Advancing understanding of land–atmosphere interactions by breaking discipline and scale barriers

Vegetation and atmosphere processes are coupled through a myriad of interactions linking plant transpiration, carbon dioxide assimilation, turbulent transport of moisture, heat and atmospheric constituents, aerosol formation, moist convection, and precipitation. Advances in our understanding are hampered by discipline barriers and challenges in understanding the role of small spatiotemporal scales. In this perspective, we propose to study the atmosphere–ecosystem interaction as a continuum by integrating leaf to regional scales (multiscale) and integrating biochemical and physical processes (multiprocesses). The challenges ahead are (1) How do clouds and canopies affect the transferring and in‐canopy penetration of radiation, thereby impacting photosynthesis and biogenic chemical transformations? (2) How is the radiative energy spatially distributed and converted into turbulent fluxes of heat, moisture, carbon, and reactive compounds? (3) How do local (leaf‐canopy‐clouds, 1 m to kilometers) biochemical and physical processes interact with regional meteorology and atmospheric composition (kilometers to 100 km)? (4) How can we integrate the feedbacks between cloud radiative effects and plant physiology to reduce uncertainties in our climate projections driven by regional warming and enhanced carbon dioxide levels? Our methodology integrates fine‐scale explicit simulations with new observational techniques to determine the role of unresolved small‐scale spatiotemporal processes in weather and climate models.


INTRODUCTION
Ecosystems are undergoing large transformations due to anthropogenic activities that have led to drastic changes in the land, 1 weather, 2 climate, 3 and atmospheric composition. 4 The changes are having dramatic effects on sensitive regions, such as the Amazonia basin. 5 Climate modifications are specifically inducing changes in the hydrological cycle, 6,7 in which the coupling between vegetation, evapotranspiration, and clouds plays a prominent role. 8 Owing to the many stabilizing feedbacks between vegetation, evapotranspiration, and clouds, ongoing changes in the climate and land surface may lead to unexpected and transformative changes. These changes could have a profound impact on clouds and atmospheric circulation 9 and the role of ecosystems as sources and sinks of greenhouse gases. 1 Currently, these effects are difficult to quantify specifically because observations and models do not reflect the relevant processes in enough details to capture all potential feedbacks. In this article, we will focus on two interrelated phenomena that are poorly understood and hard to quantify: the coupling across scales (leaf-canopy) between transpiration and photosynthesis occurring under various ecosystem conditions 10 and the relationships to cloud variations on space and time 11 . For both phenomena, the atmosphere integrates biochemical and physical processes occurring at a wide range of spatiotemporal scales and we must interrelate knowledge from the disciplines of atmospheric and plant sciences to fully unravel all potential feedbacks in the coupled vegetation-atmosphere system. Until now, the separation between disciplines has hampered the interconnection between the vegetation dynamics regulating CO 2 and transpiration on the one hand and the influence of atmospheric dynamics (including cloud feedback) on these vegetation dynamics on the other hand.
The current generation of weather and climate models does include many process representations and feedbacks between the vegetation, hydrological cycle, and atmosphere. 12 However, their spatiotemporal resolution as well as process description, for instance, the diurnal cycle of state meteorological and atmospheric composition variables, are insufficient to fully capture the detailed feedbacks between vegetation and clouds that drive meteorological processes. 2,8,12 Therefore, we believe that special attention needs to be placed on how the local dynamic interactions between leaf, canopy, and clouds lead to a large temporal and spatial variability in the photosynthesis rate. 13 As a result, the regional partitioning and properties of the surface turbulent fluxes will result in changes that yield modifications in the cycling and organization of clouds. Advancing understanding and improving on the parameterizations could lead to a reduction in the large uncertainties in the projections of cloud feedback and terrestrial CO 2 uptake under climate modification conditions. 8 Therefore, we argue that by integrating all these small spatiotemporal processes, including the combined effects of the impact of clouds and CO 2 assimilation at the regional level, we could assess how the atmosphere-ecosystem is modified by current and future climates.
