with a discussion of how the dynamical framework can be scaled up to regional or continental levels.

ABSTRACT. The relationships between forest degradation indicators (FDIs), calculated from multispectral satellite imagery stacks, representing time-series, and climate variables affecting forest health have not been fully examined, especially in relation to climate change effects on forest degradation. To ascertain how historic climate variables relate to FDIs, they were statistically related to historic climatic variables to help project FDIs into the future using projected climate variables derived from General Circulation Models. These quantitative relationships were employed to predict how future climate would determine future FDIs in the study area (Renfrew County including Algonquin Park, Ontario, Canada). The impacts of temperature (T), precipitation (P), evapotranspiration (ETo), and moisture availability (estimated from two regimes: MA1 and MA2) changes on forest vegetation conditions were studied from the perspective of temperature and water-deficit effects. To understand the quantitative relationships between climate variables and FDIs, multiple linear regression (MLR) analysis was employed to predict future dynamic changes in forest cover leading to the creation of forest degradation projection maps. Forest cover alterations—derived from a computed composite forest degradation indicator (CFDI)—was also monitored from 1972–2020. The relationships between remote sensing indicators and climate variables led to the creation of spatially-explicit maps of relevant climatic variables and long-term historical forest degradation maps. P, MA1, and MA2 had the strongest correlation (-0.95, -0.91, and -0.93, respectively) with the CFDI. This strong relationship resulted in an MLR model—with a high coefficient of determination, R2 (0.93), and low RMSE (0.28)—used for projecting the CFDI into the future. In 2050, the study area will likely experience a daily average temperature (3.0°C) increase. Yearly precipitation will decrease (73.0 mm), and evapotranspiration (109.5 mm) and moisture deficits (MA1 and MA2; 28.47 mm and 37.60 mm, respectively) will increase, resulting in increased forest degradation as determined by the FDIs.

ABSTRACT. Wildland fires can be viewed as having a “lifetime” that consists of several sequential phases. The specific sequence of phases can vary depending on how a fire is responded to (e.g., full suppression or monitoring) by a fire management agency. We investigate the lifetime distributions of two phases for fully suppressed wildland fires from a study area consisting of a response sector in Ontario’s Northwest Fire Region. The progression of phases from ignition to being under control are examined using multi-state models and joint frailty models. Several fixed and random effects are incorporated into the models, including fire weather variables, the number of fires on the landscape, and seasonality. We identify the utility of our modelling approaches for understanding the factors that drive progression through all the phases of a fire as well as those that only influence specific phases.

ABSTRACT. Getting vaccinated is the best way to help protect people from different variants of the SARS-CoV-2 Virus. In the US COVID-19 vaccines are available for free to everyone 6 months and older. However, several factors including misinformation create vaccine hesitancy and threaten to undercut the advances of the COVID 19 vaccination program. In this article, we demonstrate how different metrics and models of complex network analysis can be used in combination with the topological data analysis (TDA) based network clustering method to assess COVID-19 vaccine acceptance in the US. We evaluate the influence of different socio-demographic and neighboring factors in COVID-19 vaccine acceptance decisions based on exponential random graph models. This analysis is conducted using US county level data from the Centers for Disease Control and Prevention (CDC).

