This contains all the data used in the paper as well as R code to implement the parametric loss f... more This contains all the data used in the paper as well as R code to implement the parametric loss function model for eliciting loss functions
The latest Intergovernmental Panel on Climate Change (IPCC) report estimates that the global mean... more The latest Intergovernmental Panel on Climate Change (IPCC) report estimates that the global mean temperature increase will be up to 5.4°C. One of the most affected areas globally is the Eastern Mediterranean and Middle East (EMME), a wide and highly diverse region, identified as a hotspot that is warming twice as fast compared to the global increase. Climate change, in conjunction with poor air quality are two key factors impacting human health. In this work, we use flexible statistical modeling approaches to quantify the joint effect of temperature, humidity and air quality on human mortality for the city of Thessaloniki in Greece. In this work we briefly expose the statistical methodology we propose, and how it was used to quantify the joint effects of temperature, humidity and air quality over prolonged periods of exposure, on human mortality. More specifically, we utilize data on all-cause but also cause-specific daily mortality and model this as a function of temperature, humidity, ozone, PM10 and nitrogen dioxide – all measured from monitoring stations in Thessaloniki. We implement the well-established framework of Distributed Lag Models (DLMs) as Generalized Additive Models (GAMs), to capture the complex interactions between the aforementioned exposures on the risk of mortality. Such models capture the exposure effect through time and thus enable understanding into how prolonged periods of poor air quality and heat stress affect human health. Results confirm the intuition that exposure to extreme heat and humidity in conjunction with poor air quality significantly increases the risk of mortality. This increase in risk varies considerably by case-of-death and also by age group. We show how the type of air pollutant results in different risk profiles, but also that there are correlations across the pollutants that affect the risks.
The climate is continually changing; therefore, making appropriate adaptation and mitigation deci... more The climate is continually changing; therefore, making appropriate adaptation and mitigation decisions is essential. Accurate data is paramount for quantifying climate change impacts and performing risk assessments to support decision-making. Nowadays, climate models are the primary tool for understanding and projecting climate variability. However, their outputs systematically differ from observations, especially for climate variables characterized by strong stochasticity (e.g. precipitation). An ongoing problem concerning climate models is the “drizzling” bias. Climate models tend to overestimate the frequency and duration of rainfalls resulting in the underestimation of their intensity and severity. Due to that fact, climate models significantly underestimate consecutive dry days and overestimate wet days. The adjustment of these biases is a critical process that should precede the use of data. This work proposes a novel statistical method, the Q-GAM (quantile generalized additive models), to bias-correct daily precipitation values over Europe. This task is challenging due to the stochasticity that characterizes this variable, especially on high temporal resolution. The Q-GAM method combines Quantile Mapping (QM) and Generalized Additive Models (GAMs). It is an approach that preserves the advantages of the well-established QM and overcomes its limitations using the flexibility of GAMs. Hence, Q-GAM can significantly increase the accuracy of model-simulated rainfall, maintaining its variability and correcting the number of dry days, overcoming the “drizzling” bias. This is critical for robust and reliable impact analysis. The specific aim of this study is to use Q-GAM bias-corrected model projections of precipitation for quantifying future changes in wet and dry spells across Europe. The duration of wet and dry periods is of great importance as dry spells can serve as indicators of drought and affects several aspects of everyday life (e.g. agriculture, health and economy) and builds the intraseasonal structure of water balance. Here, daily rainfall from three EURO-CORDEX climate models is used for the period 1981-2050. The historical period 1981-2005 is for training the bias adjustment method, while the data for 2006-2050 are corrected, and then the projected wet and dry spells are quantified. The future projections are made according to two climate scenarios: the optimistic RCP2.6 and the business-as-usual RCP8.6. The results show that global warming increases rain falling over short periods, likely triggering floods and/or landslides. Additionally, the improvement of dry spell probabilities detection results in an increasing trend of the duration and severity of drought in future. Furthermore, at a seasonal scale, dry spells are enhanced in future summers and are a temporal extension to transitional and moderate seasons.
