The irrigation of urban green spaces can strengthen their thermal mitigation through increased evapotranspiration, yet its effectiveness depends on vegetation structure and local microclimatic conditions. Quantifying irrigation-induced cooling across heterogeneous urban parks is therefore essential to support efficient and cost-effective urban water management. This study integrates ground-based microclimatic observations with detailed urban vegetation layers to parameterize the ENVI-met model, with the aim to assess the thermal mitigation potential provided by eight urban parks in the Municipality of Prato (Italy) under four irrigation scenarios and tree-density across two distinct microclimatic conditions during a representative hot summer day.Model validation demonstrated overall good performance in reproducing air temperature (AirT) and relative humidity (RH) at both irrigated (R² = 0.98) and rainfed site (R² = 0.96). During the warmest hours of the day (12:00-16:00), simulated parks belonging to higher tree-density classes showed lower average AirT, with differences of approximately 1.5°C and 0.5°C between CL1 and CL4 in cooler and warmer microclimatic classes, respectively. Similarly, increasing irrigation intensity from the baseline to the most intensive irrigation scenario further reduced temperatures by about 1.5°C in both microclimatic classes. Simulated scenarios combining higher irrigation intensity with parks belonging to higher tree-density classes were associated with a clear shift toward cooler thermal classes, with mean temperatures moving from >35.0°C to <34.0°C. The economic analysis showed that overall, costs increase progressively for smaller incremental cooling gains, indicating declining marginal cost-effectiveness despite continued temperature reductions.

Understanding how photosynthetic carbon (C) is allocated to woody biomass remains a critical gap in predicting forest responses to climate change, especially in cold-limited ecosystems, due to the pervasive lack of comprehensive carbon-based data at the whole-stand level. We applied a multi-proxy approach integrating eddy covariance, process-based modelling, and quantitative wood anatomy to assess C fluxes and stem-level C allocation in two mature boreal stands in Canada—black spruce (Picea mariana Mill.) and jack pine (Pinus banksiana Lamb.)—from 1999 to 2021.At both stands, we found that stem structural C allocation (measured as cell wall area, CWA) was tightly coupled with observed and modelled gross primary productivity (GPP). Modelled non-structural carbohydrates (NSC) dynamics revealed contrasting temporal patterns between species: jack pine showed an immediate response to available NSC and annual CWA, suggesting an active role of NSC in supporting growth under fluctuating environmental conditions. In contrast, black spruce exhibited a delayed effect, suggesting a more passive and buffering role of NSC in stem structural C allocation. Notably, at the jack pine site, extreme cold years corresponded to reduced CWA alongside elevated NSC concentrations, which might indicate a shift in C allocation priorities toward storage over growth. Our findings, based on a multi-proxy approach, provide novel insights into species-specific and possible trade-offs between storage and growth, useful for improving C budget models and adaptive forest management under climate change.

Grasslands are worldwide spread ecosystems involved in the provision of multiple functional services, including biomass production and carbon storage. However, the increasingly adverse climate and non-optimised farm management are threatening these ecosystems. In this study, the original semi-mechanistic remotely senseddriven VISTOCK model, which simulates grass growth as limited by thermal and water stress, was modified and integrated with the RothC model to simulate the ecosystem fluxes. The new model (GRASSVISTOCK) showed satisfactory performance in simulating above-ground biomass (AGB) in dry matter (d.m.) and fractional transpirable soil water (FTSW) along Alps (AGB, RMSE = 85.39 g d.m. m- 2; FTSW, RMSE = 0.21) and Mediterranean (AGB, RMSE = 136.84 g d.m. m- 2; FTSW, RMSE = 0.13) grasslands. Also, GRASSVISTOCK was able to simulate the net ecosystem exchange (NEE - RMSE = 0.03 Mg C ha- 1), the gross primary production (RMSE = 0.04 Mg C ha- 1), the ecosystem respiration (RMSE = 0.04 Mg C ha- 1) and the evapotranspiration (RMSE = 1.44 mm), where these observations were available (Alps). The model was applied under present and two climate datasets characterised by temperature increase and precipitation decrease (+2 degrees C temperature, -10 % precipitation) and reference or enriched CO2 concentration (394 vs. 540.5 ppm) scenarios. The results showed that, while changes in temperature and precipitation alone had a negative impact by increasing NEE (+0.69 Mg C ha- 1) and decreasing total biomass (-0.20 Mg d.m. ha- 1) in the reference CO2 scenario, the enriched atmospheric CO2 concentration partially smoothed the NEE trend (+0.27 Mg C ha- 1) and increased total biomass (+0.60 Mg d.m. ha- 1) compared to the present period. It is concluded that the GRASSVISTOCK model represents a first step towards an integrated tool for estimating the performance of the agro-pastoral systems in terms of biomass production, water and carbon fluxes, in the face of ongoing climate change.

