Show Summary Details

Page of

Printed from Oxford Research Encyclopedias, Climate Science. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 22 June 2021

# Terrestrial Processes and Their Roles in Climate Change

• Nathalie de Noblet-DucoudréNathalie de Noblet-DucoudréClimate and Environment Sciences Laboratory, CEA-CNRS-UVSQ, Paris-Saclay University
•  and Andrew J. PitmanAndrew J. PitmanCentre of Excellence for Climate System Science, Australian Research Council

### Summary

The land surface is where humans live and where they source their water and food. The land surface plays an important role in climate and anthropogenic climate change both as a driver of change and as a system that responds to change. Soils and vegetation influence the exchanges of water, energy and carbon between the land and the overlying atmosphere and thus contribute to the variability and the evolution of climate. But the role of the land in climate is scale dependent which means different processes matter on different timescales and over different spatial scales.

Climate change alters the functioning of the land with changes in the seasonal cycle of ecosystem growth, in the extent of forests, the melt of permafrost, the magnitude and frequency of disturbances such as fire, drought, … Those changes feedback into climate at both the global and the regional scales. In addition, humans perturb the land conditions via deforestation, irrigation, urbanization, … and this directly affects climatic conditions at the local to regional scales with also sometimes global consequences via the release of greenhouse gases.

Not accounting for land surface processes in climate modelling, whatever the spatial scale, will result in biases in the climate simulations.

### Subjects

• Climate Systems and Climate Dynamics

### Introduction

The land surface is a critical component of the Earth System. Most immediately, it is obviously the place people live and source most of their food, water, and energy. The effects of climate change are most commonly thought of in terms of impacts on human systems, which are largely land based. However, the land also has a role in the climate system in terms of both anthropogenic climate change and the natural climate changes that have affected our planet. The key roles of the land surface include the storage of carbon on geological, millennium, and seasonal timescales. The land stores water, in some cases on millennium timescales in the form of deep groundwater and permafrost. Finally, the land stores energy, in the form of heat, on the diurnal, seasonal, annual, and even millennium timescales again for example in the form of permafrost.

To set the scene, on geological timescales, the land plays a central role in long-term climate change, primarily through its impact on the key greenhouse gas, carbon dioxide $CO2$. Ciais et al. (2013) discussed carbon on two timescales. A slow cycling of carbon though the Earth System on millennium timescales linked with interactions between the cycles of calcium carbonate and calcium silicate that lead to massive storage of carbon dioxide as limestone rock. Weathering of these rocks leads to leaching of dissolved calcium, carbon, and silica, which are transported to the oceans and used to form the inorganic skeletons of benthic organisms and plankton. A faster cycle exists linked with a very large exchange of fluxes between the atmosphere and the land via vegetation and soils. Butcher et al. (1992) provided a thorough assessment of a range of global biogeochemical cycles, including nitrogen, phosphorous, and sulphur. (An up-to-date overview of the carbon, methane, and nitrogen cycles is provided by Ciais et al., 2013.)

This article focuses on the timescales of relevance to human-induced climate change, but with an example of how it can also contribute to longer timescales. How the land surface contributes or is affected by human-induced climate change at timescales shorter than centuries is explained. The spatial scale will tend to be global through to regional in terms of processes. The extreme challenge of the role of land processes at the scale of a catchment or farm is beyond our scope.

The article is organized as follows: First, the basic role of the land is explained. Second, how that role is expressed on daily to seasonal and decadal to larger timescales is discussed. A discussion of how these roles act as drivers of and moderate climate change includes case studies. The role of land management as an additional driver of how land impacts climate is addressed. The most important remaining challenges are presented and the role of land processes is examined through the lens of how these processes are represented in the key tool used to understand our climate and climate change: global and regional climate models.

### Basic Theory Connecting the Land Surface With Climate

Basic physics tells us that energy and mass (in the form of water and carbon) must be conserved. Our understanding of land surface processes starts from these principles. The surface energy balance, the surface water balance, and the surface carbon balance are key to understanding how the land affects climate and how climate change might affect the land.

#### The Surface Energy Balance

The Sun has a temperature of approximately 6000 oC, and according to Wien’s displacement law the peak of the emissions curve of electromagnetic radiation by a body with this temperature is ~0.5 μ‎m. We know this as shortwave radiation, and this radiation is reflected, absorbed, or transmitted by the atmosphere. Energy from the Sun $S↓$ reaches the Earth’s surface (184.7 W m-2 averaged over all land areas and over the year; Trenberth et al., 2009) and some (39.6 W m-2) is reflected (depending on the land surface albedo α‎,1 which is about 21.44% when averaged globally and over the year), leaving 145.1 W m-2 to warm the land surface (Figure 1). Everything with a temperature warmer than absolute zero (-273.16oC) emits electromagnetic radiation. The atmosphere emits radiation known as infrared radiation, which depends on temperature and emissivity (the effectiveness of emitting thermal radiation) of the atmosphere. The annual amount of atmospheric radiation received at the land surface, when averaged globally, is 303.6 W m-2 $L↓$. The surface also emits infrared radiation ($L↑$, 383.2 W m-2), again depending on its temperature and emissivity leaving a net infrared flux at the surface of -79.6 W m-2. The net balance of the incoming and reflected shortwave radiation and the incoming infrared radiation at the Earth’s surface is called available energy, $AE$, which is basically the energy that subsequently drives all terrestrial processes. In the following for available energy, all fluxes are in W m-2:

(1) $Display mathematics$

The net balance of all the radiative fluxes is known as the net radiation $Rnet$ and is expressed as:

(2) $Display mathematics$

According to Trenberth et al. (2009), $Rnet$ is 65.5 W m-2, comprising the 145.1 of solar radiation minus 79.6 W m-2 of infrared radiation (Figure 1). By convention, the radiative fluxes are positive downward (a positive flux means the land gains energy).