To illustrate how processes interact across scales, Figure 1 shows a common transition from ocean to land whereby cloud formation and intensification predominantly occurs over the vegetated land surface. This clear-to-cloud transition is shown for climate-sensitive forest ecosystems in (sub)tropical, temperate, and boreal climates. 14 Although the transitions are visually similar among these forest ecosystems, there are significant differences in terms of species and plant type as well as the magnitude and seasonality of the energy input and available soil water. Also, the synoptic and mesoscale atmospheric patterns differ largely as a function of latitude and seasonality. As illustrated in Figure 1, cloud properties are continuously modified and shaped due to ongoing interactions with vegetation when developing over land. These interactions depend on the surface exchange fluxes and partitioning of water and energy as well as on boundary layer dynamics connected to radiation perturbation by clouds and on canopy roughness via its effects on wind shear and turbulence. 15 In all these ecosystems, horizontal gradients and vertical contrasts between the land, canopy, atmosphere, and clouds play a key role in controlling precipitation 16 and carbon dioxide uptake by vegetation at regional scales. 4 With this in mind, we offer a way to integrate these processes at the meteorologically relevant spatiotemporal scales using current 2. How is atmospheric composition (greenhouse gases and reactive species) impacted by turbulent transport, mixing, convection, weather variables (temperature, moisture, and wind), and reactivity?
3. How do physical, chemical, and biological processes interact between local and regional scales, and at different time scales? 4. How does climate change, in terms of enhancement of carbon dioxide and warming, lead to changes in cloud and plant physiology feedback?
The first challenge focuses on radiation as a primary source of energy, where clouds [17][18][19] and canopies 20,21 modify the transfer of incoming shortwave radiation, and therefore, the distribution between the direct and diffuse radiation. Small-scale and short-term variability in the partitioning between direct and diffuse radiation leads to large fluctuations in the photosynthesis rate and stomatal aperture that cannot be captured without considering these processes at high spatiotemporal resolution. Clouds also modulate the temperature and the water vapor pressure deficit (VPD) in and above the canopy. 22 This results in relationships between radiation, temperature, and moisture 23 that regulate photosynthesis and transpiration of plants, which, in turn, governs the surface energy distribution over land. 24 These relationships continuously change over time and space, and are regionally dependent. Considering atmospheric factors driven by weather and a changing climate, current efforts focus on developing consistent and reliable theories to calculate photosynthesis and plant hydraulics 25 that will need to be integrated and evaluated in weather and climate models.
The second challenge focuses on connecting atmospheric turbulence to biochemistry processes at the surface. The former is governed by wind and thermodynamic instabilities driving transport and mixing. The latter, biochemistry, governs the exchange and transformation in atmospheric composition. Vegetation properties, such as canopy structure, leaf area index (LAI), and leaf color, determine how much energy is absorbed and regulates the skin temperature and the near leaf VPD. 15 Within the entire canopy, this absorption and partly reflected radiation strongly influences vertical profiles in skin temperature and VPD. Therefore, this radiative energy drives the turbulent-canopy fluxes of energy and moisture, 26 which return to the atmosphere in the form of momentum, heat, and moisture. This resulting energy-water-carbon partitioning and redistribution over land governs atmospheric turbulence and cloud formation. 27 As a result, there are two direct effects on cloud dynamics and microphysicalchemical processes. The first effect involves the relationship between the energy and moisture fluxes at the vegetated land surface and how this drives the vertical transport of heat and moisture to higher atmospheric levels where condensation occurs. 28,29 The second effect concerns the key role played by plants in regulating the exchange of biogenic volatile organic compounds. 30,31 These highly reactive compounds act as precursors of aerosols and new particle formation as well as bio-aerosols, prior to the formation of cloud condensation nuclei. 32 A key aspect of these processes is how the aerosol composition becomes a mixture originating from natural and anthropogenic sources. Aerosols characterized by different compositions (sea, rural, and urban) could lead to changes in the cloud microphysics with subsequent influence on the formation, maturity, precipitation, and dissipation of clouds. 33 The third challenge focuses on the interactions and feedbacks between regional-scale weather and climate phenomena with the local processes described above. These cross-scale interactions can lead to large regional variations in the thermodynamic state variables and the composition of greenhouse gases and chemically active species. In particular, the variations in land properties exert a large influence on the organization of the atmospheric flow. This can be either dynamically driven by, for instance, the shading of clouds, 34,35 or by more static or gradual variations of land properties, such as albedo, roughness, and soil moisture, 36 for example. Figure 2 shows the relationships between surface heterogeneities and flow organization. Three length scales are relevant: the patch (X P ; defined as the length scale in which surface properties are disturbed), the length scale of the surface heterogeneity (X H ), and the most representative length scale of the land-atmosphere interaction, represented by the turbulence length scale (L). 37 In brief, F I G U R E 2 Nondimensional representation of the impact of surface heterogeneity, represented by the heterogeneity size (X H ) and the patch size (X P ) and the atmospheric flow organization represented by the length scale of turbulence (L). X P is the size of the disturbed surface. The figure describes three regimes that are common in land-atmosphere interactions. From top to bottom: (1) the macro-scale heterogeneity in which the patch and surface heterogeneity are larger than the turbulent representative scale, that is, land or sea breeze. (2) The meso-scale heterogeneity in which patch and surface heterogeneity have a similar length scale as the turbulence and (3) micro-scale heterogeneity in which the length scale heterogeneity is smaller than the turbulence length scale. More information in how the length of the patch and the heterogeneity influence the atmospheric flow can be found at van Heerwaarden et al. 37 in the case of macro-scale heterogeneity, the surface heterogeneity dominates turbulence which leads to the organization of flows characterized by larger length scales than turbulence, such as sea-breeze.
