Deforestation has been practiced for tens of thousands of years for agriculture, grazing, cultivation, and urban purpose. However, over the last 33 centuries deforestation has drastically increased, with around 12 million km2 of forests cleared and 40 million km2 remaining today (Ramankutty and Foley, 1999; Klein Goldewijk, 2001; http://www.fao.org/forestry/fra/41256/en/, last access: 22 May 2019).

From a physical perspective, several studies investigated the effects on climate of deforestation, or of its opposite (afforestation), mainly via a modelling approach. These studies compare the effects on climate of changes between current and pre-industrial potential vegetation, under the hypothesis of no human activities. Among its biogeophysical effects on climate, deforestation has contrasting effects on air temperature that depend on the latitude and the vegetation types involved (Claussen et al., 2001; Snyder et al., 2004; Gibbard et al., 2005; Bala et al., 2007; Betts et al., 2007; Jackson et al., 2008; Davin and de Noblet-Ducoudré, 2010; Beltràn-Przekurat et al., 2012). At high latitudes, deforestation triggers a winter and spring surface cooling due to changes in the radiation budget that compensate, at the annual scale, for the summer warming resulting from decreased latent heat flux (i.e. evaporation). In particular in boreal regions, forest removal strongly increases the surface albedo. Indeed forests mask the snow as opposed to herbaceous vegetation (Chalita and LeTreut, 1994; Betts et al., 2001; Meissner et al., 2003; Randerson et al., 2006). At low latitudes, deforestation leads to a surface warming due to changes in the water cycle. Conversion of tropical rainforests to pasture lands (as in the Amazon Basin region) strongly modifies surface evapotranspiration and roughness since, compared to pasture lands, trees have a higher surface roughness that enhances surface fluxes and thus the evapotranspiration cooling efficiency (Shukla et al., 1990; Dickinson and Kennedy, 1992; Lean and Rowntree, 1997; von Randow et al., 2004; Nogherotto et al., 2013; Lejeune et al., 2015; Spracklen and Garcia-Carreras, 2015; Llopart et al., 2018). In the long term, reduced evapotranspiration and precipitation may lengthen the dry season in the tropics, thereby increasing the risks of fire occurrence (Crutzen and Andreae, 1990). At mid-latitudes, both albedo and evapotranspiration mechanisms are at work and compete against each other, as recently confirmed by satellite-based observation analysis (Li et al., 2015; Forzieri et al., 2017). Although studies over the mid-latitudes show somewhat contradictory results and the effect on air temperature (warming/cooling) remains unclear in temperate regions such as the Mediterranean Basin region and Europe (Gaertner et al., 2001; Heck et al., 2001; Anav et al., 2010; Zampieri and Lionello, 2011; Gálos et al., 2013; Stéfanon et al., 2014; Strandberg and Kjellström, 2019), in the Northern Hemisphere, the historical land cover change has very likely led to a substantial cooling (Brovkin et al., 1999, 2006; Bonan, 1997; Betts, 2001; Govindasamy et al., 2001; Bounoua et al., 2002; Feddema et al., 2005a), comparable in magnitude with the impact of increased greenhouse gases (Boisier et al., 2012; de Noblet-Ducoudré et al., 2012). However, a recent study combines present-day observations and state-of-the-art climate simulations and shows that historical deforestation in North America and Eurasia has made the hottest day of the year warmer since pre-industrial times, contributing to at least one-third of the local present-day warming of heat extremes (Lejeune et al., 2018). In addition to modifying mean and extreme temperatures, deforestation–afforestation can also modify the hydrological cycle by enhancing or inhibiting convective clouds and precipitation in the overlying atmospheric column. Some studies show an enhancement of shallow cumulus clouds over deforested lands in Amazonia (Chagnon et al., 2004; Wang et al., 2009), while opposite results were found over deforested lands in southwest Australia (Ray et al., 2003). Two different mechanisms result from the interplay between the surface heat fluxes and the boundary layer structure (i.e. stability, temperature, and humidity): (1) dry soil and high sensible heat flux can increase the entrainment of cold air from the boundary layer top and finally increase shallow cloud cover by lowering the saturation threshold (Westra et al., 2012; Gentine et al., 2013). (2) Conversely, wet soil and high latent heat flux moisten the boundary layer and increase the relative humidity at its top in case of deforestation.

From a biological perspective, deforestation implies modifications in surface moisture and temperature that in turn might directly or indirectly affect decomposition rates and nutrient mineralization in soils (Dominski, 1971; Stone, 1973; Stone, 1979; Classen et al., 2015; Manzoni et al., 2012; Chen et al., 2014; Townsend et al., 2011; Bonan, 2008). As a result, both carbon and nitrogen release to the environment are forecasted to increase. The forest floor decomposes rapidly (Covington, 1976; Bormann and Likens, 1979) and, without forest regeneration, will eventually be partially eroded. The combination of increased decomposition (which consumes oxygen) and wetter soils (which slow oxygen diffusion) may also increase the occurrence of anaerobic microsites within the soils, which might contribute to methane (CH4) emissions (Adji et al., 2014; Jauhiainen et al., 2016). Nitrogen can be lost to the atmosphere through ammonia (NH3) volatilization, nitrous oxide (N2O) production during nitrification (Bremner and Blackmer, 1978; Veldkamp et al., 2009), or denitrification to N2O or atmospheric nitrogen (N2) (Firestone et al., 1980; Neill et al., 2005; Lammel et al., 2015). Soil properties such as soil organic carbon or soil nitrogen cycling respond to deforestation with a large spatial variety from one system to another (Powers and Schlesinger, 2002; Chaplot et al., 2010; de Blécourt 2013). However, the largest emissions of non-CO2 greenhouse gases will probably result from agricultural use and management on deforested areas.