This net energy, $Rnet$, is partially returned to the atmosphere in the form of sensible (convection of dry heat, $H$) and latent heat (convection of water vapor, $λE$) fluxes, to the deeper soil layers in the form of ground heat flux ($G$), and a very small portion of it is stored during photosynthesis and released by respiration (this amounts to less than 1% of absorbed insolation [Sellers, 1992] and is known as chemical energy, $F$). $H$ and $λE$ are 27.0 and 38.5 W m-2, respectively (Figure 1), while G is very small on annual and longer timescales unless there is a trend in temperature but can be large on daily timescales. The fluxes of $H$ and $λE$ occur in the presence of turbulence, and are therefore known as the turbulent energy fluxes. They also require that there exists a temperature (for $H$) and a humidity (for $λE$) gradient between the land and the atmosphere to drive a transfer of heat or water vapor. The latent heat flux (also known as evapotranspiration) requires water and is the physical link between the surface energy balance and the surface water balance. The land also stores energy on diurnal, seasonal, and longer timescales (thousands of years in the case of heat stored in permafrost).

Conservation of Energy states that $Rnet$ (Equation 1) must be balanced by $H$, $λE$, $G$, and $F$ such that (in W m-2):

(3) $Display mathematics$

Trenberth et al. (2009) notes $H$ and $λE$ are 27.0 and 38.5 W m-2, respectively—a total of 65.5 W m-2—matching $Rnet$ and implying $G$ and $F$ are negligible when averaged globally over all land and annually (Figure 1). By convention, the fluxes of $H$, $λE$, $G$, and $F$ are positive when the land surface loses energy (upwards for $H$, $λE$). The land directly affects Equation 1; that is, the amount of energy captured, via the surface albedo, which depends on soil color, vegetation type, density and greenness, and soil moisture. It also directly affects $Rnet$ in Equation 2 via the land emissivity and thus the emitted infrared radiation, and it also affects how $Rnet$ is partitioned or split between the fluxes on the right-hand side of Equation 3. $H$ and $λE$ are often partly controlled by the near-surface atmospheric conditions, which in turn are linked to turbulence. This follows atmospheric boundary layer theory, which is complex and beyond our scope (for details, see, e.g., Bonan, 2019; Garratt, 1992; Stull, 1988). The key properties of the atmosphere—temperature, water vapor, and wind—are inherently coupled, and they are simultaneously coupled with the sensible and latent fluxes and need to be considered together. While it is tempting to view the land surface as an independent unit, the interaction between the surface and the atmosphere via the energy and water fluxes must be understood as a coupled system. A key element of these fluxes is the surface roughness. A rougher surface slows the horizontal flow of the atmosphere more effectively than a smoother surface and generates more turbulence. As discussed by Bonan (2019), the roughness length is a complex function of the structure of the vegetation, the number of leaves, and so on. The horizontal patterns of vegetation (i.e., vegetation is broadly spatially homogeneous, there are clearings or patches of forest, etc.) and topography also affect the roughness length. Note that the aerodynamic roughness length for momentum is different from the aerodynamic roughness length for the sensible and latent heat fluxes. How $λE$ is affected by land is more thoroughly discussed in the following section (Surface Water Balance) below.

#### The Surface Water Balance

Approximately 111,000 km3 y-1 of precipitation (in the form of liquid water and snow in cold regions) reaches the land’s surface. It is either intercepted by vegetation or reaches the soil surface. If precipitation is intercepted, it subsequently either evaporates (interception loss) or drips to the surface. The combination of drip and the rainfall that reaches the surface directly then either infiltrates the soil or runs across the soil surface (surface runoff). If precipitation is in the form of snow it is stored on the ground, is sublimated, and melts during spring. Water that infiltrates the soil (liquid water or snowmelt) may subsequently evaporate from the soil surface (soil evaporation), drain through the soil and reach deeper layers, or be taken up by roots and transpired through stomates, which are the pores on the surface of leaves through which CO2 is absorbed and water vapor released. Oki and Kanae (2006) estimated total land evaporation at about 65,500 km3 y-1. Water in the soil is known as soil moisture and is estimated at about 122,000 km3 (Trenberth & Asrar, 2014), and a further 15,300,000 km3 is stored as groundwater and plays a vital role in agriculture and natural ecosystem support. When considering all components of evaporation—that is, sublimation, interception loss, soil evaporation, and transpiration—we generally refer to evapotranspiration, which is a major component of the continental water cycle and returns about 59% of all land precipitation back to the atmosphere (e.g., Oki & Kanae, 2006; Figure 2).

The land surface controls how available water (assumed here to be precipitation, $P$, but it could also include snow and permafrost melt) is partitioned or split between evapotranspiration $E$ and its various components and runoff $R$. Total runoff $R$ is commonly split between a fast component that flows across the surface $Rsurf$ and a slower component that drains through the soil column $Rdrain$. $Rsurf$ (15,300 km3 y-1) and $Rdrain$ (30,200 km3 y-1) can be summed to give R (45,500 km3 y-1; see Oki & Kanae, 2006). Overall, this balancing of incoming and outgoing fluxes of water is called the surface water balance (Figure 2) and can be expressed in many ways. Here, we assume the fluxes are in kg m−2 s−1, which, if divided by the density of water (units, kg m-3), can be converted to m s-1:

(4) $Display mathematics$
(5) $Display mathematics$

By convention, positive $P$ means an addition of water to the surface, while for those fluxes on the right-hand side a positive value means a loss of water to the surface. Note that $E$ can be negative when dew occurs. $ΔS$ is the change in storage (liquid or frozen soil moisture or snow) and is commonly assumed to be zero on long timescales, although loss of permafrost or long-term declines in soil moisture can mean that $ΔS$ is a significant term. Note that the hydrological cycle uses $Ekgm‐2s‐1$; this is linked to the surface energy balance, which uses $λEWm−2$, where $λ$ is the latent heat of vaporization (2.5 x 106 J kg−1), simply by multiplying $Ekgm‐2s‐1$ by $λJkg‐1$ to give units of $Jm‐2s‐1$ and remembering that a $Js‐1$ is equivalent to a watt $W$. $λ$ has a slightly larger value for sublimation 2.834 x 106 J kg−1.

The two runoff terms, $Rsurf$ and $Rdrain$ in Equation 5, are influenced by soil characteristics (sand-silt-clay content, depth, compaction, etc.) and the amount of water stored in the soil. A change in the distribution or type of vegetation modifies the balance between fluxes originating from the soil and those derived through canopy as well as the amount of water that is intercepted by the canopy. $S$ provides memory. The actual amount of moisture stored in the soil at a given time reflects previous variations in $P$, $E$, $Rsurf$, and $Rdrain$ at the site.