ABSTRACT Free air carbon dioxide enrichment (FACE) offers a unique approach to study the response of crops to rising carbon dioxide (CO 2 ) concentrations under field conditions. We have established a Miglietta‐type FACE facility in Northern Germany (JKI‐FACE) consisting of four rings with CO 2 fumigation and four rings without fumigation with diameters of 15.5 m. Here, we present the technical details and the performance of the JKI‐FACE facility based on thee‐year data (2022–2024) regarding temporal and spatial carbon dioxide distribution. Our results indicated that high‐frequency (1‐min) data of CO 2 concentration was within 10% and 20% of the target concentration (600 ppm) for 84.2% and 95.8% of the time, respectively. Although the rings are relatively close together in our experimental setup, CO 2 drift measurements suggest no significant interference between the rings. These results suggest that JKI‐FACE is a suitable field infrastructure for studying the effects of increasing CO 2 concentrations in cereals. With this FACE facility, we focus on the intraspecific variation of the response of crop yield and quality to elevated CO 2 to make an important contribution to the adaptation of crops and cropping systems to changing climate conditions.

Wheat has a pivotal role in food chains and human diet, and the understanding of its productive and qualitative performance under elevated CO2 (e[CO2]) is therefore a primary research target. While e[CO2] generally boosts wheat yield and biomass, major concerns remain about the impact on qualitative traits. The use of biochar as amendment, thanks to its well-known ability to improve soil fertility and crop production and quality, could be a suitable solution to cope with the negative effect of e[CO2] on qualitative traits. To test this hypothesis, we present the results of an open field experiment investigating the combined effects of biochar application and e [CO2] on quantitative and qualitative performance of two commercial durum wheat (Triticum durum) cultivars (Aureo and Claudio). The experimental design included plots treated without [B0] or with 30 t ha-1 of biochar as amendment [B30] and grown at ambient a[CO2] and e[CO2] (570 ppm). Results indicated that biochar increased soil temperatures by +1 degrees C during winter, thus favouring an earlier phenology and average biomass (+10 % and +7.2 %) and yield (+13.2 % and +8.5 %) increase in both a[CO2] and e[CO2] treatments compared to the no biochar [B0] treatments. Biochar treatments also increased flag leaf chlorophyll and decreased flavonoids, and enhanced intrinsic water use efficiency due to increases in photosynthesis and decreases in stomatal conductance. Spike density, falling number and thousand kernel weight were positively and significantly altered by biochar amendment, whilst the remaining parameters mostly differed between cultivars only. By contrast, biochar application did not alleviate nitrogen reduction in grains and straw in both cultivars neither increase grain protein content, although it remained sufficient for pasta making under all treatment combinations. Accordingly, biochar use could be preferentially adopted for increasing the productivity of high-quality cultivars whose destination and industrial processing would not suffer from a decrease in quality parameters.

