Environmental Classification System Research Articles

Study on environmental monitoring classification system based on improved bayesian algorithm

Abstract The commonly used methods for classifying water quality data are based on water quality parameters. The naive Bayesian classification method can be applied to substitute different evaluation criteria and selected water quality parameters for different water bodies. Compared to other water quality data classification methods, naive Bayesian classification methods have advantages such as simple calculation, high classification accuracy, and strong universality. However, this method overlooks the correlation between various water quality parameters and categories. To address the issues of poor universality, computational complexity, and low accuracy of traditional water quality data classification methods, a water quality data classification method based on weighted naive Bayes is proposed. This method comprehensively considers the impact of water quality attributes and their values on the classification results and replaces the original naive Bayes with weighted attribute conditional probabilities to make the classification results as close as possible to the actual category of the sample. The results indicate that this method exceeds 94% accuracy and can be directly utilized as a water quality classification module in water quality monitoring systems.

Enviromic prediction enables the characterization and mapping of Eucalyptus globulus Labill breeding zones

Genotype-environment interaction is pervasive in forest genetics. Delineation of spatial breeding zones (BZs) is fundamental for accommodating genotype-environment interaction. Here we developed a BZ classification pipeline for the forest tree Eucalyptus globulus in 2 Australian regions based on phenotypic, genomic, and pedigree data, as well on a detailed environmental characterization (“envirotyping”) and spatial mapping of BZs. First, the factor analytic method was used to model additive genetic variance and site–site genetic correlations (rB) in stem volume across 48 trials of 126,467 full-sib progeny from 2 separate breeding programs. Thirty-three trials were envirotyped using 145 environmental variables (EVs), involving soil and landscape (71), climate (73), and management (1) EVs. Next, sparse partial least squares-discriminant analysis was used to identify EVs that were required to predict classification of sites into 5 non-exclusive BZ classes based on rB. Finally, these BZs were spatially mapped across the West Australian and “Green Triangle” commercial estates by enviromic prediction using EVs for 80 locations and 15 sets of observed climate data to represent temporal variation. The factor analytic model explained 85.9% of estimated additive variance. Our environmental classification system produced within-zone mean rB between 0.76 and 0.84, which improves upon the existing values of 0.62 for Western Australia and 0.67 for Green Triangle as regional BZs. The delineation of 5 BZ classes provides a powerful framework for increasing genetic gain by matching genotypes to current and predicted future environments.

Enviromic analysis for prediction of Eucalyptus globulus genotype-environment interaction and mapping of breeding zones

Genotype-environment interaction (GE) is pervasive in forest genetics and is a substantial cause of erosion in genetic gain. Delineation of spatial breeding zones (BZs) is fundamental for controlling GE. We developed a BZ classification pipeline for Eucalyptus globulus in 2 Australian regions based on phenotypic, genomic and pedigree data, as well as on a detailed environmental characterization (‘envirotyping’) and spatial mapping of BZs. First, the factor analytic method was used to model additive genetic variance and site-site correlations (rB) in stem volume across 48 trials of 126,467 full-sib progeny from 2 separate advanced-generation breeding programs. Single-step genomic BLUP (ssGBLUP) was used to create a joint relationship matrix combining both programs for factor analytic analysis. Thirty-three trials were envirotyped using 145 environmental variables (EVs), involving soil and landscape (71), climate (73), and management (1) EVs. Next, sparse partial least squares-discriminant analysis (sPLS-DA) was used to identify EVs that were required to predict classification of sites into 5 non-exclusive BZ classes on the basis of rB. These BZs were then spatially mapped across the West Australian (WA) and ‘Green Triangle’ (GT) commercial estates by enviromic prediction using EVs for 80 locations and 15 sets of observed climate data to represent temporal variation. Finally, stem volume breeding value accuracy and genetic gain from ssGBLUP across 40 trials were compared using the new enviromic BZ classification and the previous regional BZ classification. The FA model explained 85.9% of additive genetic variation. Our environmental classification system produced within-zone mean rB between 0.76 and 0.84, which improves upon the existing values of 0.62 for WA and 0.67 for GT as regional BZs. One to two sPLS-DA components were required for each BZ definition, with 5 to 8 EV’s per BZ. The enviromic BZ classification provided for more accurate breeding values and greater stem volume gain. This work raises highly relevant topics for discussion, including a) the drivers of GE in E. globulus, b) methods of collecting and curating environmental data for envirotyping with eucalypts, c) the advantages and challenges of adopting non-exclusive (i.e. overlapping) BZs, d) the need for developing a phenological system for eucalypts, e) accommodating temporal environmental variation in GE analysis, and f) breeding and improvement strategies for a changing climate.