As the amount of water stored in the soil decreases, it affects evaporation from the soil and the ability of vegetation to take up water through the roots. There is therefore a very strong feedback between $S$ and $E$ in Equation 5 (Seneviratne et al., 2010). Evapotranspiration is a complex process and many ways to represent it have been proposed. Sellers (1992) provides a simple form:

(6) $Display mathematics$

where $e∗T9$ is the saturated vapor pressure at temperature $TsK,er$ is the vapor pressure at a reference temperature above the surface, $ρ$ is the air density (kg m-3), $cp$ is the specific heat of air (J kg-1 K-1), $γ$ is the psychrometric constant (Pa K-1), and $r<$ is the aerodynamic resistance to the exchange of water vapor (s m-1). This leaves $β$, which varies between 0 for a perfectly dry surface and 1 for a perfectly wet surface and combines resistance to the flow of water from the soil into the atmosphere, the uptake of water by the roots, resistance to the flow of water through the plant, and resistance to the flow of water through stomates on the leaf. Most land models separate these resistances and treat each independently. The key points here are that water in the soil strongly affects the latent heat flux (so evapotranspiration) and that soil moisture varies strongly with time and with place.

Combining the surface energy and water balance leads to some important concepts. The way the land interacts with the atmosphere can be energy limited or water limited (Seneviratne et al., 2010). In the Arctic tundra and in tropical forests there tends to be plenty of water; what limits evaporation is the amount of energy $Rnet$ available to drive the evapotranspiration. In contrast, in the midwestern region of the United States, the Mediterranean, and hot deserts there is very high $Rnet$ during summertime, but evapotranspiration is limited by the availability of water. In some regions the limitations to $E$ vary through the season—for example, midlatitude grasslands commonly start in spring with high water availability and relatively low $Rnet$ (energy limited), but through the season the soils dry such that the exchanges with the atmosphere become water limited. The timing of this transition depends critically on the vegetation type. A shallow-rooted grassland tends to become moisture limited far earlier in the season than a deeply rooted forest. Thus, the nature of the vegetation strongly affects the nature of the land-atmosphere interaction.

#### The Surface Carbon Balance

The surface takes up carbon dioxide via photosynthesis (also called gross primary productivity, GPP) while simultaneously losing water via transpiration since both fluxes are exchanged through stomates. Photosynthesis requires a very small amount of energy (~1%), which in the case of plants is sunlight $S↓$ at particular wavelengths known as photosynthetically active radiation. The chloroplast in the leaf captures photons of photosynthetically active radiation and uses this energy to split water molecules. The resulting hydrogen atoms are then combined with $CO2$ molecules to produce glucose with the byproduct of oxygen molecules. This can be expressed simply as:

(7) $Display mathematics$

Note the link with the surface water balance (the $6H2O$ is sourced from soil moisture). The glucose is converted into starch for storage or into cellulose for growing trunks and other physical structures. These are all basically carbon and represent removal of carbon from the atmosphere that remains on land as long as wood is not burned and plants do not die (thus from seasons to centuries). Plants also respire, which returns carbon dioxide to the atmosphere in a process known as autotrophic respiration $Ra$. When a plant dies and decomposes, or is burned, the carbon is returned to the atmosphere. In addition, respiration by microorganisms in the soil break carbon down and release $CO2$, a process known as heterotrophic respiration $Rh$. This process is very complex, but both temperature and moisture affect the rates of chemical reactions.

The scale of these fluxes, on an annual basis and accumulated over all land ecosystems, is enormous, with about 123 PgC yr-1 being taken up via $GPP$ on the land (a sink of carbon) and 118.7 PgC yr-1 being lost via respiration (a source of carbon), 60 PgC yr-1 being lost as $Ra$, 50 PgC yr-1 as $Rh$, and an additional 8.7 PgC yr-1 lost through disturbances including fire (Ciais et al., 2013). These numbers are large in comparison to human emissions from fossil fuels (~9.5 PgC yr-1; Friedlingstein et al., 2019) plus ~1.5 PgC yr-1 of net emissions from land use change (essentially deforestation). The net balance, at the vegetation level, between $GPP$ and $Ra$ is called net primary productivity ($NPP$, Equation 8). At the ecosystem level, $Rh$ needs to be removed from $NPP$, which forms the net ecosystem productivity ($NEP$, Equation 9). $NEP$ at the annual scale and over all land amounts to 4.3 PgC yr-1. It is positive because as the net effect of terrestrial ecosystems acts as a sink of $CO2$ (Shukla et al., 2019). Land therefore helps minimize increases in atmospheric $CO2$ CO2. Similar to Equation 5, we may thus write the following equations for the carbon balance over land:

(8) $Display mathematics$
(9) $Display mathematics$

The biggest reservoirs on land are the soils that hold between 1,500 and 2,400 GtC, followed by permafrost, which holds about 1,700 GtC. Vegetation is a smaller reservoir of between 450 and 650 GtC.

Warming of climate risks the stability of all these reservoirs.

### The Role of the Land Surface and Land Surface Variability in Climate

The role of the land surface in climate change is spatially and temporally scale dependent. Aggregated to the global scale, historical changes in land cover (basically an increase in cropland area at the expense of vast amounts of forests) led to the release of 5.5 ± 2.6 GtCO2 in the atmosphere (Shukla et al., 2019). This contributed to the warming of the climate by about 0.20 ± 0.05 °C to 0.25 ± 0.10 °C annually, depending on the estimates. Simultaneously, changes in surface albedo and sensible and latent heat fluxes contributed to an estimated global annual cooling of -0.10 ± 0.14 °C (known as biophysical cooling). This is smaller than the warming effect of the historical changes, but the estimates are less certain. The contribution of changes in land cover to the total change in ambient air temperature, through its biophysical effects, is small when averaged globally and annually, which led to an unfortunate misunderstanding and underestimation of the role of the land. In fact, the role of the land is highly regionalized and is greatest where people live or grow crops, and the role of the land varies from one season to another. For example, replacing forests with crops typically increases the albedo and decreases total evapotranspiration. At high latitudes, a boreal forest in midwinter can be quite dark, but a crop or grassland can be covered by highly reflective snow. A boreal deforestation will then cool the land and the overlying air during the snow season in response to the higher albedo, but will warm during summer when vegetation is green and transpiring as deforestation reduces the loss of energy through the release of latent heat. Building a city is another example of a highly regionalized impact. Urban development modifies the albedo and water availability and contributes direct input of energy via human activities, but it has little effect in the global average climate.