Finally, several studies show that there are feedbacks between tropical forests and climate change (Bonan, 2008). Carbon dioxide fertilization, for example, could have a positive effect by sustaining tropical forest growth (Lapola et al., 2009; Salazar and Nobre, 2010). This is exacerbated by N fertilization since tropical areas are not limited-N environments and N is increasing through atmospheric deposition in non-tropical areas (Magnani et al., 2007; Sutton et al., 2008; Samuelson et al., 2008; Jackson et al., 2009). Zaehle et al. (2011) showed that N inputs increased C sequestration by ecosystems, and Churkina et al. (2007) attributed 0.75–2.21 GtC yr-1 during the 1990s to regrowing forests. However Yang et al. (2010) showed that the contribution of N fertilization is lower for secondary forest regrowth (Jain et al., 2013).

As a direct effect, afforestation inevitably leads to carbon loss from the system (Feddema et al., 2005b; Foley et al., 2005; Le Quéré et al., 2013; Houghton et al., 2012). However, large uncertainties remain on (i) how these altered ecosystems will react to induced global climate change (increased CO2 concentration, increased temperature, etc.), (ii) changes in the emissions of non-CO2 greenhouse gases (N2O, CH4), and (iii) changes in the exchange of reactive trace gases.

From a chemical perspective, afforestation directly affects BVOC emissions since trees are high BVOC emitters, as documented by Purves et al. (2004) over the eastern US by combining a BVOC emission model with vegetation changes as recorded by the USDA Forest Service Inventory Analysis (FIA) over surveyed forest plots. Over the target region, emissions of the main BVOCs (i.e. isoprene and monoterpene) have increased, especially under heatwave conditions (i.e. daily air temperature above 35 ∘C), due to increase in the forest leaf area mainly driven by human disturbance via harvesting and plantation management (i.e. often plantation forestry introduces high emitters), but as well by perturbing ecological succession with fires and pollution. Enhanced BVOC emissions from forests are likely to modify the NOx–VOC–O3 regime; nevertheless the outcome critically depends on the fate of isoprene nitrates, whether they are a terminal or temporal sink of NOx (Val Martin et al., 2015). Concerning fine-mode aerosols, summer levels of PM2.5 (i.e. particulate matter, PM, with aerodynamic diameters ≤2.5 µm) are predicted to increase with afforestation due to the formation of biogenic secondary organic aerosols (BSOAs) from BVOCs (Heald et al., 2008; Trail et al., 2015; Val Martin et al., 2015). As with afforestation, deforestation to create pasture or crop lands can also exacerbate O3 levels by increasing NOx emissions from soil microbial activity, promoted with fertilization (Ganzeveld and Lelieveld, 2004; Trail et al., 2015); in winter, the enhanced NOx levels favour nitrate aerosol production, while in summer deforestation decreases aerosol deposition, by reducing surface roughness. In conclusion, fine-mode aerosols such as PM2.5 may increase year-round under deforestation (Trail et al., 2015).

Under an increasing demand and interest for fast-growing plants for food production, cattle feed, domestic products, and biofuels, plantations are rapidly expanding all over the world. The choice of crop or tree type influences BVOC emissions and the resulting O3 and BSOA levels (Hewitt et al., 2009; Ashworth et al., 2012; Warwick et al., 2013; Stavrakou et al., 2014). This is the case of oil palm crops that show much larger BVOC emission potentials compared to primitive forests (from 3 to 10 times higher for isoprene; Hewitt et al., 2009, and Fowler et al., 2011). In South East Asia, increasing BVOC emissions from oil palm plantations interplay with increasing NOx emissions resulting from the spread in mechanization, fossil fuel use, and fertilizer application associated with the oil palm industry. The complex interaction between BVOC and NOx finally enhances O3 levels at local–regional scales (Goldammer et al., 2009; Hewitt et al., 2009; Silva et al., 2016; Harper and Unger, 2018), with even transboundary effects (i.e. downwind regions) (Warwick et al., 2013). Similarly to South East Asia and oil palm production, the expansion of biofuel production in Europe could modify future LULC to satisfy the increasing demand for renewable energy sources (Beringer et al., 2011). Among biofuel feedstock, crops such as miscanthus or second-generation plantations such as poplar show higher isoprene emission potential compared to European native species. The conversion of European grass- and croplands into biofuel plantations may affect summer O3 levels with effects that strongly depend on the interaction between BVOC and NOx emissions. For example, to limit the effects on O3 production of a steep increase in isoprene emissions (∼45 %) from conversion of 5 % of European grass- and croplands into poplar plantation, NOx emissions should be reduced by 15 %–20 % (Beltman et al., 2013). Regarding Europe, Ashworth et al. (2013) showed that the extension of short-rotation coppice for biofuel feedstock could have small but yet important impacts on surface O3 concentrations, and subsequently on human mortality and crop productivity, since it would modify emitted compounds and their levels. Being BSOA precursors, enhanced BVOC emissions from afforestation are also involved in particulate matter pollution.

Using a large-scale chemistry-transport model for present-day climate, Ashworth et al. (2012) investigated the impact of realistic large-scale scenarios of biofuel feedstock production (∼100 Mha plantations) in both the tropics and the mid-latitudes on isoprene emissions, O3, and BSOA formation. These LULCCs drive an increase in global isoprene emissions of about 1 %, with a substantial impact on regional O3 levels and BSOAs. In the tropics, the expansion of oil palm plantations enhances BSOAs by 0.3 µg m-3 (+3 %–5 %, BSOA annual mean concentrations: 6–10 µg m-3). In the mid-latitudes, the establishment of short-rotation coppice increases BSOA concentrations up to 0.5 µg m-3 (+6 %, from 8 µg m-3).