Willingness to Pay for Implementation of an Environmentally Friendly Pharmaceutical Policy in Finland-A Discrete Choice Experiment Study.

The use of pharmaceuticals is their main pathway to the environment, making the public a major stakeholder in environmentally friendly pharmaceutical policies, including an environmental classification system for medicines. We studied the Finnish adult population’s (n = 2030) preferences and willingness to pay (WTP) for an environmentally friendly pharmaceutical policy by means of an online survey employing a discrete choice experiment (DCE). We also studied the relative importance of the policy attributes, namely, the environmental impact, geographical scope, available information about the environmental impact of a pharmaceutical, and the effect of the respondents’ general environmental attitudes on the WTP. The total annual WTP of the Finnish adult population ranges from 37 million to 134 million euros, depending on the attribute levels. Moreover, the environmental attitude of a respondent had a significant impact on the WTP. Generally, the environmental impact of the policy was the most important attribute, the geographical scope of the policy the second, and information about the environmental impact of pharmaceuticals was the third most important attribute. However, the most environmentally friendly respondents preferred information as the second important attribute. This study provides insights into the environmental valuations of the public to be used in preparing new pharmaceutical policy measures.

Recognize the surrounding: Development and evaluation of convolutional deep networks using gammatone spectrograms and raw audio signals

Environmental sound (ES) consists of the surrounding sounds around us. The classification of this kind of sound has specific and significant contributions in many modern world applications, especially in the internet of things (IoT). However, accurately predicting the sound class of an ES is difficult because of the uncertain pattern in most cases. Although a lot of research has been done to develop a fully accurate environmental sound classification (ESC) system using different extracted features from signal, feature selection and classifier design is still a harsh job and sometimes does not guarantee a precise result. In recent years, ESC has attracted the attention of the research community because of the breakthrough in learning algorithms. Following this trend, a study is carried out here to find and propose two CNN models, 1D and 2D, for raw signal input and Gammatone spectrogram input, respectively. Compared with 2D CNN model, 1D CNN for time series waveform is a recent idea. Assessing the models on two various datasets (ESC-10 and US-8 K), the overall accuracy for 2D CNN is found 80.2% (ESC-10), 89% (US-8 K). For 1D CNN, the accuracy is 80.4% (ESC-10), 86% (US-8 K). The performance of the proposed 1D CNN model is almost similar to that of the proposed 2D CNN model. Effective data augmentation procedure is done to achieve this high accuracy by introducing variability in the given dataset, and to reduce overfitting. Finally, the proposed models have a minimal number of parameters to be trained, and floating-point operations per second (FLOPS) are also minimal, especially for 1D CNN.

Environment Classification for Robotic Leg Prostheses and Exoskeletons Using Deep Convolutional Neural Networks.

Robotic leg prostheses and exoskeletons can provide powered locomotor assistance to older adults and/or persons with physical disabilities. However, the current locomotion mode recognition systems being developed for automated high-level control and decision-making rely on mechanical, inertial, and/or neuromuscular sensors, which inherently have limited prediction horizons (i.e., analogous to walking blindfolded). Inspired by the human vision-locomotor control system, we developed an environment classification system powered by computer vision and deep learning to predict the oncoming walking environments prior to physical interaction, therein allowing for more accurate and robust high-level control decisions. In this study, we first reviewed the development of our “ExoNet” database—the largest and most diverse open-source dataset of wearable camera images of indoor and outdoor real-world walking environments, which were annotated using a hierarchical labeling architecture. We then trained and tested over a dozen state-of-the-art deep convolutional neural networks (CNNs) on the ExoNet database for image classification and automatic feature engineering, including: EfficientNetB0, InceptionV3, MobileNet, MobileNetV2, VGG16, VGG19, Xception, ResNet50, ResNet101, ResNet152, DenseNet121, DenseNet169, and DenseNet201. Finally, we quantitatively compared the benchmarked CNN architectures and their environment classification predictions using an operational metric called “NetScore,” which balances the image classification accuracy with the computational and memory storage requirements (i.e., important for onboard real-time inference with mobile computing devices). Our comparative analyses showed that the EfficientNetB0 network achieves the highest test accuracy; VGG16 the fastest inference time; and MobileNetV2 the best NetScore, which can inform the optimal architecture design or selection depending on the desired performance. Overall, this study provides a large-scale benchmark and reference for next-generation environment classification systems for robotic leg prostheses and exoskeletons.