Discussions of the scale dependency of land interactions with climate are provided in the next sections at the shortest timescale, the daily (or diurnal) cycle, the seasonal timescale, and finally the millennial scale, which is associated with changes in the types of vegetation associated with variations in the trajectory of Earth around the Sun (Berger & Loutre, 1988). A complementary review is provided by Pielke et al. (2011).

#### Diurnal Variations of Land-Atmosphere Fluxes

Figure 3 shows the diurnal cycle for three contrasting sites: a tropical evergreen forest in Africa (green lines), a temperate C3 crop in Europe (orange lines), and a desert site in the Sahara (black line). The basic shape of the diurnal cycle for surface soil and ambient air temperatures ($Ts$ and $Ta$), evapotranspiration $E$, sensible and latent heat fluxes $HandλE$, and gross and net primary productivities $GPPandNPP$ is the result of the incident solar radiation received at the surface $S↓$. Not all variables display a diurnal cycle: leaf area index $LAI$ is basically constant through a day but of course varies strongly on a seasonal timescale (Figure 4). Note rainfall shows several rainy episodes at the rainforest site, but no rain is observed that day for the cropland and desert sites.

Figure 3 allows the nature of surface quantities, and how these vary on daily timescales, to be explored. As there is no vegetation in the desert, $LAI$, $GPP$, and $NPP$ are all effectively zero. On the day presented in Figure 3, evapotranspiration is also zero because no recent rainfall was experienced at this site. The desert site is more than 10 °C warmer than all other sites throughout the day, despite an albedo that is much larger (40% versus 15% for both tropical forest and temperate crop on this specific day) and an amount of available energy $AE$ that is similar to the other two sites (the daily mean available energy is 582 W m-2 at the desert site, 536 W m-2 at the forest site, and 603 W m-2 at the temperate cropland site). The higher soil temperature in the desert results from much less energy being dissipated in the form of turbulent fluxes (the daily mean sum of latent and sensible heat is 43 W m2 at the desert site, 105 W m-2 at the forest site, and 163 W m-2 at the temperate cropland site). This is because the desert site is relatively smooth compared to a rough forest or well-developed cropland, which highlights why assessing the role of the land purely in terms of reflectivity or impact on $S↓$ is limiting.

In addition to being warmer, the desert site exhibits the biggest diurnal cycle for temperature (25 °C, whereas the amplitude of the diurnal cycle is 23 °C for the temperate cropland and only 9 °C for the tropical forest). The large drop in temperature at night in the desert results from infrared radiation being emitted by the land in this area at night (~475W m-2). The tropical forest shows a relatively small diurnal variation and a much lower maximum daytime temperature 28.4 °C (compared to 36.2 °C for the cropland and 52.6 °C for the desert site). The differences between soil and air temperatures are the greatest at the cropland and desert sites, where surface roughness is smaller than at the forest site.

Overall, the diurnal cycles for the forest and crops are profoundly different because of the presence of water and vegetation. The higher $LAI$ for the forest (~6) than the cropland (~3) is a response to higher gross and net primary productivity. Both sites exhibit negative $NPP$ at night as canopies respire, but do not uptake $CO2$ in the absence of light. The higher $LAI$ and low water stress leads to high evapotranspiration ($E$, or latent heat flux $λE$) and low sensible heat fluxes in the forest. As the amount of available energy is larger over the cropland site owing to larger incident solar radiation at that time of the year, and because latent heat flux $λE$ is smaller over that site, there is more energy dissipated in the form of sensible heat over the cropland site.

Figure 3 highlights the role of the land. By taking an amount of energy from the Sun, the land can be cooled by both sensible and latent heat. The lower atmosphere can be heated via sensible heat, whereas latent heat only moistens it (heating occurs in the upper atmosphere when and where water vapor condenses). Both turbulent fluxes have a large impact on the temperature diurnal cycle. Forests are commonly very efficient at supplying moisture into the atmosphere. Recall that vegetation does not lose water by accident, but rather is the byproduct of photosynthesis. Figure 3 therefore shows that forests are particularly efficient at taking up CO2 from the atmosphere. Further detail is provided by Bonan (2002, 2019).

#### Case Study: Influence of Land on Heat Waves, From Daily to Seasonal Timescales

Trends have been observed in several important extremes including those linked to temperature (e.g., heatwaves) and rainfall (e.g., extreme rainfall). Any significant reductions in rainfall or increases in evapotranspiration can also deplete soil moisture and increase the risk of the occurrence or intensity of drought. The influence of the land surface on extremes on daily to seasonal timescales has evolved rapidly and is now understood to be very important under some circumstances (Miralles et al., 2018).

Virtually all extreme events require the atmosphere to provide conducive conditions. For extreme warm temperatures, this is commonly a slow-moving large-scale blocking high pressure system. Under these conditions, high net radiation can dry out the landscape and cause a repartitioning of net radiation such that sensible heat becomes the dominant flux. Intense heating of the boundary layer by sensible heating and the storage of this heat and recycling of the heat back to the surface over successive days can sustain and intensify heatwaves (Miralles et al., 2014). The key here is the limitation of soil moisture, which tends to happen faster in bare soil or shallow-rooted vegetation (grass, crops) than in a deeply rooted forest that can tap deep soil moisture. Therefore, both the short-term variation in soil moisture and the nature of the vegetation, including how humans have modified it, contribute to how a landscape influences heatwaves (Teuling et al., 2010).

If the nature of the land influences heatwaves, perhaps humans could modify the land to ameliorate heatwaves (Seneviratne et al., 2018). There is some evidence of this via agricultural strategies (Davin et al., 2014; Hirsch et al., 2017), urban design (Imran et al., 2019), and irrigation (Lobell et al., 2009; Puma and Cook, 2010). This in turn has opened up the potential to use the land to help manage heatwaves.

#### Seasonal Variations of Land-Atmosphere Fluxes

The analysis of Figure 3 reflects a single diurnal cycle. Figure 4 shows the seasonal cycle of the same variables at the same sites. The tropical rainforest site exhibits little seasonal variation in either the fluxes or the leaf area index. This is because the incoming radiation fluxes are similar throughout the year, rainfall is abundant, and the forest is evergreen. At this site, about 19% of the incoming energy is returned to the atmosphere in the form of turbulent fluxes (minimum value ~17% and maximum ~21%), primarily via the latent heat flux (~76 W m-2 on average compared to ~30 W m-2 for sensible heat). Minimum latent heat (64 W m-2) occurs during the brief dry season (July–August).