ABSTRACT. Power grids are widely recognized as one of the most complex man-made systems. As power grids constitute the core of most modern critical infrastructures, from communication to transportation to urban supply chain, the analysis of resilience, stability, and functionality of power systems is the key towards enhancing economic growth, security and sustainability of our society as a whole. Power grids are prone to a broad range of disruptive events, e.g., man-made threats, as in the case of terrorist attacks and cyber-threats, as well as from natural hazards such as earthquakes, hurricanes, and severe storms. Understanding the power system response and dynamics under such disruptive events is of critical importance to develop timely and efficient risk mitigation strategies. In this project we focus on assessing power system sustainability under weather-related threats. Nowadays there exists a growing evidence that the number and magnitude of adverse atmospheric events, especially affecting coastal areas, tend to exhibit an upward trend and will likely continue to increase. To evaluate geographically localized resilience properties of a power grid, we first represent the power grid as an edge-weighted graph and then employ the concepts of topological data analysis, particularly, persistent homology to assess the local topology and geometry of the grid under atmospheric hazardous scenarios. The derived topological properties of the grid can be then used as early warning indicators of the power system failure as well as employed for efficient islanding of the vulnerable regions. We illustrate the proposed methodology to the assessment of the impact of natural hazards on the energy system in application to synthetic and real power grids, and discuss the associated socio-economic implications. Consequently, the proposed measures can be integrated into early warning indicators for the power grid’s failure as well as employed for developing timely and efficient risk mitigation strategies.

ABSTRACT. Climate change stands to have a profound impact on human society, and on political and other conflicts in particular. We propose a Bayesian framework to address these challenges. We find that there is a strong and substantial association between temperature anomalies on aggregated material conflicts and verbal conflicts globally. Going deeper, we also find significant evidence to suggest that positive temperature anomalies are associated with social conflict primarily through government-civilian and government-rebel material conflicts, as in civilian protests, rebel attacks against government resources, or acts of state repression. We find that majority of the conflicts associated with cli-mate anomalies are triggered by rebel actors, and others react to such acts of conflict. Our results exhibit considerably nuanced relationships between temperature deviations and social conflicts that have not been noticed in previous studies. Methodologically, the proposed Bayesian framework can help social scientists explore similar domains involving large-scale spatial and temporal dependencies.

ABSTRACT. Monitoring change in species composition, often referred to as species turnover, is an informative way of measuring biodiversity. Generalised dissimilarity modelling is commonly used in ecology to understand species turnover through site-pairwise dissimilarities in species composition, which depends on environmental predictors in monotonic relations under a generalised linear model. However, this approach using predictors to explain spatial variation is unable to accommodate complex spatial dependence, especially when the ecological process underlying dissimilarity is spatially dependent. To this end, we develop a dissimilarity-analysis framework based on spatial generalised linear mixed models. The framework extends the classical model by taking into account more structured dependence and thus introduces spatial dependence among dissimilarities. Monotonic relations are modelled using nonparametric regressions based on shape-constrained Bernstein polynomials. Our modelling approach offers avenues to incorporate expert knowledge into the monotonic relations. Finally, we integrate spatial statistical downscaling into the model to allow for predictors that are only available at coarse resolutions. Our approach, which is based on conditional-probability modelling, naturally accounts for the uncertainty that may arise from the downscaling procedure. This compares favourably with the traditional protocol in which downscaling is used as a pre-processing step when preparing predictor data. We investigate model properties analytically and through simulation studies, and we illustrate our methodology with an analysis of species turnover in Antarctica.

ABSTRACT. In many developing cities, economic growth has led to an increase in atmospheric pollutant emissions, resulting in serious consequences for the environment and human health [1]. Carbon monoxide (CO) emissions, which are released into the environment daily, significantly contribute to the increase in respiratory diseases. The transportation sector is the main source of CO emissions due to the incomplete combustion of gas, oil, gasoline, coal, and oils. In March 2020, the World Health Organization (WHO) declared a global pandemic due to the outbreak and spread of COVID-19. A month later, strict social distancing measures were implemented in Guanajuato, Mexico, and activities decreased in general. It is hypothesized that during the lockdown, CO concentration levels decreased in the state of Guanajuato. Therefore, the main objective of this study is to perform a statistical analysis of CO emission reports in different municipalities of the state of Guanajuato between 2020 and 2021 and compare the data with that of the two previous years. A data collection of CO emissions was carried out using the database of the Secretary of Environment and Territorial Ordering. The municipalities of Celaya, Irapuato, and Salamanca were selected for this study. Statistical computing software RStudio [2] was used to perform a first comparison analysis between each of the months for each year.