A Finnish Environmental Classification for Building Projects: Experiences and New Developments

There are a lot of international or national environmental classification systems for building projects. The Building Information Foundation RTS noticed a need for national classification system in Finland, where the target was to improve indoor air quality, moisture management, carbon footprint, energy efficiency, material choices and worksite technology and of course to save money in construction processes and maintenance of the premises. A Finnish Environmental Classification system called Green Leadership Tool for construction processes has been developed for managing the actual building process, environmental, economic and sociocultural aspects as described in the CEN/TC 350-standards. The criteria were made with construction industry and published by The Building Information Foundation RTS in the year 2017.The major part of the criteria utilizes existing classifications and certifications developed in Finland as EPD’s, Building Performance Indicators of GBC Finland, the emission classification of building materials and cleanliness classification of HVAC devices (M1-label), Classification of Indoor Environment (CIE2018), Energy Audit (E-Value), Green Factor Tool, Drychain10 and other relevant information provided into market.The tool and criteria are suitable for all property owner’s despite of the background and business model. There are four construction project types included: new building, renovation and partial renovation, enlargement/change of the space and two main buildings types: office/service and residential buildings. The Finnish legislation has been updated on the year 2017, the new LEVEL’s certification criteria have been developed and the new Classification of Indoor Environment 2018 has been updated and consequently there was a demand to update the RTS criteria as well. The classification was launched in the spring 2017 and there are 83 active projects and 140 users in the tool on April 2019. The environmental classification of building processes follows the changes, developments and needs of the market, therefore the criteria will be updated continuously. The new criteria have been used in several projects and experiments of the criteria and projects will be presented.

Classification of Indoor Environment 2018 and Updated Criteria of RTS Environmental Classification

The Finnish Society of Indoor Air Quality and Climate (FiSIAQ) introduced over twenty years ago in 1995 a Classification of Indoor Climate, Construction Cleanliness, and Finishing Materials. After two revised versions, the fourth edition Classification of Indoor Environment was published on May 2018. Based on the criteria set in the classifications, The Building Information Foundation RTS sr started M1-labelling of building products in 1996, named Emission Classification of Building Materials and Cleanliness Classification of Air-handling Components. Indoor air quality and instruments in improving all the sectors to concern are in great importance in Finland. First part of The Classification of Indoor Environment 2018 (CIE 2018) includes new Requirements for Technical Target Values. The other part, Guidance for design and Construction, includes surface structural design, construction site design, Classification of Construction and Construction Work Cleanliness (P1), Classification of Building Materials (M1), Cleanliness Class of Air-handling components (M1) and instructions for designing mechanical systems for buildings. The main changes in the 2018 edition are revised target values for indoor operative temperature, inclusion of a PM2.5 criteria, revision of some acoustical target values and revision of recommended outdoor air flows. Regarding thermal comfort, the target values of the individual indoor climate, S1, will be slightly more demanding than the targets of the S2, good indoor climate, class. The criteria for particulate matter, PM2.5, has been included due to the significant health impact of fine particles. The targets for acoustic environment in offices have been revised to reflect the importance of speech transmission. The recommended outdoor air flows follow the principles of EN15251-2007, considering the common usage of very low-emitting M1-class materials in Finland. The outdoor air flow rates have also been aligned with the recently changed building code values. Companies have developed low-emitting building materials, fixtures, furniture and HVAC-devices during the years to meet the requirements. There are almost 5200 M1-classified low emitting building materials, fixture, furniture and HVAC-devices at the moment (May 2019). The Classification of Indoor Environment 2018 (classes S1, S2) and sub-classifications (P1) and (M1) are integrated part of the new Finnish environmental classification system for building processes called RTS Environmental Classification launched in the spring 2017 and updated in September 2018. The Classification of Indoor Environment has been used in multiple office and residential construction projects during the years. By integrating the CIE 2018 requirements to the RTS 2018, third-party audit and certification of the whole construction project is possible. Therefore, the continuous improvement will be one considerable option.