At the temperate cropland site, the marked seasonal cycle of incident solar radiation affects all other variables. Vegetation is green between April and October because crops have a sowing date and are harvested, and this is when photosynthesis occurs. Senescence begins in August when net primary productivity drops to zero. When foliage is well developed, the fraction of total energy returned to the atmosphere as turbulent fluxes is about 22%, which is as large as the tropical forest site, with similar partitioning between latent and sensible heat fluxes as discussed earlier. This means that when water is not limiting, energy is plentiful, and the canopy is well developed a cropland can be as efficient as a tropical rainforest at evaporating water and dissipating turbulent fluxes. However, as soon as soil moisture depletes and leaf area decreases, this efficiency strongly decreases and less than 5% of the incoming energy is returned to the atmosphere via turbulent fluxes. At temperate latitudes during winter, sensible heat can be negative because the atmosphere is warmer than the land. At the desert site, the ratio of turbulent fluxes to available energy never exceeds 8% because the surface is too smooth. At all sites the difference between ambient air and surface soil temperatures is relatively small at the monthly timescale, contrary to what we observe at the daily timescale.

#### Case Study: Influence of Deforestation at the Seasonal to Interannual Timescales

Deforestation can affect local and regional climate in a variety of ways (D’Almeida et al., 2007; Pielke et al., 2011; Jia et al. 2019), including via changes in albedo, leaf area, and roots that affect the surface energy and water balance (Boisier et al., 2012) and changes in surface aerodynamic roughness that can affect the magnitude of turbulent transfer of $H$ and $λE$. The replacement of forests with grasslands or crops typically increases the albedo and reduces the available energy ($AE$, prior to any atmospheric feedback), which tends to cool the land surface. This can be dampened or offset by reduced loss of energy via turbulent heat fluxes in response to a smoother surface. While the sum $H+λE$ decreases, this is not always true for the individual fluxes $HorλE$. The more shallow-rooted grassland, which often also has lower leaf area, generally results in reduced $λE$ (again prior to atmospheric feedbacks) while $H$ can either decrease following the smaller roughness length or increase to compensate a large decrease in $λE$. If $H$ increases, it warms the lower atmosphere (Bonan, 2008; Pielke et al., 2011) and deepens the convective boundary layer (Runyan et al., 2012).

In the tropical regions dominated by evergreen broadleaf forest, this change in the surface energy balance tends to be dominated by the reduction in $λE$, which leads to an overall warming during the growing season. In the boreal zone the increase in albedo is large in winter and spring because snow commonly masks part of the land surface. When forested, the snow tends to fall from the trees and leaves the dark vegetation to absorb energy and warm up in spring. In contrast, when deforested, snow increases the reflectivity of the surface, limits the absorption of $S↓$, and slows spring warming. Depending on the region, the amount of snow, and the overlying climate, this can delay spring warming (Betts, 2001). During the growing season however, and in the absence of snow, reduced $λE$ tends to dominate over the increased albedo and surface warming is observed (Strandberg & Kjellström, 2018; Schultz et al., 2017).

Deforestation also has the potential to affect rainfall because it can reduce $λE$, but how deforestation affects rainfall is complex and there are at least two counteracting feedbacks. First, Costa and Foley (2000) noted that lower rainfall recycling following tropical deforestation altered the boundary layer and tended to reduce the likelihood of rainfall. This implies a positive feedback (Runyan et al., 2012) where less forest leads to less $λE$ and less rainfall. The strength of this feedback depends on the magnitude of local rainfall recycling in the areas of tropical deforestation, which tends to be high (around 25%–35% annually, but as high as 70% during the dry season in parts of South America) in the Amazon (Eltahir & Bras, 1994; Staal et al., 2018). A second feedback is linked with moisture convergence, whereby deforestation reduces the aerodynamic roughness of the surface, which enables stronger horizontal advection. This draws more moisture into a region and might increase rainfall. The overall balance of these contributors to rainfall change is highly uncertain and some experiments simulating deforestation suggest reduced rainfall (e.g., Costa & Foley, 2000; Badger & Dirmeyer, 2015; Lawrence & Vandecar, 2015; Lejeune et al., 2015), whereas others suggest increases (Dirmeyer & Shukla, 1994; Polcher & Laval, 1994; see also, Perugini et al., 2017). This highlights the complex nature of the interactions between the land and the atmosphere. These interactions occur at small spatial scales relative to the detail represented in climate models and the spatial scale of deforestation. Using climate models to explore this set of processes therefore is limited by the skill of the models, the scale of the deforestation, and the spatial resolution of the climate models and remains an area of ongoing research.

#### Long Timescales for Land–Atmosphere Interactions

On longer timescales, decades to centuries or more, the important roles of the land are the storage of carbon, groundwater, and energy (in permafrost and ice sheets). Issues like soil moisture are extremely important on seasonal timescales but as a landscape or region enters long-term drought soil moisture tends to fall to such low values after a couple of years that the memory is effectively negligible. Under these circumstances, the memory is held in the vegetation for as long as it survives. Survival or replacement of a certain vegetation type by another is part of the long-term vegetation dynamics. Climatic changes and perturbations such as fires or insect outbreaks affect the functioning of ecosystems and their reproductive and dispersal capacities. They contribute to changes in vegetation distribution (landscapes) through time.

In general, those changes are slow unless an abrupt catastrophic event occurs—for example, a fire that destroys an entire forest. A rapid change can also occur in case of tipping points linked to land surface processes (Lenton et al., 2008). The term tipping point refers to a threshold at which a perturbation can qualitatively alter the state or development of a system. Lenton et al. (2008) identified the Amazon rainforest, the boreal forests, and the Sahara/Sahel as likely land-based candidates. The simplest of these to understand is probably the Amazon. Rainfall in this region is recycled; a lot is intercepted and almost immediately reevaporated to provide moisture for subsequent rainfall (this was examined in detail by Scott et al., 1995, and more recently in Staal et al., 2018, and Zemp et al., 2017). If something should happen to reduce rainfall, recycling would weaken and a feedback would drive subsequent reductions in rainfall across the basin. Many estimates of the impact of Amazonian deforestation identify a consequential reduction in rainfall that might trigger this feedback, but changes in ocean circulation or a change in the large-scale atmospheric flow triggered by El Nino changes could also begin to dry the Amazon (Betts et al., 2004). As a tropical forest dries, it becomes vulnerable to fire, which can cause a rapid destruction of a forest and consequential drying, leading to further reductions in rainfall. Lenton et al. (2008) suggested that land-use change could bring the Amazon to a critical threshold, lead to massive emissions of $CO2$ (e.g., ~50 GtC; Cox et al., 2004), and amplify the climate changes that started the process.