Soil microbiome is one of the most important components influencing biogeochemical cycles. Changes in the dominance of different microbial functional groups can result in a community that, due to the changes in microbial enzymes, can respond more or less rapidly to decomposition rates, synthesis of organic matter, nutrient availability and soil structure (Brangarí et al., 2021, Wu et al., 2024). The size and composition of soil microbiome is influenced by variables such as plant species, soil moisture and temperature, pH and nutrients availability (Naylor et al., 2022), which in turn are influenced by climate conditions and agronomic practices. Estimating the soil microbiome composition is therefore crucial to deeper understanding processes such as crop development, carbon (C) and nitrogen (N) uptake, soil nutrient retention, drought tolerance and pest resistance (Lutz et al., 2023).Despite the large importance of soil microbial composition at determining magnitude and patterns of biogeochemical cycles, the majority of crop and biogeochemical models currently existing are not able to well represent this process. For instance, the microbial biomass simulated by STICS (Brisson et al., 1998) and EPIC (Izaurralde et al., 2006) varies according to N availability in the soil organic matter (SOM) decomposition, without considering microbial species dynamics. Similarly, the pools of models such as RothC (Coleman and Jenkinson, 1996), CENTURY (Parton, 1996), APSIM (Probert et al., 1998), DayCent (Parton et al., 1998), FASSET (Berntsen et al., 2003) Report fixed values of C/N ratios.This poor representation is mainly related to the lack of detailed algorithms to simulate, for example, SOM turnover driven by soil microbial biomass, the partitioning of different incorporation of decomposable C pools (i.e., lignin and cellulose) from crop residues into soils, the effect of N deficiency on SOM decomposition, and gas transport in soils. These processes should be incorporated into process-based biogeochemical models as driven by soil microbiome to provide more reliable estimates of C and N while reducing uncertainties.To this end, the RothC submodel implemented within the GRASSVISTOCK model (Leolini et al., 2023) has been improved to take into account seasonal evolution of the soil microbiome and the related effect of agronomic practices. Specifically, new mathematical approaches reproducing the response of microbiota activity to soil temperature and water availability numerically quantify the seasonal trend of the enzymatic activity of the soil microbiota communities (Zhao et al., 2024; Babic et al., 2024; Ghodizadeh et al., 2024) will be integrated within the GRASSVISTOCK model and then validated against a measured available data of the grassland test site in Italy.

The Iridaceae family comprises approximately 1800 species, including Iris pallida Lam., which is widely recognized for its ornamental and aromatic properties and particularly adopted in the perfume industry. In this study, we evaluated the effects of planting density and maturity age on biomass production, morphological traits, rhizome biomass, and orris concrete yield in Iris pallida grown in Tuscany (Italy). The experiment consisted of four agricultural parcels, each one containing six plots arranged to test combinations of two planting densities (low density [LD], 8 plants/m2 and high density [HD], 15 plants/m2) and harvesting age (2, 3, and 4 years). Results indicated that planting density significantly influenced biomass variables—including rhizome, bud, and stem biomass—with the low planting density (LD) exhibiting higher total biomass (5.48 ± 0.59 kg/m2) compared to that observed under high planting density (HD) (1.82 ± 0.54 kg/m2). Orris concrete yield varied significantly across planting densities and harvesting age, consistently favoring LD (0.055 ± 0.01%) over HD (0.045 ± 0.01%). Also, orris concrete yield showed a positive correlation with floral stem number (r = 0.73, p < 0.001), root biomass (r = 0.66, p < 0.01) and floral stem biomass (r = 0.63, p < 0.01), while no significant correlations were found between orris concrete yield and total biomass or rhizome biomass. A shorter production cycle under low-density planting may improve orris concrete yield without compromising biomass productivity.

The impacts of heat stress and air pollution are both related to severe health risks for citizens. Complexity and heterogeneity of urban systems can lead to some residents being more exposed than others, possibly exacerbating social inequalities. Whilst the impacts of heath stress and air pollution on population health are known, their relationship with socioeconomic vulnerability has been less investigated. In this work, an integrated risk assessment framework for a mid-size city (Prato, Italy) was developed by combining information on concurrent hazards (summer heat stress and winter air pollution), socioeconomic vulnerabilities indices (Income deciles and Deprivation Index), and demographic exposure (elderly population fraction). Multiple data sources were merged through a novel approach incorporating observed measurements of air temperature and air pollution (PM10 and PM(2.5 )concentrations) at fine time and spatial resolution through a dense IoT sensor network. Results indicated that i) socioeconomic vulnerability was significantly and positively correlated with summer heat intensity (R > 0.8); ii) lowest and highest income classes experienced lower PM concentrations compared to middle-income classes; iii) the fraction of elderly people associated with low socioeconomic vulnerability was little impacted by heat intensity but mostly exposed to winter air pollution depending on their proximity to highly travelled roadways and industrial activities.