Under the micro-scale heterogeneity regimen, the individual surface properties exert a relatively small influence on the flow, and turbulence is able to blend the small-scale heterogeneities, above a certain height. The most challenging case is represented by the meso-scale heterogeneity. Here, the length scale of turbulence is similar to the surface heterogeneity. Note that the relationship between these length scales varies during the day due to variation of radiation and turbulence that yields different conditions in the interaction between land and the atmosphere, that is, cloudy versus clear, and thermal stratification, that is, stable, unstable, and neutral. 38,39 The challenge relies on first understanding how these differences of surface heterogeneities influence the atmospheric flow and second how to represent this coupling between surface heterogeneity and flow organization in weather 40 and climate 12 models. In these models, processes related to land-atmosphere exchange and turbulence are represented in the form of parameterized expressions. As a result, surface heterogeneities are normally calculated as an aggregate of surface properties, which could lead to misrepresentations of turbulence. 40 These inaccurate calculations of the turbulent transport of momentum, heat, moisture, and aerosols could yield erroneous calculations of cloud dynamics and physics. Turbulent-and cloud-resolving models as presented and discussed in this article are able to simulate the most energetic parts of the coupling between land-atmosphere, 35,41 and therefore, simulate the three regimens presented in Figure 2 as a continuum of linked scales. 42 With respect to the interactions between the local processes and regionally driven air masses, the recycling of moisture controlled by vegetation plays a key role. 43,44 Incoming air masses (for instance, the ones with oceanic origin at Figure 1), driven in synoptic and mesoscale patterns, are influenced by the local thermodynamic variables and atmospheric composition which leads to changes in the water budget (evaporation and precipitation). The coupling between the ecosystem and clouds, the identification of the sources and sinks of moisture, the capacity of locally recycling moisture, 43,44 and the vegetation/soil exchanges of CO 2 45 need to be properly quantified. By identifying the location and the monthly to yearly source regions of moisture or CO 2 , one can understand if those air masses have a nonlocal origin over the ocean or if their origin is locally driven by evaporation recycled by ecophysiological and hydrological processes. 46 The fourth challenge is related to how large-scale climate changes 47 influence key meteorological and ecophysiological processes at regional and global scales. 3,8 As shown in Figure 3, weather, atmospheric composition, and climate modifications drive the dynamics of ecosystem and clouds. These dynamics are strongly influenced by modifications of the atmospheric thermodynamic structure (stabilization) disturbed by more frequent weather extremes. The figure is far from complete, but it attempts to show that as part of the research challenge, some processes offset each other. In short, we have divided the interaction of these processes in two groups. The first group includes ecosystem processes affected by climate change that leads to modifications in plant assimilation and transpiration, and F I G U R E 3 Weather, climate, and atmospheric composition are changing due to the transitions between present and future climate conditions. Ecosystem (green) and cloud processes (blue) are influenced by opposite effects which, in turn, influence the dynamics of ecosystems and clouds. These opposite effects depend on thermal stabilization of the atmosphere under current climate predictions, but also on higher frequency of extreme weather events, such as drought, heatwaves, and flooding that are expected to occur. Both thermal stabilization and extreme weather effects can become dominant for longer periods and are region-specific.
