Editing IPCC:AR6/WGII/Chapter-2 (section)

=== 2.4.4 Observed Changes in Ecosystem Processes and Services ===

<div id="h2-10-siblings" class="h2-siblings"></div>

<div id="2.4.4.1" class="h3-container"></div>

<span id="observed-browning-of-rivers-and-lakes"></span>
==== 2.4.4.1 Observed Browning of Rivers and Lakes ====

<div id="h3-24-siblings" class="h3-siblings"></div>

In boreal coniferous areas, there has been an increase in the transporting of terrestrial-derived dissolved organic carbon (DOC) into rivers and lakes, which has caused increased opacity and a shift toward a brown colour (browning). There was little assessment of this in AR5. This process is driven by climate change, and stems from hydrological intensification, greening of the Northern Hemisphere and degradation of carbon sinks in peatlands ''(robust evidence, high agreement'' ) ( [[#Solomon--2015|Solomon et al., 2015]] ; [[#Catalán--2016|Catalán et al., 2016]] ; [[#de%20Wit--2016|de Wit et al., 2016]] ; [[#Finstad--2016|Finstad et al., 2016]] ; [[#Creed--2018|Creed et al., 2018]] ; [[#Hayden--2019|Hayden et al., 2019]] ). These factors enhance terrestrial productivity, alter vegetation communities and affect the hydrological control of the production and transport of DOC ( [[#Weyhenmeyer--2016|Weyhenmeyer et al., 2016]] ). Non-climate-related drivers of browning are: declining atmospheric sulphur deposition, forestry practices and LULCCs (see Table SM2.1 for detail).

Browning creates a positive feedback to climate by absorbing photosynthetically active radiation, which accelerates upper water (epilimnetic) warming ( [[#Solomon--2015|Solomon et al., 2015]] ). Browning of lakes leads to shallower and more stable thermoclines, and thus overall deep water cooling ( [[#Solomon--2015|Solomon et al., 2015]] ; [[#Williamson--2015|Williamson et al., 2015]] ), and can provoke a transition of the seasonal mixing regime from a mixed lake (polymictic) to one that is seasonally stratified ( [[#Kirillin--2016|Kirillin and Shatwell, 2016]] ).

The ecological responses of browning are a concomitant effect of climate change and nutrient status. Results from long-term, large-scale lake experiments have been variable, showing both strong synergistic effects ( [[#Urrutia-Cordero--2016|Urrutia-Cordero et al., 2016]] ) and no significant effects of browning on plankton community food webs ( [[#Rasconi--2015|Rasconi et al., 2015]] ). Browning has driven a shift from auto- to heterotrophic/mixotrophic-based production ( [[#Urrutia-Cordero--2017|Urrutia-Cordero et al., 2017]] ) and supports heterotrophic metabolism of the bacterial community ( [[#Zwart--2016|Zwart et al., 2016]] ). Browning may also accelerate primary production through the input of nutrients associated with dissolved organic matter (DOM) in nutrient-poor lakes and increase cyanobacteria, which cope better with low light intensities ( [[#Huisman--2018|Huisman et al., 2018]] ) and toxin levels ( [[#Urrutia-Cordero--2016|Urrutia-Cordero et al., 2016]] ). However, the synergistic impacts of browning and climate change on aquatic communities depends on regional precipitation patterns ( [[#Weyhenmeyer--2016|Weyhenmeyer et al., 2016]] ), watershed type ( [[#de%20Wit--2016|de Wit et al., 2016]] ) and the length of the food chain ( [[#Hansson--2013|Hansson et al., 2013]] ). Quantitative attribution of browning to climate change remains difficult ( ''medium evidence'' , ''medium agreement'' ).

In summary, new studies since AR5 have explicitly estimated the effects of warming and browning on freshwaters in boreal areas, with complex positive and negative repercussions on water temperature profiles (lower vs. upper water) ( ''high confidence'' ) and primary production ( ''medium confidence'' ).