The urban heat island (UHI) is a well-known effect where the temperature is higher in a city comp... more The urban heat island (UHI) is a well-known effect where the temperature is higher in a city compared to a rural area, defined as the temperature difference between an urban and a rural location. This is a challenging phenomenon because it exacerbates the heat stress on human health in addition to the on-going global warming. It is therefore important to understand temporal changes in the UHI effect in a warming climate. To examine the UHI effect intensity and variability, we use 40 years (1980-2019) of observational data (daily maximum, minimum, and mean temperature) from the Global Summary of the Day (GSOD), consisting of about 1000 stations of varying temporal extend, spanning the Middle East and North Africa (MENA) region where a faster warming rate has occurred than the other regions globally. The challenge in using data with such spatial and temporal extend is the need to allow for heterogeneities between each of the two stations used to define an urban-rural pair. For instance, one has to allow for the distance between the pair, elevation differences, spatial changes in urbanization as well as the distance from the coast. Here a new method is proposed based on flexible statistical methods (Generalized Additive Models –GAMs) to quantify the temporal trend in the UHI effect while allowing for the aforementioned characteristics using regression splines of appropriately defined variables. A composition of high-resolution satellite geospatial information related to urbanisation properties and population data was utilized from the Global Human Settlement (GHSL-SMOD), in order to characterise the stations in terms of their urbanization type. This was also used to quantify changes in the extend of urbanisation in an area surrounding the rural stations. Results indicate a consistently upward trend of the UHI effect, particularly at night time (Tmin), across all four seasons.
Recent approaches for large-scale mapping of continuous environmental variables by combining grou... more Recent approaches for large-scale mapping of continuous environmental variables by combining ground observations, remote sensing and machine learning have proposed incorporating computer vision capabilities into the model, so that potentially-complex regression features may be learned automatically from covariate datasets, such as of terrain elevation and other satellite imagery (e.g. see Kirkwood et al 2022; 'Bayesian deep learning for spatial interpolation in the presence of auxiliary information').Here we present new research using national-scale land-surface geochemical data to explore and compare how the incorporation of computer vision for automatic feature learning affects the predictive performance of geostastistical interpolators both within and beyond the spatial extents of the study areas in which ground observations are collected. We attempt to characterise empirically how well the predictive performance of different models is preserved with increasing distance from training observations in order to provide insights into the value of incorporating computer vision capabilities into geostatistical models, compared to more traditional approaches.
<p>The analysis of climate change impacts involves the utilisation of clima... more <p>The analysis of climate change impacts involves the utilisation of climate model output. Quite often, quantities of interest are compound events rather than “raw variables” such as temperature. Questions such as "what is the probability that temperature will exceed a high threshold for five consecutive days and how will this change in the future?" are quite common. Statistical (probabilistic) modelling of climate model output can be used to answer such questions by stochastically simulating the raw variables and then quantifying the compound events as a “by-product”. This is particularly useful since any compound event can be investigated using the same approach – since the raw variables are the ones being modelled.</p><p>Such approaches however do not always scale well with big data sets and are often too complicated to even interpret appropriately. Here we present a way of analysing such data, using the (well-established) idea of a ‘moving window’ in conjunction with penalised smoothing splines and Generalised Additive Models (GAMs). The probabilistic nature of the resulting predictions provides a way of extrapolating beyond the range of the original data to robustly quantify the likelihood of rare events and their future changes. The approach is implemented in the Bayesian framework which results in full quantification of the associated uncertainty in using this method, e.g. increased uncertainty for extreme events way outside the range of the original data.</p><p>The method is both scalable and paralleliseable and we present it in quantifying changes in regional climate model output. Due to the simplicity of the components that make up the approach, it can be argued that it is highly interpretable as well as robust to the choice of variables – we demonstrate this using temperature as well as humidity and precipitation, variables which are known to have very different statistical behaviour. We also demonstrate how the approach can be extended to capture the behaviour of more that one variable and use it to quantify the changes in compound hazard events such as the frequency of “warm-dry” days.</p>