Although wetland drainage is a relatively small proportion of the world's land surface, LULCC can have significant impacts on some areas. Wetland drainage for agriculture purposes has removed between 64 % and 71 % of natural wetlands since 1900 (Davidson, 2014).

From a physical perspective, only few studies have evaluated its impact on local and regional climate. The most documented case is that of south Florida (Pielke et al., 1999; Weaver and Avissar, 2001; Marshall et al., 2004a, b). During the 20th century, large wetland areas in south Florida were converted to large-scale crops (cereals), citrus growth, and fruit crops in general. Modelling studies show that current surface cover caused significant changes in temperature extremes with increased length of freezing events and increased magnitude of frost (lower temperature), which severely reduced the agricultural production (Marshall et al., 2004a). During night-time, water vapour evaporates from the swamps and modifies the longwave radiation budget, resulting in a less rapid infrared cooling and less cooling by +2 ∘C than for the current (drained) case. A similar study over Switzerland shows opposite results (Schneider and Eugster, 2007). The conversion of wetlands to extensive farming caused a night-time warming and a daytime cooling of a few tenths of a degree Celsius. This temperature modification was explained by the alteration of soil thermal properties and by higher albedo in the current case. During the night-time, higher thermal conductivity of the current soils resulted in upward heat fluxes, which enhanced the temperature. In another vein, Mohamed et al. (2005) studied the effect of Sudd swamp on the Nile water flow and local climate. Due to the Sudd wetland, located in the upper Nile, a substantial amount of water is lost through evapotranspiration. In a drained Sudd scenario produced by a numerical experiment, the Nile flow just downstream the wetlands increases by 46 Gm3 yr-1 over a total of 110 Gm3 yr-1. However, evapotranspiration decreases, causing a temperature increase of +4–6 ∘C during the dry season.

From a biological perspective, the drainage of peatlands and wetlands for agricultural use alters several characteristics of those areas and could thus be problematic (see Verhoeven et al., 2011, for a review). Especially in tropical areas, peatland draining releases some extra CO2 by oxidizing and subsiding peat soils used for growing oil palms (Immirzi et al., 1992; Maltby and Immirzi, 1993; Safford et al., 1998; Furukawa et al., 2005). Hoojer et al. (2006) estimate the emissions from Indonesian peatland draining at 516 Mt C yr-1 (fires excluded). Conversely, since wetlands are a considerable source of CH4, their drainage will decrease emissions of CH4 and can thus be considered a carbon gain from that point of view (Bergkamp and Orlando, 1999; Maltby and Immirzi, 1993). However, this gain is counterbalanced by increased N2O emissions, due to the lowering of the water table (Kasimir-Klemedtsson et al., 1997; Maljanen et al., 2010). On the other hand, changes in vegetation and therefore growth in those drained areas involve an increased carbon sink from vegetation. However, this additional sink rarely compensates for the greenhouse gas (GHG) losses resulting from C losses from the soil (Yeh et al., 2010; Yew et al., 2010).

From a chemical perspective, on top of decreasing CH4 emissions, wetland drainage may probably increase NOx emissions, and modify emissions of other compounds such as BVOCs, due to vegetation change, which together could contribute to significant changes in the atmospheric chemical composition. Overall, the impact of wetland conversion on compound emissions other than CH4 and on atmospheric chemistry has been poorly investigated.

Urbanization results in the replacement of (pseudo-)natural ecosystem vegetation by more or less dense and impervious built-up environments. Human activities concentrated in these areas are responsible for additional heat and gaseous releases in the atmosphere. Consequently, these LULCCs sharply modify the atmosphere, in terms of both climatic conditions and gas composition, which ultimately affect land–atmosphere exchanges and biogeochemical cycles.

From a physical perspective, urbanization results in a modification of surface radiative budget, energy balance, water balance, and land–atmosphere mass and energy exchanges (see Eqs. S1 to S5 in the Supplement), leading ultimately to (local) climate alteration in urban areas.

Firstly, urbanization affects each component of the radiative budget. On the one hand, the net radiation is potentially reduced due to the decrease in the incoming shortwave radiation that is screened out by a reflecting smog layer. In the dry season, in clear skies, Jauregui and Luyando (1999) observed that the incoming solar radiation over Mexico City was 21.6 % lower than its suburbs. This difference could increase up to 30 % under weak winds. However, the intensity of the reduction in incoming shortwave radiation was closely related to the day of the week (i.e. human activities) and meteorology (e.g. temperature, humidity, solar radiation), which both influence photochemical smog formation. Similarly, Wang et al. (2015) measured lower incoming shortwave radiation in Beijing compared to its surroundings, with values ranging between 3 % and 20 % depending on the season. Focusing on summer periods (June, July, August), Li et al. (2018) recorded lower S↓ at urban stations compared to rural stations in the city of Berlin; the authors attributed this dimming effect to the thick aerosol layer observed over the city. Based on the analysis of global radiation measurements from the Global Energy Balance Archive (GEBA), Alpert et al. (2005) and Alpert and Kishcha (2008) showed a relationship between solar dimming, population density, and atmospheric pollution such as aerosols, which absorb and scatter the incoming solar radiation. Overall, Alpert and Kishcha (2008) demonstrated that at the surface shortwave radiation is 8 % lower in urban compared to rural areas. Moreover, the net radiation is also potentially reduced by the enhanced outcoming longwave radiation due to a warmer urban environment (the so-called “urban heat island effect”; see below) since infrared radiations depend on surface temperature. On the other hand, urbanization also induces an increase in net radiation. Urbanization usually results in a decrease in surface albedo (α) and surface emissivities (εs) (Table 1), finally reducing both outgoing short- and longwave radiation. Although some building materials exhibit larger albedo and emissivity than (pseudo-)natural environments, most of them have lower ones, especially asphalt or other dark materials (e.g. Li et al., 2013; Alchapar et al., 2014; Rahdi et al., 2014). Yet, at the city scale, outgoing short- and longwave radiation is scattered and absorbed multiple times within urban canyons (i.e. light-trapping effect), thus contributing to both outgoing short- and longwave radiation reduction. Overall, both effects tend to compensate for each other and only a few differences in Q* have been observed between urban and rural environments on a yearly average (Oke and Fuggle, 1972; Christen and Vogt, 2004). Nevertheless, depending on the seasons and time of the day, larger net radiation has been observed in urban areas during daytime and in winter, when snow covers surrounding rural areas (Christen and Vogt, 2004).