consequently, the distribution of the land exchange fluxes of energy, water, and carbon. For example, the fertilization and increased growth of plants due to the enhancement of leaf-level CO 2 can be counteracted by large vegetation mortality rates due to more frequent droughts. 48 Similarly, the closure of the stomata under elevated CO 2 reduces leaf-level transpiration, although this may be offset by increases in total leaf area. 49 Both interactive processes influence how the available radiation is partitioned between evaporation and sensible heat flux. To the right in Figure 3, at the second group, the light perturbed by clouds and intercepted by canopies drives photosynthesis, and influences leaf-level transpiration via changes in leaf energy balance and stomatal conductance. Here, under future climate scenarios, the two opposite effects are the following: clouds could become less frequent and less active due to the increase in thermal stabilization of the atmosphere. However, clouds could become more dynamically active and vigorous due to larger evapotranspiration rates driven by regional warming that enhance the capacity of the atmosphere to hold water vapor. It is within this context that the organization of clouds in terms of complexity 50 can play a key role in regulating the radiation reaching the surface.
Scientific communities in the fields of ecology, ecophysiology, and atmospheric science have implemented initiatives to further develop and integrate land surface and atmospheric processes. 28 (2022) addressed this latter aspect when examining and comparing the formulation and performance of widely used leaf photosynthesis models. In the atmospheric science community, 54 Emanuel (2020) stressed the need to maintain a balance between observations, theoretical concepts, and supportive and integrative numerical experiments. 23 To accomplish this, we need to first determine the relevance of each individual process on the representative scale and any subsequent interactive effects occurring when the individual processes are interconnected.

CHALLENGE 1: RADIATION AND PHOTOSYNTHESIS
The transfer of shortwave and longwave radiation throughout the atmosphere is a crucial component of the interactions of the landatmosphere system as they represent the energy input into the ecosystem. Although routine measurements of the upward and downward/short-and longwave radiation components to obtain the radiative budget, and therefore, the net available radiation energy are regularly carried out, 39  influence the fluxes of heat, moisture, and carbon dioxide at the canopy level and the feedback with the clouds.
These more advanced measurements of radiation should be accompanied by an improvement in the radiative transfer calculations, in particular, the study of the radiative transfer in three dimensions and how this transfer is perturbed by the presence of clouds and the penetration within the canopy. Our reasoning is that despite the fact that radiation is a primary variable, our calculations still remain inaccurate when combining clouds and canopy effects, which has an impact on key processes related to photosynthesis and the surface energy balance. Figure 5 shows the differences in the global shortwave radiation and sensible and latent heat fluxes calculated by solving the radiative transfer either by one-(top panel) or three-dimensional radiative (bottom panel) transfer models. 19 These numerical experiments were performed using a fully coupled soil-plant-turbulent-cloud model, that is, using a large-eddy simulation technique (LES). LES numerical experiments calculate clouds explicitly 18,59 and the ecophysiology and soil processes with evaluated mechanistic representations. 23 As discussed by Veerman et al. (2020) 19 for the one-dimensional radiative transfer, the reduction in direct radiation below clouds is partially compensated by an increase in diffuse radiation, but it leads to an overall reduction in the sensible and latent heat fluxes. This decrease happens in the source region where the thermal plumes originate. This leads to a weakening of the turbulent transport of heat and moisture and a flow situation characterized by less cloud cover and smaller volume clouds. 59 In turn, the displaced cloud shadows with three-dimensional radiative transfer are responsible for decoupling the system between surface properties and clouds, which leads to the formation of atmospheric circulations. 60 The level of coupling between land and atmosphere exchanges also depends on the background wind 61  Note that the scintillometer technique, unlike eddy-covariances, does not rely on integration over all eddy scales that contribute to the turbulent transport. Rather, it determines turbulence variables, structure parameters of temperature, and dissipation rate TKE on eddy scales that lie in the inertial range of the refractive index spectrum which are linked to fluxes using MOST. The disadvantage of the technique is that it is more indirect because it relies on inertial range behavior of the observed eddies. The advantage is that it yields turbulent fluxes at much shorter timescales ( Figure 6).