This contains all the data used in the paper as well as R code to implement the parametric loss f... more This contains all the data used in the paper as well as R code to implement the parametric loss function model for eliciting loss functions
The latest Intergovernmental Panel on Climate Change (IPCC) report estimates that the global mean... more The latest Intergovernmental Panel on Climate Change (IPCC) report estimates that the global mean temperature increase will be up to 5.4°C. One of the most affected areas globally is the Eastern Mediterranean and Middle East (EMME), a wide and highly diverse region, identified as a hotspot that is warming twice as fast compared to the global increase. Climate change, in conjunction with poor air quality are two key factors impacting human health. In this work, we use flexible statistical modeling approaches to quantify the joint effect of temperature, humidity and air quality on human mortality for the city of Thessaloniki in Greece. In this work we briefly expose the statistical methodology we propose, and how it was used to quantify the joint effects of temperature, humidity and air quality over prolonged periods of exposure, on human mortality. More specifically, we utilize data on all-cause but also cause-specific daily mortality and model this as a function of temperature, humidity, ozone, PM10 and nitrogen dioxide – all measured from monitoring stations in Thessaloniki. We implement the well-established framework of Distributed Lag Models (DLMs) as Generalized Additive Models (GAMs), to capture the complex interactions between the aforementioned exposures on the risk of mortality. Such models capture the exposure effect through time and thus enable understanding into how prolonged periods of poor air quality and heat stress affect human health. Results confirm the intuition that exposure to extreme heat and humidity in conjunction with poor air quality significantly increases the risk of mortality. This increase in risk varies considerably by case-of-death and also by age group. We show how the type of air pollutant results in different risk profiles, but also that there are correlations across the pollutants that affect the risks.
The climate is continually changing; therefore, making appropriate adaptation and mitigation deci... more The climate is continually changing; therefore, making appropriate adaptation and mitigation decisions is essential. Accurate data is paramount for quantifying climate change impacts and performing risk assessments to support decision-making. Nowadays, climate models are the primary tool for understanding and projecting climate variability. However, their outputs systematically differ from observations, especially for climate variables characterized by strong stochasticity (e.g. precipitation). An ongoing problem concerning climate models is the “drizzling” bias. Climate models tend to overestimate the frequency and duration of rainfalls resulting in the underestimation of their intensity and severity. Due to that fact, climate models significantly underestimate consecutive dry days and overestimate wet days. The adjustment of these biases is a critical process that should precede the use of data. This work proposes a novel statistical method, the Q-GAM (quantile generalized additive models), to bias-correct daily precipitation values over Europe. This task is challenging due to the stochasticity that characterizes this variable, especially on high temporal resolution. The Q-GAM method combines Quantile Mapping (QM) and Generalized Additive Models (GAMs). It is an approach that preserves the advantages of the well-established QM and overcomes its limitations using the flexibility of GAMs. Hence, Q-GAM can significantly increase the accuracy of model-simulated rainfall, maintaining its variability and correcting the number of dry days, overcoming the “drizzling” bias. This is critical for robust and reliable impact analysis. The specific aim of this study is to use Q-GAM bias-corrected model projections of precipitation for quantifying future changes in wet and dry spells across Europe. The duration of wet and dry periods is of great importance as dry spells can serve as indicators of drought and affects several aspects of everyday life (e.g. agriculture, health and economy) and builds the intraseasonal structure of water balance. Here, daily rainfall from three EURO-CORDEX climate models is used for the period 1981-2050. The historical period 1981-2005 is for training the bias adjustment method, while the data for 2006-2050 are corrected, and then the projected wet and dry spells are quantified. The future projections are made according to two climate scenarios: the optimistic RCP2.6 and the business-as-usual RCP8.6. The results show that global warming increases rain falling over short periods, likely triggering floods and/or landslides. Additionally, the improvement of dry spell probabilities detection results in an increasing trend of the duration and severity of drought in future. Furthermore, at a seasonal scale, dry spells are enhanced in future summers and are a temporal extension to transitional and moderate seasons.