Secondly, compared to the surrounding areas, an urban environment sharply modifies the way surface energy is dissipated (i.e. the energy partitioning between sensible and latent heat fluxes). In rural environments vegetation and pervious surfaces provide larger evapotranspiration rates (i.e. latent sensible heat flux), and therefore lower sensible heat flux, whereas in urban areas energy is mainly dissipated through sensible heat flux. A non-natural term, sensible heat flux, due to heat release by human activities (e.g. building heating or cooling), adds to a natural sensible heat flux, further increasing sensible heat flux in urban areas (Arnfield, 2003). As a result, the Bowen ratio is amplified in urban areas (Table 1). Such a large dissipation of energy through sensible heat flux, which transfers heat from the surface to the air, leads to the so-called urban heat island (UHI) effect, the most well-known alteration of (local) climate due to urbanization that corresponds to a warmer climate in urban environments compared to surrounding rural environments (around 2–3 ∘C). UHI is defined as a temperature difference between the city and its surroundings, depending on the local land use. Nevertheless, UHI intensity is sharply variable according to the time of day (e.g. Pearlmutter et al., 1999), the season (e.g. Eliasson, 1996; Zhou et al., 2014), the geographical location, spatial organization of the urban fabric (e.g. building size and density, human use, fraction of vegetation) (e.g. Emmanuel and Fernando, 2007; Hart and Sailor, 2009), and rural land use (e.g. forests, crops, bare soil) (Chen et al., 2006). Recently, Yao et al. (2019) combined satellite-based observations of land surface temperature (LST) and enhanced vegetation index and showed that rural greening has contributed by +0.09 ∘C per decade (23 %) over the period 2000–2017 to the increase in daytime surface UHI intensity (i.e. urban LST minus rural LST). By modifying the local energy budget, urbanization modifies the boundary layer structure and lastly influences the water budget. Urban signatures (e.g. change in magnitude, intensity, and spatial patterns) have been observed in precipitation (see Shepherd, 2005, and Pielke et al., 2007, for a review on urban precipitation). Moreover, complex urban terrain amplifies regional gradients in temperature, pressure, moisture, and wind that act as a source of vorticity for storm ingestion and development into tornadoes (Kellner and Niyogi, 2014). Moreover, urban areas can attenuate, split, or deflect extreme storm events (e.g. Lorenz et al., 2019) and modify their intensity and occurrence. Over the Beijing metropolitan area, 60 %–95 % of the selected weather stations show that the intensity and occurrence of extreme rainfalls have slightly decreased throughout 1975–2015, periods with consecutive rainy days (CRDs) have lengthened, and the Julian dates of daily maximum precipitation have been delayed (Zhang et al., 2019). Furthermore, cities are important source of aerosols that help initiate thunderstorms (Haberlie et al., 2015). However, the joint study of UHI and urban pollution island is still in its infancy and the indirect radiative effect of aerosols (i.e. impact on cloud properties and formation) on UHI needs further investigations (Li et al., 2018).

To mitigate UHI-induced warming, vegetated or highly reflective roofs are being integrated in the built environment and have received a growing interest in climate modelling studies. Cool roofs absorb less incoming shortwave radiation than dark roofs. They decrease the local and regional summer surface temperature by 0.1–0.9 ∘C (Millstein and Menon, 2011; Georgescu et al., 2012; Salamanca et al., 2016; Vahmani et al., 2016). Their impact on climate is not just limited to surface energy budget as, for example, precipitation decrease was put forward in a modelling framework (Georgescu et al., 2012). Benefits from green roofs are analogous to cool roofs, as vegetation contributes to cooling via increased albedo and water evapotranspiration. In situ experiments with different species have surface temperature differences up to 3 ∘C (MacIvor and LundHolm, 2011). However at the regional scale and over urban areas, simulated cooling is greater for the cool roofs relative to the green roofs because of the vegetation seasonality and sensitivity to dryness (Georgescu et al., 2014).

From a biological perspective, at a local scale, the development of urban areas and the related activities directly affect air quality and local temperatures, which leads to modifications in the biology of organisms. Studies based on the analysis of tree traits along an urban–rural gradient showed that tree growth and phenology are affected by the vicinity of an urban area mainly due to increase in temperature (Gillner et al., 2014; Mimet et al., 2009; Dale and Frank, 2014), CO2 concentrations (Calfapietra et al., 2010; Ziska and George, 2004), ozone deposition (Gregg et al., 2003; MacKenzie et al., 1995), and through the enhanced effect on air quality via the increased emissions of BVOCs (Calfapietra et al., 2013; Lathière et al., 2006). Recent studies have also focused on the effects of soil waterproofing in urban areas that reduce water availability and exacerbate water stress in urban forests, significantly affecting growth (Vico et al., 2014; Volo et al., 2014; Scalenghe and Marsan, 2009).