<div id="_idContainer027" class="Figure"></div>

[[File:91f7befa6456321ae09212fa9420be41 IPCC_AR6_WGII_Figure_2_005.png]]

'''Figure 2.5 | Large-scale observed changes in freshwater ecosystems attributed to climate change over more than four decades.''' For description and references, see Sections 2.3.3, 2.4.2 and 2.5.3.6.2.

<div id="2.4.4.2" class="h3-container"></div>

<span id="observed-changes-in-wildfire"></span>
==== 2.4.4.2 Observed Changes in Wildfire ====

<div id="h3-25-siblings" class="h3-siblings"></div>

<div id="2.4.4.2.1" class="h4-container"></div>

<span id="detection-and-attribution-of-observed-changes-in-wildfire"></span>
===== 2.4.4.2.1 Detection and attribution of observed changes in wildfire =====

<div id="h4-17-siblings" class="h4-siblings"></div>

Wildfire is a natural and essential component of many forest and other terrestrial ecosystems. Excessive wildfire, however, can kill people, cause respiratory disease, destroy houses, emit carbon dioxide and damage ecosystem integrity (see Sections 2.4.4.2 and 2.4.4.4). Anthropogenic climate change increases wildfire by exacerbating its three principal driving factors: heat, fuel and ignition ( [[#Moritz--2012|Moritz et al., 2012]] ; [[#Jolly--2015|Jolly et al., 2015]] ). Non-climatic factors also contribute to wildfires—in tropical areas, fires are set intentionally to clear forest for agricultural fields and livestock pastures ( [[#Bowman--2020|Bowman et al., 2020]] ). Urban areas and roads create ignition hazards. Governments in many temperate-zone countries implement policies to suppress fires, even natural ones, producing unnatural accumulations of fuel in the form of coarse woody debris and high densities of small trees ( [[#Ruffault--2015|Ruffault and Mouillot, 2015]] ; [[#Hessburg--2016|Hessburg et al., 2016]] ; [[#Andela--2017|Andela et al., 2017]] ; [[#Balch--2017|Balch et al., 2017]] ; [[#Lasslop--2017|Lasslop and Kloster, 2017]] ; [[#Aragao--2018|Aragao et al., 2018]] ; [[#Kelley--2019|Kelley et al., 2019]] ). Globally, 4.2 million km 2 of land per year burned on average from 2002 to 2016 ( [[#Giglio--2018|Giglio et al., 2018]] ), with the highest fire frequencies in the Amazon rainforest, deciduous forests and savannas in Africa and deciduous forests in northern Australia ( [[#Earl--2018|Earl and Simmonds, 2018]] ; [[#Andela--2019|Andela et al., 2019]] ).

Since the AR5 and the IPCC Special Report on Land, published research has detected increases in the area burned by wildfire, analysed relative contributions of climate and non-climate factors and attributed burned area increases above natural (recent historical) levels to anthropogenic climate change in one part of the world, western North America ( ''robust evidence'' , ''high agreement)'' ( [[#Abatzoglou--2016|Abatzoglou and Williams, 2016]] ; [[#Partain--2016|Partain et al., 2016]] ; [[#Kirchmeier-Young--2019|Kirchmeier-Young et al., 2019]] ; [[#Mansuy--2019|Mansuy et al., 2019]] ; [[#Bowman--2020|Bowman et al., 2020]] ). Across the western USA, increases in vegetation aridity due to higher temperatures from anthropogenic climate change doubled burned area from 1984 to 2015 over what would have burned due to non-climate factors including unnatural fuel accumulation from fire suppression, with the burned area attributed to climate change accounting for 49% (32–76%, 95% confidence interval) of cumulative burned area ( [[#Abatzoglou--2016|Abatzoglou and Williams, 2016]] ). Anthropogenic climate change doubled the severity of a southwest North American drought from 2000 to 2020 that has reduced soil moisture to its lowest levels since the 1500s ( [[#Williams--2020|Williams et al., 2020]] ), driving half of the increase in burned area ( [[#Abatzoglou--2016|Abatzoglou and Williams, 2016]] ; [[#Holden--2018|Holden et al., 2018]] ; [[#Williams--2019|Williams et al., 2019]] ). In British Columbia, Canada, the increased maximum temperatures due to anthropogenic climate change increased burned area in 2017 to its highest extent in the 1950–2017 record, seven to eleven times the area that would have burned without climate change ( [[#Kirchmeier-Young--2019|Kirchmeier-Young et al., 2019]] ). In Alaska, USA, the high maximum temperatures and extremely low relative humidity due to anthropogenic climate change accounted for 33–60% of the probability of wildfire in 2015, when the area burned was the second highest in the 1940–2015 record ( [[#Partain--2016|Partain et al., 2016]] ). In protected areas of Canada and the USA, climate factors (temperature, precipitation, relative humidity and evapotranspiration) accounted for 60% of burned area from local human and natural ignitions from 1984 to 2014, outweighing local human factors (population density, roads and built area) ( [[#Mansuy--2019|Mansuy et al., 2019]] ).

