Statistical Inference for Natural Resources and Biodiversity

Addressing environmental challenges and their interconnected aspects represents a critical priority in current scientific research. The rapid decline of biodiversity at global and local scales, the increase in the number of damaged species, the impact of climate change on natural resources, the monitoring of air and water quality, the management of forests and animal populations are just a few of the pressing issues that demand immediate attention. To tackle these challenges effectively, the development of advanced statistical methods as well as the application of enhanced methodologies in the field of environmental and biodiversity research is crucial, and this need for innovation was the focus of the Second International Workshop on “Statistical Inference for Assessing and Monitoring Natural Resources and Biodiversity.” Hosted by the Department of Economics and Statistics at the University of Siena in November 2022, the workshop brought together experts in environmental statistics and quantitative natural sciences with the aim of fostering interdisciplinary collaboration. The event was held under the patronage of the Italian Statistical Society and the GRASPA-SIS group. This Special Issue features a selection of papers presented at the workshop, alongside additional contributions highlighting methodological advancements and applications in the field of environmental and ecological studies. More in detail, the quantitative assessment of biodiversity is addressed in the papers by Golini et al., Altieri et al., and Di Biase et al. The paper by Golini et al. propose to quantify species diversity by means of biodiversity profiles, which are analyzed through functional data analysis, and to apply a penalized model-based clustering procedure. This allows to identify ecological patterns based on species distributions for providing a detailed spatial representation of biodiversity. Altieri et al. illustrate an approach to biodiversity evaluation based on entropy measures, providing new insights into the nesting behavior of two ant species. The implementation of the adopted techniques is supported by the new release of their R package SpatEntropy. Alternatively, in a design-based framework, the paper by Di Biase et al. addresses the estimation of totals of ecological attributes as, commonly, diversity indices can be expressed as suitable functions of totals. In particular, the paper compares the performance of tessellation sampling schemes against schemes specifically tailored for achieving spatial balance to increase the precision of the Horvitz-Thompson total estimator. A design-based approach is also adopted in the paper by Di Biase and Mecatti which delves into the application of Conditional Poisson Sequential Adaptive sampling to rare and spatially clustered animal populations. The considered scheme aims to intensify the detection of cases and provides sample sizes significantly smaller and more stable than the classical Adaptive Cluster sampling while allowing to manage logistics and budget constraints. The monitoring of animal populations is also addressed in the papers by Ketwaroo et al., Panunzi et al., and Pia et al. In the paper by Ketwaroo et al. capture-recapture data on European badgers are used to identify how the dynamics of bovine tuberculosis vary in space and time and as a function of population density. The proposed model integrates disease data from multiple imperfect tests with spatial capture-recapture data to jointly analyze population demographics and disease dynamics in a spatiotemporal framework. Panunzi et al. use available occurrence records from diverse sources to construct a spatial log-Gaussian Cox process for modeling the white shark presence intensity in the Mediterranean Sea. Their approach takes into account data-source specific detection functions, thinning and physical barriers like islands and coastlines. This allows for a high-resolution continuous space map of presence intensity of white sharks, offering insights into their spatial distribution and supporting conservation efforts. A contribution to the management and conservation of fishery resources is given by the paper by Pia et al. who develop a spatially explicit population growth model for annually spawning fish, combining population dynamics and species distribution modeling with data integration approaches. The model analyses data with varying spatial resolutions in a hierarchical Bayesian framework, using local offspring counts and areal fisheries data. Applied to whitefish in the Gulf of Bothnia, it provides insights into how environmental factors influence reproduction at a local scale. In the framework of monitoring and managing natural resources, Koski and Eidsvik introduce advanced sampling design methods for lake management in Finland, focusing on optimizing monitoring efforts. Using a Bayesian spatial logistic regression model for the ecological status of lakes, they rely on functional approximations and Laplace approximation to get closed form expressions for the value of information of a sampling design. The combinatorially difficult design problem is approached by comparing various heuristic algorithms. The paper by Ambrosini et al. addresses the growing importance of synthesizing information from complex networks in ecology and environmental sciences. The authors discuss the use of two alternative methods, similarity network fusion and similarity matrix average, for integrating multiplex similarity networks. The effectiveness of the two methods is evaluated using a real-world dataset on the abundance of microbial species in glacier ecosystems, providing insights into ecological network analysis. The growing importance of remote sensing data in environmental studies is explicit in the papers by Arima et al. and Saarela et al. Arima et al. apply a spatial extension of the Poisson zero-inflated model to wildfire counts at administrative areal units obtained by satellite data, known to be affected by measurement error. Within a Bayesian hierarchical framework, the model incorporates fixed effects and spatially correlated random effects to capture unexplained variability. Spatial autocorrelation is modeled using a conditional autoregressive structure informed by the adjacency of spatial units. The paper by Saarela at el. refines variance estimation within the hierarchical model-based statistical method used by NASA's Global Ecosystem Dynamics Investigation (GEDI) mission to assess forest aboveground biomass in areas with limited GEDI footprint data. The authors propose a bootstrap procedure leveraging a theoretical decomposition of the hierarchical model-based variance into components, enabling their separate and simultaneous estimation, significantly reducing computational time. Finally, two papers focus on air quality and meteorological conditions. In particular, in the paper by Evengelista et al. a multiple function-on-function linear regression is adopted to analyze the impact of local meteorological conditions on PM10 variations in Abruzzo, Italy. The authors propose to integrate functional data analysis with the group Lasso method, employing a principal component regression approach to accommodate multiple functional predictors and functional responses. De Iaco et al. apply a Spatio-Temporal Linear Coregionalization Model (ST-LCM) to describe spatio-temporal correlations between evapotranspiration and meteorological conditions measured over the Veneto Region, Italy. The approach based on the joint diagonalization of the empirical covariance matrix evaluated at different spatio-temporal lags allows to extend the use of ST-LCM to five relevant agrometeorological variables. We would like to warmly thank all the contributors to this special issue for the enthusiasm and the dedication they demonstrated in preparing their manuscripts. We are also deeply grateful to the researchers who served as reviewers; their careful and thoughtful work is sincerely acknowledged. Our genuine thanks also go to Alessandro Fassò, Editor-in-Chief of Environmetrics, for accepting to attend the conference with an invited talk and for hosting this special issue in Environmetrics.