A second example proposed by Lenton et al. (2008) was the boreal forest system, which has previously been highlighted as important to climate (Bonan et al., 1992). This is a complex environment with strong impacts from snow and permafrost and with tree mortality linked with temperature and associated diseases. If moisture stresses limit the productivity of these systems, and if very cold winter temperatures become too rare to kill some pests, increased large-scale mortality can leave these forests vulnerable to fire. Note how relatively slow processes in both the Amazon and boreal regions can become very fast if fire becomes a dominant driver of regional change.

There are also abrupt changes possible in soil carbon. About half of the carbon taken up through the vegetation is stored in the soil. Indeed, soils store about 300 times the amount of carbon now released annually through burning fossil fuels (Schulze & Freibauer, 2005) on a global scale. Evidence has emerged that soil carbon is rather vulnerable to changes in temperature (Bellamy et al., 2005) due to warmer temperatures accelerating soil respiration. Again, there is a feedback operating on very large stores of carbon that leads to increases in atmospheric $CO2$ that then acts to increase warming.

These types of abrupt changes are very likely low-probability events. It is not known for certain how likely they are and it is not clear that they can be predicted. They are, however, high-impact events with profound implications for the climate system. The interested reader can refer to Alley et al. (2003), Kriegler et al. (2009), Lenton (2011), and Lenton et al. (2008) for further details.

#### Case Study: The Role of Vegetation Dynamics During the Last Glacial Inception

Climate changes naturally in response to the variations of Earth’s orbit, with an approximately 100,000-year cycle (Berger & Loutre, 1988). The last glacial inception began 115,000 years ago and culminated 21,000 years ago, a time period known as the last glacial maximum when large ice sheets covered North America (the Laurentide ice sheet) and northern Europe (the Fennoscandian ice sheet).

At the start of the glaciation, the eccentricity of Earth’s orbit was greater than it is today. The northern hemisphere was experiencing warmer winters and cooler summers. The changes in solar radiation were insufficient to explain the buildup of the very large ice sheets. Amplifying feedback processes need to be accounted for, and vegetation dynamics are among these feedbacks.

The cooler summers triggered by lower insolation were sufficient to decrease the energy required by boreal trees to grow—this energy is often calculated as growing degree days above a minimum threshold. The colder temperatures led to the southward expansion of tundra vegetation in boreal regions at the expense of taiga and boreal forest in a natural deforestation. Grass and shrubs are more prone to be covered with snow and exhibit a much larger surface albedo than forests during the snow season. Increased albedo reduces the amount of solar energy absorbed by the land and induces cooling. During winter this deforestation-induced cooling dominated the orbital-increased warming, but it enhanced the orbital-induced cooling during other seasons (de Noblet-Ducoudré et al., 1996). Colder temperatures were established at those latitudes all year long and delayed snow melting, snow became perennial in some areas, which led to the buildup of ice sheets.

Figure 5 shows two simulations from a climate model for the start of the last glacial inception: one allowing vegetation to respond to climate change and another with vegetation distribution imposed throughout the 20,000-year simulation (Kageyama et al., 2004). The buildup of ice sheets is significantly limited when vegetation dynamics are not accounted for (thin red-black-green curves on the left figure), whereas ice sheets grow quickly if forests are replaced by herbaceous vegetation. This illustrates how vegetation dynamics affect the initial natural climate change. Here it propagates the orbitally induced climatic signal of the snow-free season (cooling) to the snow season via increased surface albedo. As the vegetation change occurs in the high northern latitudes, land-atmosphere interactions influence oceans by increasing sea ice duration, which progressively cools the entire globe.

### Climate Change as a Driver of the Surface Energy, Water, Carbon Balance

There are three aspects of anthropogenic climate change that affect the land surface. The first one is the effect of increasing $CO2$ on $L↓$, which increases via the enhanced greenhouse effect (Myhre et al., 2013). The second aspect is the impact of increasing $CO2$ on the terrestrial carbon balance. Change in the surface carbon balance is dominated by three processes. First, under elevated $CO2$, many plants use water more efficiently such that water becomes less limited (all other things being equal) and this tends to enhance landscape greening (Zhu et al., 2016). Second, higher $CO2$ acts as a fertilizer (all other things being equal), potentially enabling plants to sustain deeper roots, larger leaves, and so on (Donohue et al., 2013; Wright et al., 2018). This fertilizing effect increases photosynthesis, which draws $CO2$ down from the atmosphere. In both cases, note the “all other things being equal”. In reality, changes in rainfall and increased frequency of extreme events may limit the benefits plants gain from elevated $CO2$, and in other parts of the world limits in nitrogen, phosphorous, and potassium (trace elements vital for plant growth) mean the plants cannot take full advantage of the higher $CO2$ (Jia et al., 2019). Settele et al. (2014) assessed the complexities of how land responds to these changes. Respiration increases significantly in response to warmer land and releases $CO2$ to the atmosphere. If respiration increases more than photosynthesis, the land contributes to enhanced warming as it adds $CO2$ to the atmosphere in addition to the anthropogenic $CO2$. Over the historical period land has taken up more $CO2$ than it has released and has absorbed 29% of the emitted anthropogenic $CO2$ (Jia et al., 2019). It has thus dampened the warming from anthropogenic $CO2$ release (see Friedlingstein et al., 2006).

The third and possibly the most important aspect, at least on timescales up to several decades, is the change in the meteorology and climatology that the land experiences. Changes in rainfall, temperature, temperature extremes, rainfall versus snowfall, and a host of other possible changes matter to the land surface. It is not merely the amount of rainfall that affects the land, but how the amount of rain is distributed in time, which affects runoff generation. Similarly, a small increase in the mean temperature may hide large increases in daily temperatures with impacts on ecosystems. Observations point to evolving extremes, and a major assessment was undertaken by the Intergovernmental Panel on Climate Change (Field et al., 2012). A briefer contribution from Coumou and Rahmstorf (2012) provides a useful entry point.

