The results indicated that Ag+ has rapid and highly toxic effects on cowpea roots at concentrations similar to those which are toxic to freshwater biota. Similarly, high phytotoxicity effects of Ni2+  and Cd2+  have also been reported by Kopittke et al.  and Li et al.,respectively. The highly toxic effects of these ions may be due to their accumulation in the outer cortex and the meristem of roots. They usually compete for biologically active sites with toxic ions, thereby reducing the toxicity of the latter. However, they can also inhibit plant growth at high concentrations. Kopittke et al.  studied the toxic effects of 26 metals on cowpea root and reported a 50% inhibition in root elongation at approximately 14,000 μM for Mg2+, indicating that the toxicity of Mg was far lower than that of toxic ions. Based on the QICAR method, separation of ions on the basis of their classification according to HSAB theory improved the linear regression predictive effect for root toxicity. log {EC50} of soft ions was significantly correlated with eight physicochemical properties. Z, which had the greatest correlation coefficient with log {EC50} for soft ions, was also the most effective variable for establishing predictive equation. Z is one of the auxiliary criteria by which soft and hard ions were distinguished. Generally, the toxicity of soft ions is attributed to soft–soft interactions,and soft receptors are usually characterized by low charge or large radius. Thus, the ion charge is of overriding importance for the toxicity of soft ions. Hard ions exhibited significant correlations between log {EC50} and seven physicochemical properties. Similar to the results from soft ions, the single-variable linear relationship based on Z of hard ions showed the beat prediction of toxicity effects.

In general, the establishment of a good toxicity predictive model requires a large volume of data, grow table which are lacking for many metals. Modeling based on the soft–hard ions classification required more toxicity data for metal elements, which caused difficulties in achieving modeling. In addition, the types and number of metal elements decreased after soft–hard ion grouping. Although R2 was improved, this effect may be the result of spurious correlation due to a reduction in sample number. As a consequence, the classification of soft–hard ions, using QICAR, needs to be verified over a larger number of elements. Alternatively, we explored the relationship between metal toxicity and the physicochemical parameters σCon, HLScale and log K, based on the QICAR method. The parameters showed different predictive effects for metal phytotoxicity. The correlation between toxicity and σCon values was not significant when considering all metal ions,but was improved after excluding La3+ and Mg2+. This indicated that σCon may not explain the toxicity of all ions. Kinraide  investigated the relationship between the σCon values and toxicity of low-valence ions  and found that the combination of charge and σCon could predict toxicity well. However, this relationship was limited to only low-valence metal elements. In the current study, we found that the toxicity predictive effect based on the HLScale method was even worse  than that based on σCon. On the other hand, Kopittke et al.  investigated the relationship between HLScale and rhizotoxicity of 26 metal ions to cowpea, and found that the closeness of fit was good. There are several possible reasons for these disparate results between the two studies. Firstly, roots from different plant species may contain different ligands. The cowpea  used by Kopittke et al.  may contain more of the same ligands that Kinraide  used to calculate HLScale. Thus, the HLScale can predict the toxicity of metals to cowpea roots better than to the roots of the wheat  we used. Secondly, there are limited data on metal species and toxicity involved in the HLScale method.

The HLScale method we used only involves eight metal elements, including two subsequently excluded outliers. This small number of elements has a strong influence on the regression relationship between HLScale and metal toxicity. Thirdly, there are different phytotoxicity mechanisms for different metal elements. Previous studies have pointed out that binding of metal ions to hard ligands is an important, non-specific mechanism that causes toxicity directly by inhibiting the controlled relaxation of the cell wall required for cell elongation. However, this may not apply to all ions. In other words, toxicity may not be associated with the binding strength to the hard ligands for all ions. For example, some soft ions, such as Ag+, may strongly bind to the R-S-functional group  in metallothionein, and in this process, the metal complex is bound to glutathione and sent to the vacuole, thereby exerting toxicity effect. Analogously, Kopittke et al.  investigated the toxicity mechanism of As3+ in cowpea roots and found that the phytotoxicity of As3+ is due to its reaction with dithiol groups on proteins and the inhibition of enzyme reactions that require free sulfhydryl groups. The theory of different phytotoxicity mechanisms for different metal elements is also supported in the current study by the SEM images,whereby the soft metal ions, Cd2+ ) and Ni2+ ), seriously damaged the roots and were more toxic than Cr6+ or Mg2+, indicating that they bound strongly to the soft ligands. On the contrary, Cr6+ ) and Mg2+ ) were less toxic and bound only weakly to the hard ligands, indicating that the binding strength of these metal ions to different ligands varied. Our results confirmed that the K value, based on BLM theory, can predict the phytotoxicity caused by metal elements well. The BLM considers all ligands together, calculating parameters based on toxicity. This parameter  is obtained for a specific plant species tested, so it has more advantages in predicting phytotoxicity. However, the available K values are limited as the BLM involves only a few metal elements. In addition, although some elements, especially those which are essential for plant growth, such as Mg and Ca, have K values, these need to be excluded in the establishment of QICAR due to their low toxicity or non-toxic nature. Further study on the toxicity of different elements toward different organisms is required. A common ligand parameter suitable for all organisms by averaging and normalizing their corresponding K values may further improve the application of this parameter in QICAR.School buildings require high indoor comfort levels because children are sensitive to their environment.