Plants of the genus Lavandula are widely studied for their pharmaceutical and food relevance. The composition of lavender essential oil is primarily genotype-dependent but also influenced by environment, developmental stage, and morphology. This study assessed biomass, morphology, oil yield, and chemical composition of seven cultivars (L. angustifolia Boston Blue, L. angustifolia Dwarf Blue, L. Abrialis, L. Super A, L. Super Z, L. Maime, and L. sumiens) cultivated in Tuscany (Italy) over two growing seasons years (2019-2020 and 2020-2021) at two sites (IT and VR). Most morphological traits were significantly affected by cultivar, site, and year, with IT and lavandin cultivars outperforming VR and true lavender. Cultivar strongly influenced compound concentrations, confirming genetic control. True lavender oils showed distinctive profiles compared to ISO 3515:2002/Cor 1:2004 and the literature: lower linalool (similar to 12.8% vs. 25-38%), higher linalyl acetate (similar to 22.7% vs. 25-45%), negligible camphor (similar to 0%), and very low 1,8-cineole (0.7%). Lavandin oils matched ISO 8902:2009 and the literature for major compounds (1,8-cineole 7%, camphor 8.9%, and linalool 23.4%), except for linalyl acetate (14.2%), below the standard range (20-38%). Overall, cultivar choice significantly shaped essential oil yield and chemical profiles, highlighting genetic and environmental interactions that are crucial for lavender breeding and industrial applications.

The accuracy of Copernicus Atmosphere Monitoring Service (CAMS) European forecasts of PM2.5 and PM10 hourly concentrations was assessed against hourly observations collected from low-cost stations during the 2022-2023 heating season in the Padana Plain (Italy). The intercomparison of all 11 air quality models integrated into the CAMS framework returned root mean square error (RMSE) values ranging 20.3-37.5 (PM2.5) and 22.2-37.8 mu g/m3 (PM10 concentrations), while hourly variation of observations was poorly captured (r = 0.16-0.41 and 0.25-0.47, respectively). Agreeing with prior research, CAMS models exhibited a marked daily variability in forecasting particulate matter (PM) observations, with the largest discrepancies occurring during the early morning and evening hours. PM2.5 observations were best predicted by the CHIMERE model, while PM10 observations by the MINNI model. CAMS Ensemble returned the best r values among all models, while, since all (or the majority of) models over-predicted the observations, it failed to best fit their magnitude, returning mean bias of +8.1 for PM2.5 and +4.0 mu g/m3 for PM10 concentrations. This study demonstrated that further efforts are still needed to improve the performance of CAMS models in estimating PM concentrations. However, rather than acting on model final output, e.g. by implementing biascorrection techniques, a more robust strategy could be to act upstream, i.e. by adjusting the settings of the individual CAMS models. The latter could include a more region-specific characterisation of the emission input data to avoid unrealistic overweighting of anthropogenic emissions, increasing the number of surface stations used for PM concentration assimilation, or adjusting PM chemical composition.

