With the explosive growth of atmospheric data, machine learning models have achieved great success in air pollution forecasting because of their higher computational efficiency than the traditional... Show moreWith the explosive growth of atmospheric data, machine learning models have achieved great success in air pollution forecasting because of their higher computational efficiency than the traditional chemical transport models. However, in previous studies, new prediction algorithms have only been tested at stations or in a small region; a large-scale air quality forecasting model remains lacking to date. Huge dimensionality also means that redundant input data may lead to increased complexity and therefore the over-fitting of machine learning models. Feature selection is a key topic in machine learning development, but it has not yet been explored in atmosphere-related applications. In this work, a regional feature selection-based machine learning (RFSML) system was developed, which is capable of predicting air quality in the short term with high accuracy at the national scale. Ensemble-Shapley additive global importance analysis is combined with the RFSML system to extract significant regional features and eliminate redundant variables at an affordable computational expense. The significance of the regional features is also explained physically. Compared with a standard machine learning system fed with relative features, the RFSML system driven by the selected key features results in superior interpretability, less training time, and more accurate predictions. This study also provides insights into the difference in interpretability among machine learning models (i.e., random forest, gradient boosting, and multi-layer perceptron models). Show less
Bosch, M.; Locatelli, M.; Hamel, P.; Remme, R.P.; Chenal, J.; Joost, S. 2021
Mitigating urban heat islands has become an important objective for many cities experiencing heat waves. Despite notable progress, the spatial relationship between land use and/or land cover... Show moreMitigating urban heat islands has become an important objective for many cities experiencing heat waves. Despite notable progress, the spatial relationship between land use and/or land cover patterns and the distribution of air temperature remains poorly understood. This article presents a reusable computational workflow to simulate the spatial distribution of air temperature in urban areas from their land use and/or land cover data. The approach employs the InVEST urban cooling model, which estimates the cooling capacity of the urban fabric based on three biophysical mechanisms: tree shade, evapotranspiration and albedo. An automated procedure is proposed to calibrate the parameters of the model to best fit air temperature observations from monitoring stations. In a case study in Lausanne, Switzerland, spatial estimates of air temperature obtained with the calibrated model show that the urban cooling model outperforms spatial regressions based on satellite data. This represents two major advances in urban heat island modeling. First, unlike in black-box approaches, the calibrated parameters of the urban cooling model can be interpreted in terms of the physical mechanisms that they represent; therefore, they can help promote an understanding of how urban heat islands emerge in a particular context. Second, the urban cooling model requires only land use and/or land cover and reference temperature data and can, therefore, be used to evaluate synthetic scenarios such as master plans, urbanization prospects and climate scenarios. The proposed approach provides valuable insights into the emergence of urban heat islands which can serve to inform urban planning and assist the design of heat mitigation policies. Show less
Graaf S.C. van der; Kranenburg, R.; Segers, A.J.; Schaap, M.; Erisman, J.W. 2020
The nitrogen cycle has been continuously disrupted by human activity over the past century, resulting in almost a tripling of the total reactive nitrogen fixation in Europe. Consequently, excessive... Show moreThe nitrogen cycle has been continuously disrupted by human activity over the past century, resulting in almost a tripling of the total reactive nitrogen fixation in Europe. Consequently, excessive amounts of reactive nitrogen (N-r) have manifested in the environment, leading to a cascade of adverse effects, such as acidification and eutrophication of terrestrial and aquatic ecosystems, and particulate matter formation. Chemistry transport models (CTMs) are frequently used as tools to simulate the complex chain of processes that determine atmospheric N-r flows. In these models, the parameterization of the atmosphere-biosphere exchange of N-r is largely based on few surface exchange measurement and is therefore known to be highly uncertain. In addition to this, the input parameters that are used here are often fixed values, only linked to specific land use classes. In an attempt to improve this, a combination of multiple satellite products is used to derive updated, time-variant leaf area index (LAI) and roughness length (z(0)) input maps. As LAI, we use the Moderate Resolution Imaging Spectroradiometer (MODIS) MCD15A2H product. The monthly z(0) input maps presented in this paper are a function of satellite-derived normalized difference vegetation index (NDVI) values (MYD13A3 product) for short vegetation types (such as grass and arable land) and a combination of satellite-derived forest canopy height and LAI for forests. The use of these growth-dependent satellite products allows us to represent the growing season more realistically. For urban areas, the z(0) values are updated, too, and linked to a population density map. The approach to derive these dynamic z(0) estimates can be linked to any land use map and is as such transferable to other models. We evaluated the sensitivity of the modelled N-r deposition fields in LOng Term Ozone Simulation - EURopean Operational Smog (LOTOS-EUROS) v2.0 to the abovementioned changes in LAI and z(0) inputs, focusing on Germany, the Netherlands and Belgium. We computed z(0) values from FLUXNET sites and compared these to the default and updated z(0) values in LOTOS-EUROS. The root mean square difference (RMSD) for both short vegetation and forest sites improved. Comparing all sites, the RMSD decreased from 0.76 (default z(0)) to 0.60 (updated z(0)). The implementation of these updated LAI and z(0) input maps led to local changes in the total N-r deposition of up to similar to 30% and a general shift from wet to dry deposition. The most distinct changes are observed in land-use-specific deposition fluxes. These fluxes may show relatively large deviations, locally affecting estimated critical load exceedances for specific natural ecosystems. Show less
The Integrated Model to Assess the Global Environment–Global Nutrient Model (IMAGE–GNM) is a global distributed, spatially explicit model using hydrology as the basis for describing nitrogen (N)... Show moreThe Integrated Model to Assess the Global Environment–Global Nutrient Model (IMAGE–GNM) is a global distributed, spatially explicit model using hydrology as the basis for describing nitrogen (N) and phosphorus (P) delivery to surface water, transport and in-stream retention in rivers, lakes, wetlands and reservoirs. It is part of the integrated assessment model IMAGE, which studies the interaction between society and the environment over prolonged time periods. In the IMAGE–GNM model, grid cells receive water with dissolved and suspended N and P from upstream grid cells; inside grid cells, N and P are delivered to water bodies via diffuse sources (surface runoff, shallow and deep groundwater, riparian zones; litterfall in floodplains; atmospheric deposition) and point sources (wastewater); N and P retention in a water body is calculated on the basis of the residence time of the water and nutrient uptake velocity; subsequently, water and nutrients are transported to downstream grid cells. Differences between model results and observed concentrations for a range of global rivers are acceptable given the global scale of the uncalibrated model. Sensitivity analysis with data for the year 2000 showed that runoff is a major factor for N and P delivery, retention and river export. For both N and P, uptake velocity and all factors used to compute the subgrid in-stream retention are important for total in-stream retention and river export. Soil N budgets, wastewater and all factors determining litterfall in floodplains are important for N delivery to surface water. For P the factors that determine the P content of the soil (soil P content and bulk density) are important factors for delivery and river export. Show less