Revised Finnish classification of indoor climate 2018

The Finnish Society of Indoor Air Quality and Climate (FiSIAQ) introduced a Classification of Indoor Climate, Construction Cleanliness, and Finishing Materials in 1995. The Classification of Indoor Climate has been revised to meet the new Decree on indoor air quality and ventilation, European standards and experience from users of the classification. The most significant change is that target values for concentration and the in/out ratio of fine particles have been added. Other adjustments have been made to ensure good indoor environment and energy efficiency, but with reasonable investments. The criteria for emissions from building material and furniture were also updated. The Building Information Foundation RTS sr has run the M1-labelling of building products since 1996. The voluntary approach has been proven to improve the IAQ in new buildings and to reduce emissions from building materials. The Classification of Indoor Environment 2018 is integrated part of the new RTS Environmental Classification system.

Environment Classification for Urban Big Data Using Deep Learning

With the increased population in urban areas, it has become a challenge to a service provider to manage big data coming from various users and devices. One of the fastest growing devices used by the citizens of urban areas is the smartphone. Today we can find almost nobody without at least one smartphone in their possession. While talking through a smartphone, surrounding environmental noises are added to the speech of the talker. Sometimes these noises are very annoying to the listeners. In this article, we propose a communication framework for urban mobile big data integrating an urban environment classification system. The proposed system utilizes a deep learning approach to classify environmental noises so that an appropriate noise cancellation algorithm can be used to reduce the effect of noise while having a conversation through smartphones. The proposed framework uses mobile edge computing technology to provide low-latency and efficient transmission. Experimental results demonstrate that the proposed system is very efficient in classifying environmental noises.

Mercury Pollution of Lake Imandra Sediments, the Murmansk Region, Russia

Research on the chemical composition of Lake Imandra sediments (including mercury) was carried out. Lake Imandra is known to be the largest one in the Murmansk region, North-West of Russia. It is the subject of intense pollution by the industrial enterprises and municipal wastewater. It was established that chalcophilic Hg metal, being a high-toxic and dangerous one for the lake ecosystem, entered the lake due to activities of apatite–nepheline and copper–nickel industries. In the vertical Hg distribution, within sediments of the stations which were close to the drainage of apatite–nepheline deposit development, subsurface maximum of sediment cores from 0.6 up to 2.3 μg/g at the depths from 7 up to 14 cm was marked. That was ten times higher than the average background content (0.044 μg/g). These values were higher up to an order of magnitude than the Interim freshwater sediment quality guidelines (ISQG) and 5 times higher than the Probable effect level (PEL) for Hg according to Canadian Sediment Quality Guidelines for the Protection of Aquatic Life (2018) ( http://st-ts.ccme.ca/en/index.html ). The Bolshaya Imandra should be classified as “Class III” with the marked contamination according to the Environmental Classification System of the Norwegian Pollution Control Authority (Classification of environmental quality in fjords and coastal waters. A guide. Norwegian Pollution Control Authority, 1997). Hg content was in a wide range from 0.02 up to 1.0 μg/g in the surface of 1-cm sediment layer. The maximum Hg concentrations were confined to the water areas of the effluents from apatite–nepheline and copper–nickel industries.

Assessing the Yield and Load of Contaminants with Stream Order: Would Policy Requiring Livestock to Be Fenced Out of High-Order Streams Decrease Catchment Contaminant Loads?

Catchment contaminant loads vary with stream order as catchment characteristics influence inputs and in-stream processing. However, the relative influence and policy significance of these characteristics across a number of contaminants and at a national scale is unclear. We modeled the significance of catchment characteristics (e.g., climate, topography, geology, land cover), as captured by a national-scale River Environment Classification (REC) system, and stream order in the estimation of contaminant yields. We used this model to test if potential regulation in New Zealand requiring livestock to be fenced off from large (high)-order streams would substantially decrease catchment contaminant loads. Concentration and flow data for 1998 to 2009 were used to calculate catchment load and yields of nitrogen (N) and phosphorus (P) species, suspended sediment, and at 728 water quality monitoring sites. On average, the yields of all contaminants increased with increasing stream order in catchments dominated by agriculture (generally lowland and pastoral REC land cover classes). Loads from low-order small streams (<1 m wide, 30 cm deep, and in flat catchments dominated by pasture) exempt from potential fencing regulations accounted for an average of 77% of the national load (varying from 73% for total N to 84% for dissolved reactive P). This means that to substantially reduce contaminant losses, other mitigations should be investigated in small streams, particularly where fencing of larger streams has low efficacy.