A study in primary schools in the UK found that classroom environment affected academic performance by 25%. Similarly, Chatzidiakou et al. determined that, for temperatures between 20 and 25 ◦C, a reduction of 1 ◦C resulted in a 2–4% increment in learning performance. The majority of the European and South American school buildings stocks were built between the 1960s and 1980s; therefore, it is unsurprising that non-refurbished buildings’ indoor environmental quality  does not comply with current comfort standards and indoor air pollution levels. The educational stock’s high energy use intensity  relates to infrastructure deficiency, only worsened by age, intensive use and lack of maintenance. Many energy efficiency retrofit projects exist for educational stocks, especially in Europe, where combinations of envelope improvements and efficient active equipment achieved near-zero energy goals. However, extrapolating these strategies to free-running stocks is unsuitable as passive strategies should be prioritised. The urgent need for replicable envelope retrofit strategies is known. However, there are scarce studies on passive retrofits for school buildings. Recently, schools began incorporating orchards and farms because of the benefits in children’s development fostered by nature. Rooftop farms  diminish the roof’s heat exchange with the outdoor environment whether because of the increment in thermal mass – soil in green roofs – or the creation of a new external envelope, e.g. in rooftop greenhouses  the building’s roof becomes the greenhouse’s floor. As such, RFs are suitable strategies to improve the energy performance of uninsulated schools. Additionally, schools’ RFs have pedagogical, nutritional and health-related benefits,can foster a circular economy and social cohesion, reduce the environmental impact of food chains, and increase food security. Since the 2000s, the advancement in agricultural technologies sprouted diverse RFs systems as low-tech soil-beds, high-tech controlled greenhouses, and, more recently, building-integrated agriculture. BIA refers to the bi-directional integration of exhaust flows between buildings and farms. RFs are constrained to the roof’s load capacity and require technical solutions for water, energy and waste management not relevant to ground agriculture. On this subject, there is a growing body of literature comparing building-based agriculture to its ground counterpart regarding environmental performance, cost-efficiency, and triple-bottom sustainability. To date, limited studies assess RFs systems from a building energy perspective. Initial studies quantify the performance of a specific farm system, their potential improvement, and the benefits of one-directional integration.

Difficulty in modelling transient heat and mass transfer between plants and their surrounding environment led researchers to neglect these in energy quantifications, to use static loads to represent plants, ebb flow table or to use offline independent models. Furthermore, only a few studies compare the energy and thermal benefits of building-based agriculture. Though these studies provide the first quantification of vegetation effects in buildings, there is a need for a more accurate energy modelling of RFs through the online closed-loop calculation of building and crop energy balances. Crops mathematical models are easily implemented in numeric computing platforms. Therefore, energy co-simulation is the most practical option. In addition, considering the deficiencies in school building stocks and the educational and social benefits of green envelopes, there is the need to evaluate the feasibility and replicability of RFs as passive retrofit strategies. To understand better the energy and thermal benefits of buildingbased agriculture, this study quantifies the improvement in building performance by thermodynamic coupling of different RFs. For this purpose, this study assesses two Ecuadorian archetype free-running schools and three RFs systems: edible green roofs,hydroponic rooftop greenhouses,and thermally integrated rooftop greenhouses. To move forward the simulation of building-based agriculture, this study couples the heat and mass balance of crops with building simulation engines through co-simulation. This cosimulation incorporates crops growth as a time function, includes airflow exchange between farm and building environments, includes the electricity requirements for farms operation, and relies on validated archetype buildings to extrapolate results to educational stocks. The following sections are: 1) theoretical background, 2) methodology description, 3) results and discussion on the three RFs studied, and 4) conclusions.Two real-reference building schools  were used as host buildings for RFs, one for typology A and the other for typology B. Historically listed buildings were not considered because of their specific requirements. These archetypes were selected using an engineering bottom-up urban energy model; a full description of their selection and validation is in Ref.

Fig. 3 depicts the archetypes, and Table 1 describes their envelopes. Archetype A is a building complex with two blocks, one cross-shaped and the other rectangular. Both blocks are single storey, have lightweight construction systems, pitched roofs, and a window-to-wall ratio  of 30%. The pitched roofing in archetype A is incompatible with any RF system. As such, for RFs scenarios evaluation, this was replaced with a flat 10 cm composite deck-slab with a 5 cm cement screed and indoor ceiling. Archetype B is a two-storey L-shaped concrete-frame building. Its classrooms are single-sided, accessible from an exterior hallway, and have a WWR of 21.3%. Schools standard working schedule is from 7 a.m. to 1 p.m. for weekdays and unoccupied during weekends. Summer vacations are from mid-July to the first week of September. Buildings are occupied for a total of 1477 hours annually. The average number of students per classroom is 35, and the children’s median metabolic rate is 74 W/ person. Both archetypes are uninsulated, naturally ventilated, and free-running, i.e. there is no heating, cooling, mechanical ventilation, or domestic hot water. Both schools have LED lighting fixtures on classrooms and teachers’ areas with an average power density of 7.5 W/m2 ; and compact fluorescent lighting  on halls, toilets and storage areas with a power density of 10 W/m2. Appendix A presents a detailed description of internal gains. At normal pressure, archetype A infiltration rate is 0.5 ACH, and Archetype B is 0.4 ACH. These values were taken from a detailed air network simulation of both buildings.For evaluating RFs’ energy performance, three theoretical RFs scenarios on these two real-reference school buildings were assessed using dynamic simulation: edible green roofs,hydroponic rooftop greenhouses,and thermally integrated rooftop greenhouses. Fig. 4 shows these RF scenarios in both case-study schools.