Landfills play a key role as greenhouse gas (GHGs) emitters, and urgently need assessment and management plans development to swiftly reduce their climate impact. In this context, accurate emission measurements from landfills under different climate and management would reduce the uncertainty in emission accounting. In this study, more than one year of long-term high-frequency data of CO2 and CH4 fluxes were collected in two Italian landfills (Giugliano and Case Passerini) with contrasting management (gas recovery VS no management) using eddy covariance (EC), with the aim to i) investigate the relation between climate drivers and CO2 and CH4 fluxes at different time intervals and ii) to assess the overall GHG balances including the biogas extraction and energy recovery components. Results indicated a higher net atmospheric CO2 source (5.7 ± 5.3 g m2 d−1) at Giugliano compared to Case Passerini (2.4 ± 4.9 g m2 d−1) as well as one order of magnitude higher atmospheric CH4 fluxes (6.0 ± 5.7 g m2 d−1 and 0.7 ± 0.6 g m2 d−1 respectively). Statistical analysis highlighted that fluxes were mainly driven by thermal variables, followed by water availability, with their relative importance changing according to the time-interval considered. The rate of change in barometric pressure (dP/dt) influenced CH4 patterns and magnitude in the classes ranging from −1.25 to +1.25 Pa h−1, with reduction when dP/dt > 0 and increase when dP/dt < 0, whilst a clear pattern was not observed when all dP/dt classes were analyzed. When including management, the total atmospheric GHG balance computed for the two landfills of Giugliano and Case Passerini was 174 g m2 d−1 and 79 g m2 d−1 respectively, of which 168 g m2 d−1 and 20 g m2 d−1 constituted by CH4 fluxes.

Two low-cost (LC) monitoring networks, PurpleAir (instrumented by Plantower PMS5003 sensors) and AirQino (Novasense SDS011), were assessed in monitoring PM2.5 and PM10 daily concentrations in the Padana Plain (Northern Italy). A total of 19 LC stations for PM2.5 and 20 for PM10 concentrations were compared vs. regulatory-grade stations during a full “heating season” (15 October 2022–15 April 2023). Both LC sensor networks showed higher accuracy in fitting the magnitude of PM10 than PM2.5 reference observations, while lower accuracy was shown in terms of RMSE, MAE and R2. AirQino stations under-estimated both PM2.5 and PM10 reference concentrations (MB = −4.8 and −2.9 μg/m3, respectively), while PurpleAir stations over-estimated PM2.5 concentrations (MB = +5.4 μg/m3) and slightly under-estimated PM10 concentrations (MB = −0.4 μg/m3). PurpleAir stations were finer than AirQino at capturing the time variation of both PM2.5 and PM10 daily concentrations (R2 = 0.68–0.75 vs. 0.59–0.61). LC sensors from both monitoring networks failed to capture the magnitude and dynamics of the PM2.5/PM10 ratio, confirming their well-known issues in correctly discriminating the size of individual particles. These findings suggest the need for further efforts in the implementation of mass conversion algorithms within LC units to improve the tuning of PM2.5 vs. PM10 outputs.

Outdoor air pollution in urban areas, especially particulate matter (PM), is harmful to human health. Urban trees and shrubs provide crucial ecosystem services such as air pollution mitigation by acting as natural filters. However, urban greenery comprises a particular biodiversity, and different plant species vary in their capacity to accumulate PM. Twenty-two plant species were analyzed and selected according to their leaf traits, the different fractions of PM accumulated on the leaves (large - PML, coarse - PMC, and fine - PMF) and their chemical composition. The study was conducted in four city zones: urban traffic (UT), urban background (UB), industrial (IND), and rural (RUR), comparing winter (W) and summer (S) seasons. The average PM levels in the air and accumulated on the leaves were higher in W than in S season. During both seasons, the highest PM accumulated on the leaves was recorded at the UT zone. Nine species were selected as the most suitable for accumulating PML, seven as the most efficient for accumulating PMC, and six for accumulating PMF. The leaf area and leaf roundness were correlated negatively with PM accumulation. The evergreen species L. nobilis was indicated as suitable for dealing with air pollution based on PM10 and PM2.5 values recorded in the air. Regarding the PM element and metal composition, L. nobilis, Photinia x fraseri, Olea europaea, Quercus ilex and Nerium oleander were selected as species with notable elements and metal accumulation. In summary, the study identified species with higher PM accumulation capacity and assessed the seasonal PM accumulation patterns in different city zones, providing insights into the species interactions with PM and their potential for monitoring and coping with air pollution.