From a chemical perspective, at local scales, urbanization directly affects both O3 and aerosol levels by increasing the number of emission sources on a limited area (e.g. traffic, domestic heating). In the literature, there is an increasing interest in the direct impacts of urbanization on air quality (special issues in the Atmospheric Chemistry and Physics journal related to the Megapoli-Paris 2009/2010 campaign and the MILAGRO and the City-zen projects, Baklanov et al., 2011, 2018; Zhu et al., 2019; Ooi et al., 2019), with a special focus on O3 levels, summer pollution (Nowak et al., 2000; Civerolo et al., 2007; Jiang et al., 2008), and the role of urban trees in O3 pollution via BVOC emission changes (Chameides et al., 1988; Cardelino and Chameides, 1990; Corchnoy et al., 1992; Benjamin et al., 1996; Taha, 1996; Benjamin and Winer, 1998; Yang et al., 2005; Taha et al., 2016; Livesley et al., 2016; Churkina et al., 2017; Bonn et al., 2018).

Increase in urban LU following population growth exacerbates O3 pollution during summer, mainly due to changes in NOx emissions (Zhu et al., 2019). In the greater Houston area (Texas), under a projected increase in urban LU by 62 %, together with changes in anthropogenic and biogenic emissions, the number of extreme O3 days in August rose by up to 4–5 days, with LUCs contributing to a 2–3 d increase (Jiang et al., 2008). In the greater New York City region, future urban LU changes may enhance episode-average O3 levels by about 1–5 ppb and episode-maximum 8 h ozone levels by more than 6 ppb (Civerolo et al., 2007). In metropolitan regions, changes in O3 levels show a heterogeneous spatial pattern: they decrease in the urban core, likely due to high NOx levels (O3 titration), while they generally increase downwind of precursor sources (Civerolo et al., 2007; Jiang et al., 2008). In an urban environment, BVOC emissions from urban trees seem to have a negligible effect on summer O3 levels (<1 ppb compared to increases of 1–7 ppb due to urban LUCs; Nowak et al., 2000, vs. Jang et al., 2008). However, the effect of urban green areas on BVOC emissions and O3 pollution depends on tree species (Taha, 1996; Taha et al., 2016); for this reason, the choice of urban trees based on their BVOC potential may be addressed as a critical urban land management practice (Benjamin et al., 1996; Benjamin and Winer, 1998; Churkina et al., 2015; Calfapietra et al., 2015; Grote et al., 2016). For example, in Beijing, deciduous trees dominate (76 %) and some of the main species are high BVOC emitters (e.g. Sophora Japonica L., Populus tomentosa L., and Robinia pseudoacacia L.), which may favour a worsening in O3 pollution due to the rapid increase in NOx emissions (Yang et al., 2005). In the Los Angeles metropolitan area, Corchnoy et al. (1992) measured BVOC emission rates of 11 tree species to underpin the selection of potential shade trees, whose planting should reduce the urban heat island effect. Accounting for California climate, the authors suggested the best (e.g. crape myrtle and camphor tree) and poor (e.g. liquidambar and carrotwood tree) choices for urban trees, and underlined that a large difference in BVOC emissions should be factored into decision-making about shade trees to plant. In California's South Coast Air Basin, medium- and high-emitting trees may lead to hazardous O3 levels (>50 ppbv) (Taha, 1996). In the same geographical area, the most effective scenario to reduce the peak ozone involves replacing 4.5 Mha of high BVOC emitters with low BVOC emitters, while to target all-hour ozone the best choice consists in planting 2.5 Mha of low BVOC emitters in urbanizing areas and switching 4.5 Mha from high- to low-emitting species (Taha et al., 2016). It is important to emphasize that, although BVOC concentrations are usually lower than anthropogenic volatile organic compound (AVOC) concentrations in urban areas, BVOCs react faster than AVOCs and can thus have significant effects in urban areas, as shown by Chameides et al. (1988) in the Atlanta metropolitan region.

At the regional scale, Chen et al. (2009) demonstrated that LULCCs can offset the impact of temperature on biogenic emissions and concluded that LULC evolution should be factored in the study of future regional air quality. Other than land use, land cover, and land management changes (LULCCs and LMCs) discussed here, changes in climate conditions and anthropogenic pollutant emissions (e.g. due to “clean air” policies) directly and indirectly influence air quality and interact in a non-linear fashion with LULC and LMCs; for this reason the climate–emission–land system should be considered as a whole when studying changes in surface O3 and aerosols.

The main aim of agricultural management is to increase productivity and therefore has an immediate effect on the agricultural ecosystem functioning (Tillman et al., 2002). Most of these agricultural practices will also have direct or indirect impacts on the environment other than the biosphere (e.g. atmosphere, water, soils) (Sutton, 2011). Agricultural intensification also enhances the export of organic matter from the affected ecosystems with consequences such as the reduction of carbon and nitrogen cycling and soil degradation and erosion (Mattson et al., 1997; Ruysschaert et al., 2004). Examples of agricultural intensification are the conversion of pasture or grasslands into agricultural land or including rotations of agricultural and grasslands.

