Search

Random Forest (RF) is a widely used machine learning algorithm known for its flexibility, user-friendliness, and high predictive performance across various domains. However, it is non-interpretable. This can limit its usefulness in applied sciences, where understanding the relationships between predictors and response variable is crucial from a decision-making perspective. In the literature, several methods have been proposed to explain RF, but none of them addresses the challenge of explaining RF in the context of spatially dependent data. Therefore, this work aims to explain regression RF in the case of spatially dependent data by extracting a compact and simple list of rules. In this respect, we propose S-SIRUS, a spatial extension of SIRUS, the latter being a well-established regression rule algorithm able to extract a stable and short list of rules from the classical regression RF algorithm. A simulation study was conducted to evaluate the explainability capability of the proposed S-SIRUS, in comparison to SIRUS, by considering different levels of spatial dependence among the data. The results suggest that S-SIRUS exhibits a higher test predictive accuracy than SIRUS when spatial correlation is present. Moreover, for higher levels of spatial correlation, S-SIRUS produces a shorter list of rules, easing the explanation of the mechanism behind the predictions.

Spatial mapping of biodiversity is crucial to investigate spatial variations in natural communities. Several indices have been proposed in the literature to represent biodiversity as a single statistic. However, these indices only provide information on individual dimensions of biodiversity, thus failing to grasp its complexity comprehensively. Consequently, relying solely on these single indices can lead to misleading conclusions about the actual state of biodiversity. In this work, we focus on biodiversity profiles, which provide a more flexible framework to express biodiversity through non-negative and convex curves, which can be analyzed by means of functional data analysis. By treating the whole curves as single entities, we propose to achieve a functional zoning of the region of interest by means of a penalized model-based clustering procedure. This provides a spatial clustering of the biodiversity profiles, which is useful for policy-makers both for conserving and managing natural resources and revealing patterns of interest. Our approach is discussed through the analysis of Harvard Forest Data, which provides information on the spatial distribution of woody stems within a plot of the Harvard Forest.

This study presents a comparative analysis of three predictive models with an increasing degree of flexibility: hidden dynamic geostatistical models (HDGM), generalised additive mixed models (GAMM), and the random forest spatiotemporal kriging models (RFSTK). These models are evaluated for their effectiveness in predicting PM$_{2.5}$ concentrations in Lombardy (North Italy) from 2016 to 2020. Despite differing methodologies, all models demonstrate proficient capture of spatiotemporal patterns within air pollution data with similar out-of-sample performance. Furthermore, the study delves into station-specific analyses, revealing variable model performance contingent on localised conditions. Model interpretation, facilitated by parametric coefficient analysis and partial dependence plots, unveils consistent associations between predictor variables and PM$_{2.5}$ concentrations. Despite nuanced variations in modelling spatiotemporal correlations, all models effectively accounted for the underlying dependence. In summary, this study underscores the efficacy of conventional techniques in modelling correlated spatiotemporal data, concurrently highlighting the complementary potential of Machine Learning and classical statistical approaches.

Random Forest (RF) is a well-known data-driven algorithm applied in several fields thanks to its flexibility in modeling the relationship between the response variable and the predictors, also in case of strong non-linearities. In environmental applications, it often occurs that the phenomenon of interest may present spatial and/or temporal dependence that is not taken explicitly into account by RF in its standard version. In this work, we propose a taxonomy to classify strategies according to when (Pre-, In- and/or Post-processing) they try to include the spatial information into regression RF. Moreover, we provide a systematic review and classify the most recent strategies adopted to "adjust" regression RF to spatially dependent data, based on the criteria provided by the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA). The latter consists of a reproducible methodology for collecting and processing existing literature on a specified topic from different sources. PRISMA starts with a query and ends with a set of scientific documents to review: we performed an online query on the 25$^{th}$ October 2022 and, in the end, 32 documents were considered for review. The employed methodological strategies and the application fields considered in the 32 scientific documents are described and discussed. This work falls inside the Agriculture Impact On Italian Air (AgrImOnIA) project.

The air in the Lombardy region, Italy, is one of the most polluted in Europe because of limited air circulation and high emission levels. There is a large scientific consensus that the agricultural sector has a significant impact on air quality. To support studies quantifying the role of the agricultural and livestock sectors on the Lombardy air quality, this paper presents a harmonised dataset containing daily values of air quality, weather, emissions, livestock, and land and soil use in the years 2016 - 2021, for the Lombardy region. The pollutant data come from the European Environmental Agency and the Lombardy Regional Environment Protection Agency, weather and emissions data from the European Copernicus programme, livestock data from the Italian zootechnical registry, and land and soil use data from the CORINE Land Cover project. The resulting dataset is designed to be used as is by those using air quality data for research.

The increasing interest in spatially correlated functional data has led to the development of appropriate geostatistical techniques that allow to predict a curve at an unmonitored location using a functional kriging with external drift model that takes into account the effect of exogenous variables (either scalar or functional). Nevertheless uncertainty evaluation for functional spatial prediction remains an open issue. We propose a semi-parametric bootstrap for spatially correlated functional data that allows to evaluate the uncertainty of a predicted curve, ensuring that the spatial dependence structure is maintained in the bootstrap samples. The performance of the proposed methodology is assessed via a simulation study. Moreover, the approach is illustrated on a well known data set of Canadian temperature and on a real data set of PM$_{10}$ concentration in the Piemonte region, Italy. Based on the results it can be concluded that the method is computationally feasible and suitable for quantifying the uncertainty around a predicted curve. Supplementary material including R code is available upon request.

Air pollution is a great concern because of its impact on human health and on the environment. Statistical models play an important role in improving knowledge of this complex spatio-temporal phenomenon and in supporting public agencies and policy makers. We focus on the class of hierarchical models that provides a flexible framework for incorporating spatio-temporal interactions at different hierarchical levels. The challenge is to choose a model that is satisfactory in terms of goodness of fit, interpretability, parsimoniousness, prediction capability and computational costs. In order to support this choice, we propose a comparison approach based on a set of criteria summarized in a table that can be easily communicated to non-statisticians. Our proposal - simple in principle but articulated in practice - holds true for many environmental phenomena where a hierarchical structure is suitable, a large-scale trend is included and a spatio-temporal covariance function has to be chosen. We illustrate the details of our proposal through a case study concerning particulate matter concentrations in Piemonte region (Italy) during the cold season October 2005-March 2006. From the evaluation of the proposed criteria for our case study we draw some conclusions. First, a model with a complex hierarchical structure is globally preferable to one with a complex spatio-temporal covariance function. Moreover, in the absence of suitable computational resources, a model simple in structure and with a simple covariance function can be chosen, since it shows good prediction performance at reasonable computational costs., Comment: 53 pages, 6 figures

The Commission of the European Union, as well the United States Environmental Protection Agency, have set limit values for some pollutants in the ambient air that have been shown to have adverse effects on human and environmental health. It is therefore important to identify regions where the probability of exceeding those limits is high. We propose a two-step procedure for estimating the probability of exceeding the legal limits that combines smoothing in the time domain with spatial interpolation. For illustration, we show an application to particulate matter with diameter less than 10 microns (PM$_{10}$) in the North-Italian region Piemonte., Comment: Published in at http://dx.doi.org/10.1214/08-EJS252 the Electronic Journal of Statistics (http://www.i-journals.org/ejs/) by the Institute of Mathematical Statistics (http://www.imstat.org)