Recent advances in low-cost (LC) sensor technology fostered their deployment in low-income and undersampled countries such as Sub-Saharan Africa (SSA) regions, affected by the highest particulate matter (PM) concentrations and population exposure. The present study is the first addressed in Niamey, Niger, and focuses on assessing LC sensor data and global reanalysis products. Three LC PM2.5 and PM10 monitoring stations were deployed and successfully operated across ∼8 months at different (urban, suburban and rural) locations.Observed PM2.5 and PM10 concentrations revealed consistent patterns, higher during the dry Harmattan season, while appreciably lower during the humid monsoon season. In Niamey, PM2.5 mean concentrations (6.1–20.1 μg/m3) were similar to those observed over higher-income countries, confirming the hypothesis of strictly depending on urbanisation, and thus on anthropogenic activities. Conversely, PM10 concentrations (55.3–142.8 μg/m3) were remarkably higher than most of those measured elsewhere worldwide, and predominantly constituted (81–89%) by coarse fraction. PM10 origin, inferred by backtrajectory analysis, was mainly natural (Saharan dust) during the Harmattan season, and both natural and anthropogenic during the monsoon season.Low-resolution gridded estimations by the Copernicus Atmosphere Monitoring Service (CAMS) were not capable of adequately resolving the spatial variability of PM2.5 and PM10 observations in Niamey, further highlighting the importance of sensor network data to improve air quality knowledge. To tackle the harmful effects of Saharan dust on population, and create robust datasets integrated with gridded products in this challenging region, effort should be put toward creation of trans-national integrated monitoring networks based on LC sensors across SSA.

The present study aimed at testing the benefits of protecting woodchips with an acrylic crusting product developed for the coal energy industry. In the test carried out, four conical wood chips piles were built, two consisting of fresh biomass, the other two of dry wood chips. A fourth larger pile was built as a reference. One dry and one fresh pile were superficially treated with 25 kg of protective acrylic solution diluted in 250 L of water, providing an average application of coating agent of approximately 85 g m−2, while the other two worked as controls. To monitor the piles’ temperature variation, thermal sensors were placed in the inner part of the five piles during their construction. Moisture content (MC) and dry matter (DM) variations in woodchip piles were recorded. The piles treated with the coating agent did not show any significant differences with the untreated piles: in wet material, the protective film slightly reduced the moisture dispersal from the pile from evaporation rather than limiting water intake from rain; in dry material, this confirms the inability of the coating agent to limit water intake from rainfall.

Multi-model ensembles are becoming increasingly accepted for the estimation of agricultural carbon-nitrogen fluxes, productivity and sustainability. There is mounting evidence that with some site-specific observations available for model calibration (with vegetation data as a minimum requirement), median outputs assimilated from biogeochemical models (multi-model medians) provide more accurate simulations than individual models. Here, we evaluate potential deficiencies in how model ensembles represent (in relation to climatic factors) the processes underlying biogeochemical outputs in complex agricultural systems such as grassland and crop rotations including fallow periods. We do that by exploring the correlation of model residuals. We restricted the distinction between partial and full calibration to the two most relevant calibration stages, i.e. with plant data only (partial) and with a combination of plant, soil physical and biogeochemical data (full). It introduces and evaluates the trade-off between (1) what is practical to apply for model users and beneficiaries, and (2) what constitutes best modelling practice. The lower correlations obtained overall with fully calibrated models highlight the centrality of the full calibration scenario for identifying areas of model structures that require further development.