Correspondence between zooplankton assemblages and the Estuary Environment Classification system

We tested whether variability in zooplankton assemblages was consistent with the categories of estuarine environments proposed by the ‘Estuary Environment Classification’ system (EEC) (Hume et al., 2007) across a variety of North Island, New Zealand, estuaries. The EEC classifies estuaries in to eight categories (A to F) based primarily on a combination of three abiotic controlling factors: ocean forcing, river forcing and basin morphometry. Additionally, we tested whether Remane's curve, which predicts higher diversities of benthic macrofauna and high and low salinities, can be applied to zooplankton assemblages. We focused on three of the eight EEC categories (B, D and F), which covered the range of estuaries with river inputs dominating (B) to ocean influence dominating (F). Additionally, we included samples from river (FW) and sea (MW) to encompass the entire salinity range. Zooplankton assemblages varied across the categories examined in accordance with a salinity gradient predicted by the EEC. Three groups of zooplankton were distinguishable: the first formed by the most freshwater categories, FW and B, and dominated by rotifers (primarily Bdelloidea) and estuarine copepods (Gladioferans pectinatus), a second group formed by categories D and F, of intermediate salinity, dominated by copepods (Euterpina acutifrons), and a final group including the purely marine category MW and dominated also by E. acutifrons along with other marine taxa. Zooplankton diversity responded to the salinity gradient in a manner expected from Remane's curve. The results of this study support others which have shown salinity to be the main factor driving zooplankton community composition and diversity.

Assessing the effect of management changes and environmental features on the spatio- temporal pattern of fire in an African Savanna: Fire spatio-temporal pattern

Spatio-temporal variability of the fire regime in the Kruger National Park (KNP) has been analyzed for the 1957–2011 period. Generalized linear mixed models have been used to assess the variability of fire metrics, such as the burnt area, the fire frequency and intensity for the entire area as well as for the main environmental classification systems (geology, soil, vegetation) and fire management zones. This analysis supports the hypothesis that the spatial pattern of fire regime is strongly influenced by the environmental template defined by climatic, geological, pedological and vegetation features. Moreover, contrary to some previous studies, together with the precipitation trend, fire management changes had a significant role in determining the temporal variability of the fire regime at the scale of KNP, particularly regarding the burnt area and the fire frequency.The complex interaction between the environmental template and management in determining the variability of the fire regime, as underscored by our study, defends the merit of adopting an adaptive planning/management based approach supported by quantitative analytic tools.

FULLY LEARNING CLASSIFICATION SYSTEM AND METHOD FOR HEARING AIDS

A method for operating a hearing aid in a hearing aid system where the hearing aid is continuously learnable for the particular user. A sound environment classification system is provided for tracking and defining sound environment classes relevant to the user. In an ongoing learning process, the classes are redefined based on new environments to which the hearing aid is subjected by the user.

Effects-Directed Analysis of Sediments From Polluted Marine Sites in Norway

The environmental status of two polluted marine sites in Norway was investigated by a combination of target chemical analysis and effect-directed analysis (EDA). The two selected sites, the Grenland area and Oslo harbor, in addition to two reference sites, were classified according to the Norwegian environmental classification system based upon results of the target chemical analyses. The polluted sites were characterized by high levels of metals, polycyclic aromatic hydrocarbons (PAH), and polychlorinated biphenyls (PCB). High levels of organotin compounds were also detected in Oslo harbor. The aryl hydrocarbon receptor (AhR) agonist activity in extracts of sediments from marine sites close to Oslo, Oslo harbor, and Grenland were investigated using the CALUX (chemical-activated luciferase expression) assay, which showed elevated levels of activity. As expected from the history of dioxin release into the Grenland area, the results were highest in this area. The presence of estrogen receptor (ER) and androgen receptor (AR) antagonists was also detected in the sediment extracts. Following fractionation of the sediment extracts, EDA was used to tentatively identify the AhR agonists. The compounds responsible for AhR agonist activity in samples from Oslo harbor were isolated in fraction 13, and to a lesser extent in fractions 9–11. In Grenland, the main activity was found in the more polar fractions, namely fractions 14–18. The AhR agonists identified in Oslo harbor were mainly PAH, while in the Grenland area the compounds identified were mainly nitrogen/oxygen-containing polyaromatic compounds (N/O-PAC).