As shown in Figure 6, and compared to the widely used temporal resolution of the 30-min average eddy-covariance technique, the 1-min DBLS observations not only closely follow the cloud passages that lead to radiation fluctuations, but also reveal that there are delays in the responses of evapotranspiration and NEE fluxes to radiation. For this concrete situation, the lag between radiation and evaporation and the NEE is 2 min. 23 We argue that this way of measuring NEE could lead to more precise estimations in quantifying the CO 2 assimilation by plants . Under similar and symmetric shortwave incoming radiation levels, the normally higher temperature in the afternoon leads to higher water VPD which causes the stomatal aperture to close. As a result, there might be a shift in the partitioning of the surface energy balance to higher sensible heat flux that could influence the boundary layer dynamics. As shown in Figure 7A, the asymmetry is characterized by more elliptical shapes in the case of low vegetation (winter wheat) as compared to the tall canopy of the Pine Forest ( Figure 7B).
Our explanation is that more active sweeping and ejection motions at the canopy-atmosphere interface as shown in Figure 8 introduce additional random motions to the turbulent ones leading to more chaotic patterns in the diurnal variability in the canopy fluxes of heat, water, and carbon.
As discussed by Zhang et al., 70 the understanding and analysis of these asymmetric curves enable us to better identify which are the connecting it to observations of turbulent fluxes (Figures 6 and 7) to simultaneously analyze the dynamics of the ABL (Figure 9).
In connecting these findings to atmospheric chemistry, it is important to mention that this level of spatiotemporal detail in solving the interaction between the canopy and the atmosphere is also necessary to study the production of new particles and product formation due to chemical reactions within the canopy. These processes are controlled by the distribution of sunlit/shaded radiation, that is, photolysis rate (see Figures 4 and 5), as well as the three-dimensional spatiotemporal distribution of temperature, differences in water vapor pressure, and turbulent mixing, which strongly vary in the canopy, as shown by to an extra input of energy ( Figure 10B). However, the effect of adding extra energy was clearly larger when the energy was added to the sensible heat flux. In the Warm run, the maximum cloud cover amounted to 4.3%, but in the Wet experiment, it was 2.8% ( Figure 10B). Additionally, during the period from 1400 LT to the end of the simulation period (2100 LT), the mean cloud cover amounted to 2.1% for Warm, and only 1.4% for Wet. The differences are not only visible in the cloud cover, but also in the liquid water path which directly impacts the disturbance of radiation by the cloud (Figures 4 and 5). Differences as large as a factor of 2 occur between the numerical experiments ( Figure 10C). To a certain extent, this result is counterintuitive. The explanation is that due to the more intense turbulent eddies that reach higher heights, condensation is favored by colder temperatures. As such, the vertical and horizontal organization of clouds is strongly driven by the land conditions that might bring the ABL conditions out of equilibrium with local (surface) processes and nonlocal processes (entrainment, advection, and subsidence).
It is necessary to stress that this case is highly idealized and other relevant factors, such as the presence of aerosols, thermodynamic, and boundary-layer dynamics, as well as synoptic/mesoscale features, such as subsidence and the presence of localized shear (see discussion in Figure 11), also play a relevant role as described by Teuling et al. 100 Expanding on this, and illustrated by Figure 1, it is important to understand how clouds organize during the transition from marine to land conditions. Here, we propose to connect studies on self-organization by clouds carried out under marine ABL conditions 50 to research studies that focus on the spatial distributions of biophysical and surface properties, such as albedo and canopy resistance. 37,101 Inspired by these marine studies, we argue here that it is necessary to study whether these findings of cloud self-organization over the sea, 102 key in determining the cloud dynamics, radiation disturbance, and precipitation, are relevant over land. 16 In short, these studies such as the Normalized Difference Vegetation Index or near-infrared reflectance of terrestrial vegetation. 104 By identifying regions with different photosynthetic capacities due to leaf ages, C3/C4 pathways, and/or water availability, 105 we will attempt to obtain correlations between vegetation-state patterns and cloud organization.