Quercus ilex L. dieback has been reported in several Mediterranean forests, revealing different degree of crown damages even in close sites, as observed in two Q. ilex forest stands in southern Tuscany (IT). In this work, we applied a novel approach combining dendrochronological, tree-ring δ13C and genetic analysis to test the hypothesis that different damage levels observed in a declining (D) and non-declining (ND) Q. ilex stands are connected to population features linked to distinct response to drought. Furthermore, we investigated the impact of two major drought events (2012 and 2017), that occurred in the last fifteen years in central Italy, on Q. ilex growth and intrinsic water use efficiency (WUEi). Overall, Q. ilex showed slightly different ring-width patterns between the two stands, suggesting a lower responsiveness to seasonal climatic variations for trees at D stand, while Q. ilex at ND stand showed changes in the relationship between climatic parameters and growth across time. The strong divergence in δ13C signals between the two stands suggested a more conservative use of water for Q. ilex at ND compared to D stand that may be genetically driven. Q. ilex at ND resulted more resilient to drought compared to trees at D, probably thanks to its safer water strategy. Genotyping analysis based on simple-sequence repeat (SSR) markers revealed the presence of different Q. ilex populations at D and ND stands. Our study shows intraspecific variations in drought response among trees grown in close. In addition, it highlights the potential of combining tree-ring δ13C data with SSR genotyping for the selection of seed-bearing genotypes aimed to preserve Mediterranean holm oak ecosystem and improve its forest management.

The agricultural land represents the most important form of land use, accounting for almost 48% of the European land area. Europe- Despite its relatively small share of global agriculture land total area (9.8%), has been one of the world’s largest and most productive suppliers of food and fibre. Europe accounted for 17.6% of the global cereal production during 2014–2018, and average yields in EU countries were 6.8% higher than the world average in the same period. As this chapter shows the climate change will pose substantial challenge for provisioning but also for other ecosystem services. The role of modelling in managing impacts and developing adaptation strategies is crucial as demonstrated through several examples. As the agroecosystem models are treasure troves of agronomy knowledge, their further development should be seen as on of the research priorities within the continent´s agricultural research.

A pre-deployment calibration and a field validation of two low-cost (LC) stations equipped with O3 and NO2 metal oxide sensors were addressed. Pre-deployment calibration was performed after developing and implementing a comprehensive calibration framework including several supervised learning models, such as univariate linear and non-linear algorithms, and multiple linear and non-linear algorithms. Univariate linear models included linear and robust regression, while univariate non-linear models included a support vector machine, random forest, and gradient boosting. Multiple models consisted of both parametric and non-parametric algorithms. Internal temperature, relative humidity, and gaseous interference compounds proved to be the most suitable predictors for multiple models, as they helped effectively mitigate the impact of environmental conditions and pollutant cross-sensitivity on sensor accuracy. A feature analysis, implementing dominance analysis, feature permutations, and the SHapley Additive exPlanations method, was also performed to provide further insight into the role played by each individual predictor and its impact on sensor performances. This study demonstrated that while multiple random forest (MRF) returned a higher accuracy than multiple linear regression (MLR), it did not accurately represent physical models beyond the pre-deployment calibration dataset, so a linear approach may overall be a more suitable solution. Furthermore, as well as being less computationally demanding and generally more suitable for non-experts, parametric models such as MLR have a defined equation that also includes a few parameters, which allows easy adjustments for possible changes over time. Thus, drift correction or periodic automatable recalibration operations can be easily scheduled, which is particularly relevant for NO2 and O3 metal oxide sensors. As demonstrated in this study, they performed well with the same linear model form but required unique parameter values due to intersensor variability.

Mountain pastures are essential for maintainig biodiversity and local economies. Despite the great value and fragility of these ecosystems, an up-to-date overview of extent and type of alpine pastures is lacking in many areas of the Alps. In this study, the interpretation of ancillary information combined with expeditious field campaigns, and the harmonization of classification methodologies allowed us to: (1) define the spatial extent of mountain pastures; (2) identify the non-grazeable percentage in these areas; (3) Characterize and map pasture types within the Gran Paradiso National Park (Italy), where 4596 ha of grazeable areas were mapped. Among the 13 categories identified, the three most represented in the park are Bare thermophile grasslands (38%), Nardus swards (20%), and Alpine intermediate grasslands (18%). The maps obtained in this study are useful for animal management during the grazing season, and have the capability of geographically assessing potential forage avaibility through modeling and remote sensing data.