From a physical perspective, among land management practices, irrigation is one of the most common all over the world, and it significantly modifies the surface water and energy budget. The amount of additional water put into the soils tends to increase the latent heat flux at the expense of sensible heat flux, leading to an irrigation cooling effect (ICE) of the ambient air. In California, for example, this effect was observed during daytime over a long-term dataset and estimated to several degrees (-1.8 to -3.2 ∘C since the beginning of irrigation – Lobell and Bonfils, 2008; Bonfils and Lobell, 2007). However, there are two opposite indirect heating effects. First, the high-albedo desert is converted into a low-albedo vegetated plain (Christy et al., 2006), which results from a combination of crop planting and irrigation and can therefore be classified as a land cover change rather than an agricultural intensification. Second, the greenhouse warming is enhanced due to the increase in water vapour. The greenhouse effect is less important than the transpiration effect on temperature and dominates during the night-time. Several modelling studies assess both greenhouse and transpiration effects (Boucher et al., 2004; Sacks et al., 2009; Puma and Cook, 2010; Cook et al., 2011, 2015; Kueppers and Snyder, 2012) and highlight that locally the ICE may have partly masked the 20th century climate warming due to increased greenhouse gases (Kueppers et al., 2007). Meteorological studies suggest that irrigation can also lead to an increase in summer cloud cover and precipitation, as observed over the Great Plains region in the United States, downwind of the major irrigation areas (Segal et al., 1998; Adegoke et al., 2003; DeAngelis et al., 2010). In China, paddy cultivation requires water to stay on the ground during the rice-growing season, leading to a moistening of the land surface, an increase in the latent heat flux, and a decrease in the near-surface temperature from May to July in the Sichuan Basin (Sugimoto et al., 2019). Thiery et al. (2017) demonstrated that irrigation influences temperature extremes and leads to a pronounced cooling during the hottest day of the year (-0.78 K averaged over irrigated land). In addition, this impact of irrigation on temperature is not limited to an agricultural environment as the same cooling effect has also been reproduced for urban irrigation in a water-scarce region (Los Angeles area), with the largest influence in low-intensity residential areas (average cooling of 1.64 ∘C) (Vahmani and Hogue, 2015). Affecting soil moisture and surface temperature, changes in irrigation could also affect soil processes and exchanges of greenhouse gases and chemically reactive compounds between the surface and the atmosphere (Liu et al., 2008). Performing irrigation experiments on the Inner Mongolian steppe, Liu et al. (2008) observed a significant sensitivity of the ecosystem CO2 respiration to increased water input during the vegetation period, whereas the effects on CH4 and N2O fluxes were much more moderate. In order to study the impact of irrigation on ozone and pollutants in the Central Valley of California, Li et al. (2016) implemented an irrigation method in the model WRF-Chem and showed an increase in surface primary pollutant concentrations within the irrigation zone. They also calculated an enhancement in the horizontal transport of ozone and other pollutants from irrigated to unirrigated areas near the ground surface. However, few studies have been published so far on this topic from a biological or chemical perspective, and the effect of irrigation on biological processes or on the atmospheric chemical composition therefore remains poorly quantified.

Since the Second World War, the use of synthetic N fertilizers largely increased, with half of the quantity ever used being applied in the last 20 years (Erisman et al., 2007). The growth of nitrogen fertilization threatens water sources (e.g. eutrophication of surface waters, pollution of groundwater, acid rains), soils (e.g. soil acidification), climate via GHG emissions, and air quality.

Few studies have investigated the impact of fertilizer use from a physical perspective, and yet physical interactions between the surface and the atmosphere could be affected. Based on a long-term experiment of fertilizer and amendment application running for 70 years, Pernes-Debuyser and Tessier (2004) observed that physical properties of plots were significantly affected, especially those related to soil–water relations. In spite of the preservation of their porosity, plots became more sensitive to the degradation of their hydraulic properties. Similarly, Hati et al. (2008) showed, in the case of an intensive conventional cultivation in subhumid tropics in India (acidic Alfisols), the importance of soil management practices in maintaining the soil physical environment, with a potential impact on soil aggregation, soil water retention, microporosity, available water capacity, or bulk density.

From a biological perspective, the additional source of nitrogen has different impacts on the atmosphere, mainly linked to an increase in reactive nitrogenous emissions (NH3, NOx) (Fowler et al., 2009, 2013; Galloway et al., 2003) but also in emissions of a GHG such as N2O. Increase in production also affects leaf area index and plant height and therefore surface properties and physical exchanges with the atmosphere. Finally, fertilization also influences soil microbial characteristics and, consequently, exchanges of several gaseous compounds (Marschner et al., 2003; Cinnadurai et al., 2013; Joergensen et al., 2010; Murugan and Kumar, 2013). Grassland usually stores considerable amounts of carbon in the soils, mainly due to a permanent plant cover and to a relatively large below-ground biomass (Bouwman, 1990; Casella, 1997). However, the amount of stored carbon and the emission of greenhouse gases depend on the management of this grassland (ploughing, fertilization, pasture, etc.) (Soussana et al., 2004; Lal, 2004) and on climatic conditions (Hu et al., 2001). Some studies suggest that increased nitrogen fertilization can enhance C storage in grassland. Conversely, nitrogen fertilization increases leaching and emissions of N2O and other nitrogen species (e.g. NH3, NO) to the atmosphere, with negative consequences on air quality (Flechard et al., 2005; Senapati et al., 2014; Chabbi et al., 2015).

From a chemical perspective, the increase in NH3 emissions to the atmosphere can have a serious impact on air quality through the formation of secondary organic aerosols. Agricultural practices and techniques that reduce the evaporation of manure and urea and the use of N fertilizers help in lowering ammonia emissions from agriculture as documented in Europe, where 90 % of the total ammonia emissions come from agriculture (-9 % over 1990–2002; Erisman et al., 2008). In China, where N fertilizer application rose by 271 % over the 1977–2002 period, with an increase of 71 % only in grain production (Ju et al., 2009), Ju et al. (2009) suggested reducing N application rates by 30 %–60 %. This agricultural management practice would still ensure crop yields and N balance in between rotations and would reduce economical costs for farmers, while substantially reducing N losses to the environment.