Challenge 3.2: Short-and long-range transport of moisture and greenhouse gases
When connecting and coupling local processes with regional patterns of vegetation and clouds, it is essential to identify how remote air masses are interacting with the local radiative, turbulent, and atmo-spheric composition conditions. The description shown in section 3.1 is a step forward, but the methods and results remain static in time and space. To identify how air masses change as they come into contact with ecosystems conditions, we can use Lagrangian or other tracking methodologies at short-range (less than 50 km) and long-range (up to 1000 km). Here, we provide two examples of transport studies using (1) turbulent explicit models and (2) regional-global models. They can help to illustrate how we determine source regions of moisture or carbon dioxide on a wide range of spatial scales (local to regional) and temporal (daily to yearly) and show the advantages of using these tracking tools in land-atmosphere studies analyzing either turbulent explicit model results or regional-global model results. Figure 11 displays an example of the tracking of simulated shallow clouds in the Amazonian basin during the dry season (July-October). 23,106 The numerical experiments were done using high-resolution simulations constrained by the GOAMAZON14 experiment. 107 Here, we focus on the canopy-cloud scales. We are The remaining questions concern how these findings connect to the specific land properties and how the long-range transport of moisture connects to the specific ecosystem ( Figure 1).
Moving to a larger-scale perspective, over the last few years, several methods have been used to identify the sources and sinks of moisture related to specific weather situations and their dependencies on seasonality and climate. Here, we present an example of the anomalous moisture source regions during the drought which occurred over western Europe in the summer of 2018. 108 The drought was the result of an anomalous weather situation, with a persistent high-pressure system over Scandinavia for most of the summer (illustrated with the anomalous 500 hPa geopotential height in Figure 13). The moisture sources during this period were identified with the Eulerian offline tracking tool WAM-2layer, 109 using ERA-Interim reanalysis data for May-August, and the results were compared against tracking results for the long-period (1979-2018) summer mean. The anomalous sources ( Figure 13) were of continental origin more than oceanic origin as compared to climatology. The high-pressure system redirected the westerly moist flow from the Atlantic away from western Europe toward the southern Alps and southeastern Europe. The precipitation that fell in western Europe during the drought was mostly from local origin or originated from eastern Europe following the anomalous anticyclonic flow. The evaporation recycling ratio over southern Scandinavia was 6% in 2018 compared to 10% for the base period. This indicates that the drought in that region self-intensified due to positive soil moisture-evaporation-atmosphere feedbacks. In the context of this paper, it shows that the importance of land-surface feedback might be enhanced during certain weather events, such as droughts, 79,110 which might become more dominant in the future ( Figure 3). Therefore, we emphasize the need to include a detailed analysis of the surface exchange and transport of heat, moisture, momentum, and atmospheric composition in regional and climatological studies, with the aim of investigating how the land-atmosphere interactions are changing under current climate conditions and more frequent weather extremes. 79,111 Furthermore, on a regional level, it is necessary to integrate the long-range tracking of moisture and carbon to determine the similarities and differences in their sources and sinks. For example, several studies have focused on deriving the CO 2 balance for the Amazon. 112,113 Here, local and regional scale long-term observations of the Amazon region. The combination of these local-to regionalscale observations allows us to derive the carbon balance for the Amazon, where droughts and fires lead to a reduced capacity of the carbon sink, and therefore, are highly relevant to our understanding.
In summary, observations and models on local ( Figure 11) to regional ( Figure 13) scales enable us to identify and separate the local and nonlocal contributions from moisture and carbon dioxide budgets.

CHALLENGE 4: INTEGRATING SCALES AND PROCESSES: PRESENT AND FUTURE
In this perspective article, we argue that both the short-and long-range tracking methodologies will benefit from new numerical techniques that combine both the explicit treatment of the land use and atmosphere exchange as well as clouds as a continuum, that is, LES, with realistic and accurate embedding on large meso-and synopticweather scales. To be consistent with the fine mesh of the numerical atmospheric dynamics and physics part, this should be done with land and topographic information at the same fine spatial resolution as the atmospheric model. Figure 14 shows This research strategy can be extended and completed using suites of numerical experiments with similar numerical settings to determine how strongly processes such as CO 2 assimilation by photosynthesis and plant transportation are interacting with clouds under future scenarios characterized by enhanced levels of CO 2 concentrations and regional warming. 117 We also argue that this integrative approach in connecting scales should be extended to chemically active species. As shown in the recent work by Ye et al., 118 surface heterogeneities (in the case of river-land contrast) drive atmospheric circulation of important precursors of aerosol formation, such as isoprenes.