Adaptive environment classification system for hearing aids

An adaptive sound classification framework is proposed for hearing aid applications. The long-term goal is to develop fully trainable instruments in which both the acoustical environments encountered in daily life and the hearing aid settings preferred by the user in each environmental class could be learned. Two adaptive classifiers are described, one based on minimum distance clustering and one on Bayesian classification. Through unsupervised learning, the adaptive systems allow classes to split or merge based on changes in the ongoing acoustical environments. Performance was evaluated using real-world sounds from a wide range of acoustical environments. The systems were first initialized using two classes, speech and noise, followed by a testing period when a third class, music, was introduced. Both systems were successful in detecting the presence of an additional class and estimating its underlying parameters, reaching a testing accuracy close to the target rates obtained from best-case scenarios derived from non-adaptive supervised versions of the classifiers (about 3% lower performance). The adaptive Bayesian classifier resulted in a 4% higher overall accuracy upon splitting adaptation than the minimum distance classifier. Merging accuracy was found to be the same in the two systems and within 1%-2% of the best-case supervised versions.

Evaluation of the accuracy and consistency of the Swedish Environmental Classification and Information System for pharmaceuticals

The Swedish environmental and classification system for pharmaceuticals is a voluntary, industry-owned system with the purpose to provide environmental information about active pharmaceutical ingredients in the Swedish market. In this paper we report the results from a detailed evaluation of the accuracy and consistency of the risk assessments conducted within this system. The evaluation focused on the following three aspects: 1) comparison of the companies' risk assessments with the classification system's own guidance document, 2) how the risk assessments are affected if additional effect data is used, and 3) the consistency of different risk assessments for the same pharmaceutical substance. The analyses show that the system's guidance gives no criteria for when to consider a study “long-term” or “short-term”, and that this confusion affected the outcome of the risk assessments in some cases. Furthermore, when the system's guidance document is followed and the risk assessment was supplemented with effect data from the open scientific literature, then the risk classification for a substantial number of the evaluated substances was altered. Our analyses also revealed that in some cases risk assessors disagree on the outcome of the assessment for the same active pharmaceutical ingredient. Finally we list some recommendations to improve the classification system. The recommendations include clarifying concepts and instructions in the guidance document, introduction of a standardized way of reporting data to the website, and promotion of use of non-standard test data when considered the most relevant.

Experiences with the Swedish environmental classification system of pharmaceuticals on www.fass.se

Experiences with the Swedish environmental classification system of pharmaceuticals on www.fass.se

Pharmaceutical management through environmental product labeling in Sweden

There is an increased awareness that medicinal products for human use may cause negative effects in the environment. In Sweden a voluntary environmental classification system for drugs has been established in collaboration between producers, authorities and the public health care, and used for five years. The idea is to enhance the market demand for medicines with less environmental impact, which in turn will stimulate the producers to design future medicines to be more environmentally friendly. The system is open to the public and based on assessment of the active ingredient in the medicinal product into several classes of risk and hazard, respectively. It is closely related to the EMEA guidelines. Risk is expressed as the ratio between the predicted environmental concentration (PEC) of the active ingredient (AI) and its predicted no effect concentration (PNEC). The hazard is expressed in terms of the AI's persistence, potential to bioaccumulation, and eco-toxicity. Drug data for the classification are delivered by the respective producers. Hitherto more than 300 AI, representing more than 50% of the Swedish volume of drug use, have been classified. Data for risk assessment were missing in 47% of AI. Among drugs with data 7% had a PEC/PNEC ratio > 1, and another 7% had a ratio between 0.1 and 1. The AIs with highest ratio (> 10) were two estrogens. Data for hazard assessment were lacking in 16% of the AI. Among drugs with environmental data 92% were not ready biodegradable, 23% had potential to bioaccumulation, and 61% were toxic to aquatic organisms at a concentration below 1 mg/l. These data are utilized by regional pharmaceutical expert groups when selecting substances to be recommended in public health care in Sweden. They may also be used by prescribing doctors who want to identify the environmentally most favourable substance among several with equivalent medical effect. We conclude that environmental data on human medicinal products are often missing, or reveal unfavourable environmental properties. A proper judgement of the environmental impact of an AI requires a joint evaluation of its risk and hazard. We suggest that the pharmaceutical producers should highlight environmental precaution when designing new AIs, and that the environmental data should be transparent to the general public.