From a physical perspective, several crop management techniques (e.g. cover crops, double-cropping, no tillage) have a direct effect on regional climate through changes in surface–atmosphere fluxes and surface climate conditions, and are considered among geoengineering options. When tillage is suppressed, crop residues are left on the field, resulting in two counteracting mechanisms: albedo increases while evaporation reduces (Lobell et al., 2006; Davin et al., 2014; Wilhelm et al., 2015). Surface albedo increases by 10 % and lowers hot temperature values by about 2 ∘C; however the effect on the mean climate is negligible. Climate effect of two growing seasons per year has been largely untested. Only Lobell et al. (2006) have shown via modelling that this experiment has a small impact on temperature on multi-decadal timescales when compared to practices as irrigation. However, more recently Houspanossian et al. (2017) have observed through satellite imagery a difference in reflected radiation between single and double-cropping of up to 5 W m-2. Similar to a tillage/no-tillage mechanism, differences over South America were induced by a longer fallow period in the simple cropping case. Seed-sowing dates also likely plays a role in surface energy balance, due to the modification of the growing season length (Sacks and Kucharik, 2011).

Among agriculture practices, as an alternative to biomass burning and natural decomposition, the use of charcoal from biomass pyrolysis to enrich soils may reduce CO2 emissions. However, as a side effect, the resulting darker soil increases the local radiative forcing through albedo change and offsets the sequestration effect up to 30 % according to Bozzi et al. (2015), who carried out the analysis based on observations of agricultural field albedo. Biochar has similar effects (Usowicz et al., 2016; Meyer et al., 2012).

From a chemical perspective, fallow lands are potential sources of dust and coarse aerosols (PM10), especially in regions where gusty winds dominate. Insufficient crop residues on the surface and finely divided soils by multiple tillage operations expose fallow land to wind erosion, thus contributing to poor air quality (López et al., 2000; Sharrat et al., 2007). In addition, wind erosion is likely to reduce crop yields by removing the richest fraction of soils, reducing the water-holding capacity of soils and enhancing soil degradation. Compared to conventional tillage (i.e. mouldboard ploughing followed by a compacting roller), alternative or reduced tillage practices (e.g. chisel ploughing) prevent wind erosion during fallow periods in semiarid Aragon (López et al., 2007). In addition, reduced tillage improves soil protection by lowering the wind-erodible fraction of soil surface (-10 %), increasing fraction of soil covered with crop residues and clods (+30 %), and enhancing soil roughness (15 % compared to 4 % under conventional tillage). These agricultural practices therefore have the potential to modify aerosol sources by modifying the state of surfaces.

From a biological point of view, the conditions of the soil surface and the management of crop residues highly affect soil quality as well as the functioning and the abundance of soil microorganisms (Smith et al., 2015, 2016). In terms of exchange with the atmosphere, this results in soil structural changes affecting soil porosity and directly influencing the emissions of NOx and BVOCs (Gray et al., 2010; Bertram et al., 2005). Effects can also be seen on soil organic matter content and degree and rate of decomposition therefore affecting emissions of several nitrogen compounds therefore affecting GHG balance (emissions of N2O vs. storage of carbon) (Xia et al., 2018) and air quality (NH3, NOx emissions) (de Ruijter et al., 2010). Conversely, soil surface conditions also influence the deposition of O3 (Stella et al., 2019) and potentially other highly reactive atmospheric compounds such as pesticides (Alletto et al., 2010).

Fire is still largely used as a traditional agricultural practice (e.g. slash-and-burn agriculture, pest control, promotion of the growth of fresh grass for grazing) and to convert forests to pasture/croplands, especially in tropical regions (Yevich and Logan, 2003). On a local scale, intensive mechanized grain agriculture reduces the use of fire. However, the wealth generated from intensive agriculture may be reinvested in traditional extensive land uses that promote fire (Wright et al., 2004).

Generally, fires can impact soil colour, pH, bulk density, and soil texture, and are therefore critical for physical surface–atmosphere exchanges, together with biological properties of soil such as species richness and microorganism content (Thomaz et al., 2014; Verma and Jayakumar, 2012; Savadogo et al., 2007). However the impact of fires from a physical or a biological perspective has been poorly investigated, especially regarding the long-term effect (Dooley and Treseder, 2012; Pressler et al., 2019).

From a chemical perspective, fire has impacts on both photochemical pollution (O3 production) and aerosol loading. During fire episodes, O3 production switches from a VOC-sensitive regime in nascent smoke plumes (i.e. first hours of burning and close to the ignition point) to a NOx-sensitive regime as the plume ages. In nascent smoke plumes NOx levels are high and photochemical activity is low. Smoke plume ageing decreases NOx levels via atmospheric dilution and chemical reactions, resulting in increased O3 production (e.g. Jost et al., 2003; Trentmann et al., 2003; Yokelson et al., 2003; Mason et al., 2006; Singh et al., 2012). During fire episodes, O3 levels may reach hazardous values, with the 8 h average O3 concentration often exceeding air quality standards (around 50–75 ppbv; Bytnerowicz et al., 2010). Fires also release huge amounts of both coarse- and fine-mode aerosols, leading to concentrations that largely exceed background levels (Phuleria et al., 2005; Hu et al., 2008) and that substantially affect visibility (Val Martin et al., 2015). Over Singapore, Indonesian fires caused the average daily minimum horizontal visibility to decrease, firstly, to less than 2 km, and later to 500 m (Goldammer et al., 2009). Fire emissions encompass aerosol precursors such as NH3 and BVOCs as well.

Forest management mainly relies on tree species selection, fertilization, litter raking, thinning, and clear-cutting (Eriksson et al., 2007), together with planting and harvest types, burning, and understory treatment.