This integrative approach of scales and processes is also suitable to investigate how surface exchange processes will evolve under climate change (see Figure 15). 117 Numerical experiments performed with the LES technique produced three different scenarios influenced by climate change ( Figure 3): (1) fertilization effects at the ecosystem level due to more optimal enhanced CO 2 conditions, (2) regional warming due to the enhancement of greenhouse gases, and (3) the combined effects of the first two scenarios. As an example, Figure 15 shows the most representative variables and flux exchange. Focusing on CO 2 assimilation and evaporation, mainly governed by plant transpiration in this numerical example, we find increases of up to 25% CO 2 uptake rates compared to present conditions in the scenario with only CO 2 fertilization, where the increase of CO 2 assimilation due to warming is only 7%. The combined scenario with high CO 2 concentrations and warming shows a nonlinear increase of 35%. With respect to evaporation, we find a decrease of 13% in the CO 2 assimilation scenario due to stomatal closure and an increase of up to 11% in the enhanced warming due to the higher capacity of the atmosphere to hold water.
Here, we need to stress the offsetting of both, CO 2 fertilization and regional warming, which makes it more difficult to determine which effect will be dominant. The findings discussed above have an advantage in that processes and their couplings are explicitly represented under future conditions. However, the calculations are limited in the horizontal extent (maximum domain surface are hundred kilometers).
To study the impact of land-atmospheric interactions from regional to global scales, we require the use of global climate models.
To understand the counteracting effects of enhanced warming and enhanced CO 2 concentrations, it is necessary to study the impact of plant physiology effects at the regional level. To show this, we base our discussion on the recent study by Park et al. 8 Figure 16 shows the results of 12 state-of-the-art Earth System Models (ESM) from Coupled Model Intercomparison Project Phase 6 (CMIP6). ESM models allow us not only to study the dominant effects as described in Figure 3, but also to determine how land-atmospheric interactions change at the regional level and at climatic temporal scales. They also reveal the discrepancies among the models with regard to representing these interactions. It also enables us to quantify the level of uncertainty in our future predictions of key variables, such as the NEE and cloud cover. We focus our analysis on the Amazonian basin and we show in Figure 16 the variability of cloud cover and NEE. Here, to isolate the plant physiology effect, the radiative effects are kept at preindustrial values but carbon cycle sees 1% yr −1 CO 2 increase to quadrupling for 140 years (from 285 to 1140 ppm). Therefore, the temperature changes in Figure 16 are solely driven by modifications in plant physiological processes.
The two upper panels of Figure 16 show how surface temperatures yield modifications in cloud cover and the NEE with respect to the preindustrial period. Here, two opposite effects are at play: F I G U R E 1 6 Upper two panels: Scatter plot of the differences in near surface temperature and cloud cover (left) and net ecosystem exchange (right). Middle and bottom panels: Composite maps of annual mean changes resulting from CO 2 physiological forcing in total cloud fraction (top), and net ecosystem exchange (bottom) for the top four models with the greatest temperature responses (left) and the bottom four models characterized by the weakest temperature anomalies (right) in the Amazon region from Park et al. 8 First, we see the effect of fertilization of CO 2 that leads to a stomatal closure, and therefore, a decrease in plant transpiration that leads to a higher sensible heat flux. The second effect, opposite to CO 2 fertilization, is an increase in biomass and a higher leaf area, which leads to increased light interception and canopy transpiration. These effects enhance evapotranspiration and therefore lead to a decrease in the sensible heat flux. In spite of significant differences among the model results, the relatively high correlation between surface temperature and cloud cover indicates that the models more consistently reproduce the effect of enhanced warming leading to a reduction in the cloud cover. Our explanation for this, which is supported by previous studies of vegetation-boundary layer interactions by means of a conceptual model 24 and by vegetation-regional climate models, 8 120 show that the increase or decrease in precipitation is the result of a combination of local and remote effects in which the ocean-land contrast and dominance play a key role.
The difference in CMIP6 biophysical representations and its impacts are shown in the spatial maps: middle and bottom of Figure 16.