Thus, following from the three mechanisms described, climate change has an impact on the functioning of land ecosystems. For example, it affects the timing and duration of the growing season and the extent of forests in northern regions and at higher altitude than today. These changes modify the water, energy, and carbon cycles and ultimately the atmosphere. Land therefore has the potential to dampen or amplify greenhouse gas–induced global climate warming. In boreal regions, the projected combined northward migration of the treeline and increased growing season length in response to increased temperatures and $CO2$ in those regions will act as a positive feedback on global and regional annual warming (Garnaud & Sushama, 2015; Jeong et al., 2014; O’ishi & Abe-Ouchi, 2009; Port et al., 2012; Strengers et al., 2010). The warming that results from the decreased surface albedo is the dominant signal during the snow season, while cooling occurs during the growing season in response to enhanced evapotranspiration (Figure 6, right panel).

In the tropics, climate change will cause both greening and browning. Where global warming causes a decrease in rainfall, the induced decrease in biomass production leads to decreased evapotranspiration and increased local warming (Port et al., 2012; Wu et al., 2016; Yu et al., 2016). The reverse is true where warming generates increases in rainfall and thus greening. As an example, Port et al. (2012) simulated decreases in tree cover and a shortened growing season in the Amazon, despite the $CO2$ fertilization effects, in response to both future tropical warming and reduced precipitation (Figure 6, left panel). This browning of the land decreases both evapotranspiration and atmospheric humidity. The warming driven by lower evapotranspiration is enhanced by decreased cloudiness, which increases incoming solar radiation and is dampened by reduced water vapor leading to lower incoming infrared radiation.

### Land Management as an Additional Driver of Climate

Humans affect ~70% of the global, ice-free land surface (Arneth et al. 2019). Direct effects of human activity on the land include (a) shifting land cover (e.g., deforestation, conversion of prairies to crops, urbanization) and (b) land management within the same land cover (e.g., irrigation, ploughing, intercropping, selection of crops to improve water use efficiency). The effects of shifting land cover have been previously addressed through the case study of deforestation. Urbanization is a very specific land cover change. It limits $λE$, and $H$ is larger in urban areas than in vegetated landscapes. This leads to increases in surface air temperature over cities and in their surroundings. This phenomenon is referred to as the urban heat island effect (Bader et al., 2018; Li et al., 2018; Torres-Valcárcel et al., 2015). Nighttime temperatures are affected more than daytime (Alghamdi & Moore, 2015; Alizadeh-Choobari et al., 2016) because cities release large amounts of stored energy at night. This “anthropogenic heat” needs to be added to the left-hand side of Equation 3. Precipitation can also be impacted by urbanization. Mean and extremes increase over and downwind of some cities, especially in the afternoon and early evening (McLeod et al., 2017).

Land management affects many characteristics of the land and thus the various fluxes in the energy, water, and carbon equations. The structure of the soil for example can be modified through ploughing after harvest. This decreases surface albedo and increases soil evaporation (Davin et al., 2014). It also increases soil respiration by accelerating the decomposition of soil organic matter. Surface albedo is affected by the choice of crop variety and its seasonality: darker surfaces for example will be obtained if crop management includes intercropping as soils will always be covered with vegetation. Irrigation increases soil evaporation, which cools daytime temperatures within the irrigated area and during irrigation (Bonfils & Lobell, 2007; Chen & Jeong, 2018). Irrigation also increases the total amount of water vapor in the atmosphere and thus increases the downwelling long-wave radiation (Boucher et al., 2004). There is evidence from modeling studies that implementing irrigation enhances rainfall, although there is no agreement on where this increase occurs. For example, irrigation in India occurs prior to the start of the monsoon season and the resulting land cooling decreases the land-sea temperature contrast. This can delay the onset of the Indian monsoon and decrease its intensity (Guimberteau et al., 2012; Niyogi et al., 2010).

There are scale implications of human management relevant to climate modeling. An atmospheric model using spatial resolution of 1 km—something that is done for specific regions or cities—is a very different proposition to a global model using 100 x 100 km. At a 1 km resolution, orientation of hill slopes, lateral flow and redistribution of water in the soil, urban canyons, the heterogeneity in vegetation and soils, catchment geometry, how a crop is managed, and so on can all have a major influence on the surface energy and water balance. The land models in climate models do not typically simulate these processes. It is not known at what spatial resolution these processes begin to have a significant impact on the surface energy and water balance, but users of land models do need to think about the scale dependencies of the processes they work to represent.

### Emerging Challenges for Land Surface Science

Even if some of the mechanisms through which the land surface influences climate are reasonably well understood, the quantification of by how much and where land changes influence local, regional, and global climate remains uncertain, especially if we are to understand or predict seasonal changes, including weather extremes. Emerging challenges can differ depending on the target. In this discussion we envisage two of those targets: (a) realistic global climate projections and (b) reliable regional climate projections that will be used for impact studies.

#### Challenges for Trustable Projections at the Global Scale

The magnitude of warming of the Earth for a particular increase in atmospheric $CO2$ is influenced by the land via the snow-albedo feedback, soil and vegetation carbon cycle feedbacks (Friedlingstein et al., 2003), changes in the distribution of vegetation, stability of permafrost, and so on. These can affect the hemispheric energy balance, storm tracks, and monsoon systems in the case of the snow-albedo feedback, or the growth of greenhouse gases in the atmosphere in the case of the other feedbacks. Global climate models were originally designed to simulate very large spatial scales (spatial resolution of 100 km), and the land surface component of these models is probably reasonable. The assessments of climate change at large spatial scales and over timescales of a few decades are unlikely to be significantly affected by the remaining uncertainties in how we represent the land. Longer-term projections, however, depend crucially on our ability to simulate how the dynamics of soil carbon (e.g., through changes in microbial activity) will be affected by ongoing climate change and land uses, which remains an area of uncertainty (Jia et al., 2019). It also depends on whether, in areas of melting permafrost, the release of $CO2$ and $CH4$ will be partially compensated by growing vegetation, which is also an area of uncertainty (Jia et al., 2019).