The urban heat island (UHI) is a well-known effect where the temperature is higher in a city comp... more The urban heat island (UHI) is a well-known effect where the temperature is higher in a city compared to a rural area, defined as the temperature difference between an urban and a rural location. This is a challenging phenomenon because it exacerbates the heat stress on human health in addition to the on-going global warming. It is therefore important to understand temporal changes in the UHI effect in a warming climate. To examine the UHI effect intensity and variability, we use 40 years (1980-2019) of observational data (daily maximum, minimum, and mean temperature) from the Global Summary of the Day (GSOD), consisting of about 1000 stations of varying temporal extend, spanning the Middle East and North Africa (MENA) region where a faster warming rate has occurred than the other regions globally. The challenge in using data with such spatial and temporal extend is the need to allow for heterogeneities between each of the two stations used to define an urban-rural pair. For instance, one has to allow for the distance between the pair, elevation differences, spatial changes in urbanization as well as the distance from the coast. Here a new method is proposed based on flexible statistical methods (Generalized Additive Models –GAMs) to quantify the temporal trend in the UHI effect while allowing for the aforementioned characteristics using regression splines of appropriately defined variables. A composition of high-resolution satellite geospatial information related to urbanisation properties and population data was utilized from the Global Human Settlement (GHSL-SMOD), in order to characterise the stations in terms of their urbanization type. This was also used to quantify changes in the extend of urbanisation in an area surrounding the rural stations. Results indicate a consistently upward trend of the UHI effect, particularly at night time (Tmin), across all four seasons.
Recent approaches for large-scale mapping of continuous environmental variables by combining grou... more Recent approaches for large-scale mapping of continuous environmental variables by combining ground observations, remote sensing and machine learning have proposed incorporating computer vision capabilities into the model, so that potentially-complex regression features may be learned automatically from covariate datasets, such as of terrain elevation and other satellite imagery (e.g. see Kirkwood et al 2022; 'Bayesian deep learning for spatial interpolation in the presence of auxiliary information').Here we present new research using national-scale land-surface geochemical data to explore and compare how the incorporation of computer vision for automatic feature learning affects the predictive performance of geostastistical interpolators both within and beyond the spatial extents of the study areas in which ground observations are collected. We attempt to characterise empirically how well the predictive performance of different models is preserved with increasing distance from training observations in order to provide insights into the value of incorporating computer vision capabilities into geostatistical models, compared to more traditional approaches.
<p>The analysis of climate change impacts involves the utilisation of clima... more <p>The analysis of climate change impacts involves the utilisation of climate model output. Quite often, quantities of interest are compound events rather than “raw variables” such as temperature. Questions such as "what is the probability that temperature will exceed a high threshold for five consecutive days and how will this change in the future?" are quite common. Statistical (probabilistic) modelling of climate model output can be used to answer such questions by stochastically simulating the raw variables and then quantifying the compound events as a “by-product”. This is particularly useful since any compound event can be investigated using the same approach – since the raw variables are the ones being modelled.</p><p>Such approaches however do not always scale well with big data sets and are often too complicated to even interpret appropriately. Here we present a way of analysing such data, using the (well-established) idea of a ‘moving window’ in conjunction with penalised smoothing splines and Generalised Additive Models (GAMs). The probabilistic nature of the resulting predictions provides a way of extrapolating beyond the range of the original data to robustly quantify the likelihood of rare events and their future changes. The approach is implemented in the Bayesian framework which results in full quantification of the associated uncertainty in using this method, e.g. increased uncertainty for extreme events way outside the range of the original data.</p><p>The method is both scalable and paralleliseable and we present it in quantifying changes in regional climate model output. Due to the simplicity of the components that make up the approach, it can be argued that it is highly interpretable as well as robust to the choice of variables – we demonstrate this using temperature as well as humidity and precipitation, variables which are known to have very different statistical behaviour. We also demonstrate how the approach can be extended to capture the behaviour of more that one variable and use it to quantify the changes in compound hazard events such as the frequency of “warm-dry” days.</p>
Uploads