Focusing on the Amazonian basin, the carbon-climate models characterized by large surface temperature sensitivity show a decrease in the cloud cover. Several interrelated processes can explain this behavior, but here we focus on the most relevant of the land-atmosphere interaction. As shown in Figure 13, CO 2 fertilization and enhanced warming lead to changes and shifts in evaporation and sensible heat flux. In model results in which the evaporation is dominant, and under similar conditions of available radiative energy, the sensible heat flux at the surface decreases, constrained by the surface energy balance. As a result, turbulence will become less intense and will have a smaller capacity of transporting moisture to the level of condensation, and consequently, less clouds will be formed.
The models that are less sensitive to surface temperature show a minor decrease in the cloud cover. Similar patterns are found for NEE with much larger decreases for temperature-sensitive models. These larger differences in the spatial variability among model results illustrate the difficulties in accurate projection of plant physiology on meteorology and atmospheric composition, which in the long term may influence the regional climate. Here, we advocate for dedicated studies using CMIP6 models to determine and quantify the

CONCLUSIONS AND PERSPECTIVES
This perspective article aims to advance our ability to connect and integrate processes that are usually treated separately in the disciplines of biology, chemistry, or atmospheric physics. As key components of the water and carbon cycles, we focus on intrinsic relationships in photosynthesis and plant transpiration as the most representative land processes that influence cloud formation and carbon exchanges depending on the scales where they are acting, from leaf to regional scales. These processes play key roles in determining the surface energy balance, the boundary-layer dynamics, and clouds. In connecting as a continuum these local land-atmosphere interaction processes on the one hand with weather and climate on the other hand, we aim at improving our understanding of the transport and fate of greenhouse gases, reactive species, and aerosol formation. By improving the representations of these processes, normally characterized by small and short spatiotemporal scales, and their couplings, we aim to reduce uncertainties in estimations of NEEs and cloud feedback at regional scales. The four interactive challenges set out in this paper argue for the need to move forward in: This is key to disentangling the contrasting effects of individual processes. For instance, under rising CO 2 concentrations, a reduction in leaf-level transpiration due to stomatal closure might be compensated by an increase in the total-leaf area that leads to shifts in the partition between sensible and latent heat flux at the surface. All these canopy-cloud interactions are region-specific and dependent on climate change.
Challenge 3: Organization of atmospheric flow influenced by land properties and regional weather and atmospheric composition. We suggest integrating these small and short spatiotemporal processes into regional weather and atmospheric composition since these are factors that continuously interact with the local conditions. In extending how these local conditions are influencing the spatial patterns of vegetation and clouds, we present methods that enable us to quantify which variables or processes govern the organization of clouds. A natural extension of these studies is to relate them to vegetation patterns to determine the level of coupling between cloud and ecosystem organizations. Both aspects are relevant to improving our current estimations of the regional precipitation and NEE. To advance our understanding, we propose methodologies that track clouds and air masses from local to regional scales to determine simultaneously sources and sinks of moisture and carbon dioxide.
Challenge 4: Integrating scales and processes: present and future.
We advocate studying present and future scenarios by performing LES embedded in numerical weather prediction and carbon-climate models to attempt to reduce the uncertainties in the feedback of clouds on surface processes. We also need to investigate how locally driven surface processes influence weather and climate in forming and intensifying clouds. In the current analysis of climate model results, we have evidence that there are important differences between the present situation compared to preindustrial conditions that are keys for representative variables, such as cloud cover and NEE. We argue that these differences are due not only to crude representations of plant physiology effects and clouds, but also to the need to treat the atmospheric flow as a continuum that connects short-to large-scale processes and phenomena. If these effects prove to be relevant, we need to take them into account in weather and climate models along with the effects of spatiotemporal changes in atmospheric composition. Here, we should pay special attention to the fact that some effects occur at shorter time scales, such as stomatal closure or boundary-layer cloud formation. Others may take longer to emerge, such as canopy effects because of an increase of the LAI and the atmospheric radiative effects driven by an increase of the greenhouse gases.

AUTHOR CONTRIBUTIONS
J.V.-G.A. designed and wrote the paper. The rest of the co-authors provided figures and feedback during the last 5 years. All the authors read the paper and commented.

ACKNOWLEDGMENTS
The following grants are acknowledged in providing observations and model calculations used in some figures of the paper: project Cloud-

COMPETING INTERESTS
There are no competing interests.