In addition, climate and land surface models often lack a full coupling between the carbon cycle and the physical climate system. In the real world, the exchanges of $CO2$ between the land and ocean surfaces and the atmosphere are constantly updating the atmospheric $CO2$ concentration. However, this update is not always included in global climate models, which prevents the calculation of the net effects of any change in the land on climate. Net effects are still often estimated from models that calculate the biophysical and the biogeochemical effects separately.

#### Challenges for Projections of Climate Change at the Regional Scale

Climate projections are, of course, particularly valuable where they are reliable at the spatial scale at which humans grow crops, source water, and live and in capturing extremes including heatwaves and drought. Errors in climate projections at a regional scale can lead to major uncertainties in how crop production, water resources, or human settlements will be affected. This raises multiple challenges, some of which are linked to land processes.

First of all, if we are to improve our land models and properly capture the energy, water, and carbon balance, and their influence on the atmosphere, better observations are required. Organizations like FLUXNET have provided data for many surface types, including $Rnet$, $λE$, $H$, and carbon fluxes, and sometimes soil moisture and temperature.

These data are invaluable to improving our understanding and translating that understanding into modeling (e.g., Best et al., 2015). The effort to take good quality observations, quality control them, and provide them to the community is enormous, and it is important that the modeling community supports the maintenance of this effort. Many parts of the world remain uncovered by such observations, which affects our ability to properly simulate all the world’s ecosystems. There is also a need to strengthen our capacity to measure how changes in land cover or management and specific changes in climatic conditions (e.g., more frequent or prolonged weather extremes) influence land-atmosphere interactions.

Second, at scales of regions and over timescales of days, seasons, and perhaps decades, the land is highly influential and there are many areas in which we need to improve our understanding and modeling capacity. The following is a nonexhaustive list of focus areas. It is important to note that there are elements in this list that are not important in some regions but are necessary in others; the land’s key role in this context is highly region-specific.

There is a difference between what we understand and how well we can model it in a land surface model. Accurately reflecting the basic physics of the surface energy and water balance remains an ongoing challenge. There are opportunities to improve every aspect of land surface biophysics and biogeochemistry to better reflect observations and our growing understanding.

Although the mechanisms through which ecosystems partition incoming energy are reasonably well understood, land models still have problems with accurately partitioning turbulent heat fluxes between latent and sensible heat fluxes and how this partitioning affects the overlying atmosphere.

An area of focus in the ecology and ecophysiology communities is broadly defined as vegetation-water interactions. How do different vegetation types respond to drought and soil moisture deficits? What strategies do plants use to optimize carbon gain and minimize water loss? How does this influence the portioning of $Rnet$ between $H$ and $λE$? And, more broadly, if plants can take up more carbon under higher atmospheric $CO2$, what does the plant do with the carbon? A plant that uses more $CO2$ to grow leaves may be at higher risk of drought while a plant that allocates the $CO2$ to roots might be less vulnerable. A great deal of work understanding these areas has been done, but it has not been implemented into the land models in part due to information needs that are lacking at the global scale.

Groundwater is an important element of the global water cycle and is rarely included in land models in part because it is very difficult to do this well at the global scale. It seems logical that a region with groundwater reserves and with vegetation that can tap these reserves would behave differently to a warming climate than one without groundwater. Further, under climate change groundwater stores will change, and this is potentially catastrophic in some regions where humans depend on it for agriculture and fresh water. Integrating groundwater models into climate models is therefore an important priority.

There are several processes that have strong regional and possibly larger-scale roles. These include fire, soil nutrients (particularly nitrogen and phosphorous), biological volatile organic compounds, and dust from soil erosion. These areas are discussed in Massad et al. (2019) and in Jia et al. (2019), with references herein.

Urban landscapes have traditionally not been represented in land models because at the scale climate models resolve the land (100 x 100 km) they could arguably be omitted. However, the resolution of climate models is increasing and it is untenable to not represent urban areas in land surface models in the immediate future. There is a very large community of urban land modelers, a community that is quite separate from the land modelers in the climate community, but efforts to bring these together are now emerging and some land surface models account for urban areas more reasonably.

Few climate models include land management despite evidence that regional climate can be significantly influenced by human activity. This is not only a question of how to represent energy, water, and carbon processes but also a societal question of how humans decide what crops to grow, how to manage forests, when to irrigate, and when to harvest.

A new area of how the land influences the magnitude, frequency, and duration of extreme events is now emerging. In particular, capturing the timing of dry-down in models—when soil moisture strongly limits $λE$ and $H$ increases—is extremely important to how temperature increases during a heatwave. This highlights new challenges, such as gaining detailed knowledge of how the evolution of the land affects extreme events, and offers some exciting research opportunities.

The land surface model is a complex software system with details required on the energy, water, carbon, and human management aspects of the land. While early land models were built by an individual, they now represent teams of people with both scientific and software engineering expertise. Systems approaches to model development around science advances, software management, and model evaluation have developed in some groups. International collaboration in areas of model intercomparison and model evaluation has helped accelerate progress. The demands on the land surface model have, however increased. The need to simulate the diurnal, seasonal, and interannual cycles of water-energy-carbon (and other chemical constituents) for all ecosystems under all climatic conditions persists. However, the land model must simulate shifts in ecosystem distributions, capture responses to atmospheric or other perturbations, and simulate specific phenomenon including droughts and heatwaves. Fully resolving how to model the land surface in climate models to provide the right contribution from the land to the atmosphere, and the right response by the land to the atmosphere, is likely to remain a research challenge for the foreseeable future.

### Acknowledgments

We warmly thank Josefine Ghattas (IPSL, France) who carried out the simulations with the ORCHIDEE dynamic global vegetation model that allowed the construction of figures 3 and 4. This work was facilitated by a grant overseen by the French National Research Agency (ANR) as part of the “Investissements d’Avenir” Program (LabEx BASC; ANR-11-LABX-0034). AJP was supported via the ARC Centre of Excellence for Climate Extremes (CE1700100023).

### Notes

• 1. The albedo is a coefficient that measures the proportion of solar radiation reflected by a surface or an object. It is often expressed as a percentage or a relative number between 0 and 1. Clouds, snow, and ice usually have higher albedo than soil surfaces and vegetation. The Earth’s planetary albedo changes mainly through varying cloudiness, snow, ice, leaf area, and land cover changes.