From a physical perspective, along with crop management, forest management could have a similar impact on local climate but is still poorly investigated (Bellassen and Luyssaert, 2014; Luyssaert et al., 2014), although forested areas cover one-third of the global land surface (Klein Goldewijk, 2001). The large conversion of broadleaved to managed conifer forest resulted in biogeophysical changes, which contributed to higher temperatures instead of attenuating them.

From a biological perspective, through modelling, Naudts et al. (2016) showed that 2.5 centuries of forest management in Europe may not have mitigated climate warming, contrary to what was sometimes assumed until now. With regard to atmospheric carbon budget, forests were altered from acting as a carbon sink to a carbon source because of the removal of litter, dead wood, and soil carbon pools.

From a chemical perspective, by modifying the surface characteristics, forest management can change sources and sinks of reactive compounds, and therefore affect air quality. Conversely, forest management can also be a tool when targeting air pollution reduction. Using a coupled-model approach, Baumgardner et al. (2012) analysed the improvement of air quality by a forested peri-urban national park in the Mexico City megalopolis and underlined that their results can be used to understand the air quality regulation potentially provided by peri-urban forests as an ecosystem service, together with the regional dynamics of air pollution emissions from major urban areas.

While agriculture has been criticized for several decades for its impacts on water quality (nitrate and pesticides) and for its contribution to climate change (emissions of nitrous oxide and methane), the question of its contribution to air pollution in urban and peri-urban areas has emerged only recently in the public debate, with a particular resurgence in recent spring episodes of aerosol pollution. Ammonia, which is largely emitted by animal excreta and by the application of mineral and organic fertilizers, contributes to the formation of secondary aerosols. Hence, the reduction of its emissions is an important stake for the improvement of air quality. In recent years, control of ammonia emissions has become a major concern at regional, national, and international levels and, since the end of the 1990s, a set of regulations has been put in place. To further reduce ammonia emissions, improve air quality, and optimize costs and benefits requires a better knowledge and quantification of ammonia sources and also an analysis of long-term strategies. France regularly undergoes peaks of aerosol pollution (PM10–PM2.5), especially at the end of winter–early spring, when favourable weather conditions coincide with the beginning of fertilizer spreading. In March 2014, high PM2.5 concentrations were observed in the Paris region, leading to the introduction of alternating traffic, and therefore made citizens particularly aware of the issues of air quality. Predicting air quality at the regional level is crucial to understand these episodes and to recommend appropriate levers of action in the short term to limit the magnitude of these episodes. Air pollution affects not only human health, but also the overall productivity of ecosystems and crop yields, through increased dry deposition of N compounds and O3, which in turn could affect BVOC emissions. In addition, by modifying plant functioning in terms of evapotranspiration and soil moisture status, ozone deposition may affect the hydrological cycle, which in turn will affect surface but also wet deposition of pollutants and nutrients.

We have here a typical example where scientists involved in agronomy, physics, biology, and chemistry should interact to improve predictions of ammonia emissions, transport and reactions related to weather conditions, soil biological processes, and plant phenology, to estimate feedbacks of air pollution on the functioning of involved ecosystems. However, to solve the problem, cooperation between farmers, urban planners, and decision makers is required to define optimal fertilization dates and a territorial planning of urban and peri-urban areas that accounts for the distribution of agricultural activities around the city.

Many studies have explored techniques to counterbalance the deleterious effects of urbanization on the local environment. Among the numerous solutions already proposed, urban greening is one of the most interesting since it could allow (i) an attenuation of the UHI (e.g. Shashua-Bar and Hoffman, 2000; Alexandri and Jones, 2008; Feyisa et al., 2014), (ii) a direct mitigation of air pollution via the absorption of pollutants by plants (Hill, 1971), and (iii) an indirect improvement of air quality through UHI mitigation since temperature partly drives and controls pollutant emission, dispersion, and formation (Sini et al., 1996; Kim and Baik, 1999; Stathopoulou et al., 2008).

On the one hand, green surfaces such as parks, gardens, or green roofs and walls contribute to mitigating the UHI and currently receive strong attention from both scientists and urban planners (e.g. Shashua-Bar and Hoffman, 2000; Akbari et al., 2001; Kumar and Kaushik, 2005; Alexandri and Jones, 2008; Feyisa et al., 2014) with some interdisciplinary and inter-community experiences already established (e.g. the Urban Climate Change Research Network, http://uccrn.org/, last access date: 22 May 2019; the MAPUCE project in Toulouse, https://anr.fr/Project-ANR-13-VBDU-0004, last access date: 7 June 2019, local projects in Stuttgart, New York). On the other hand, a growing number of studies focus on urban air quality assessment to quantify impacts of urban vegetation (e.g. Yang et al., 2005; Novak et al., 2006; Escobedo et al., 2011; Selmi et al., 2016). Changes in planted species and their surfaces can indeed significantly impact the amount and fate of reactive compounds emitted, such as biogenic VOCs or nitrogen compounds, and therefore affect the air chemical composition in terms of gases and aerosols (Ghirardo et al., 2016; Janhäll, 2015; Taha et al., 2016). Nevertheless, feedbacks on air quality by UHI mitigation are not accounted for but could lead to air quality degradation, by affecting pollutant and especially ozone precursor dispersion (Lai and Cheng, 2009). To quantify to which extent urban greening can help to mitigate urban local climate and atmospheric pollution, and its subsequent effects at the regional scale, it is therefore necessary to adopt interdisciplinary approaches (Baró et al., 2014), involving atmospheric physics and chemistry, but also urban planners. Indeed, although the role of urban form, urban fabric, and building arrangement and orientation in UHI mitigation was explored in previous studies (Stone and Norman, 2006; Emmanuel and Fernando, 2007; Shahmohamadi et al., 2010; Middel et al., 2014), this was not the case for atmospheric composition.