In summary, field evidence shows that anthropogenic climate change has increased the area burned by wildfire above natural levels across western North America in the period 1984–2017, at GMST increases of 0.6°C–0.9°C, increasing burned area up to 11 times in one extreme year and doubling it (over natural levels) in a 32-year period ( ''high confidence'' ).

<div id="2.4.4.2.2" class="h4-container"></div>

<span id="observed-changes-in-wildfire-globally"></span>
===== 2.4.4.2.2 Observed changes in wildfire globally =====

<div id="h4-18-siblings" class="h4-siblings"></div>

Regarding global terrestrial area as a whole, wildfire trends vary depending on the time period of analysis. From 1900 to 2000, global average fire frequency, based on field data, increased 0.4% but the change was not statistically significant ( [[#Gonzalez--2010|Gonzalez et al., 2010]] ). Fire frequency increased on one-third of global land, mainly from burning for agricultural clearing in Africa, Asia and South America, slightly less than the area of fire frequency decrease, mainly from fire suppression across Australia, North America and Russia ( [[#Gonzalez--2010|Gonzalez et al., 2010]] ). Analyses of the Global Fire Emissions Database document shows that, from 1996 to 2015, global burned area decreased at a rate of −0.7% yr -1 ( [[#Forkel--2019|Forkel et al., 2019]] ) but the change was not statistically significant ( [[#Giglio--2013|Giglio et al., 2013]] ). From 1998 to 2015, global burned area decreased at a rate of −1.4 ± 0.5% yr -1 ( [[#Andela--2017|Andela et al., 2017]] ). The area of fire increases was one-third of the area of decreases, due to reduced vegetation cover from agricultural expansion and intensification ( [[#Andela--2017|Andela et al., 2017]] ) and from increased precipitation ( [[#Forkel--2019|Forkel et al., 2019]] ). Furthermore, much of the decreasing trend derives from two years: 1998 with a high burned area and 2013 with low burned area ( [[#Forkel--2019|Forkel et al., 2019]] ). Wildfire does not show a clear long-term trend for the world as a whole because of increases and decreases in different regions ( ''medium evidence'' , ''medium agreement'' ).

Where the global average burned area has decreased in the past two decades, higher correlations of rates of change in burning to human population density, cropland area and livestock density than to precipitation indicate that agricultural expansion and intensification were the main causes ( [[#Andela--2017|Andela et al., 2017]] ). The global decrease of fire frequency from 2000 to 2010 is correlated with increasing human population density ( [[#Knorr--2014|Knorr et al., 2014]] ). The fire-reducing effect of reduced vegetation cover following expansion of agriculture and livestock herding can counteract the fire-increasing effect of the increased heat and drying associated with climate change ( [[#Lasslop--2017|Lasslop and Kloster, 2017]] ; [[#Arora--2018|Arora and Melton, 2018]] ; [[#Forkel--2019|Forkel et al., 2019]] ). The reduced burning needed after the initial clearing for agricultural expansion drives much of the decline in fires in the Tropics ( [[#Andela--2017|Andela et al., 2017]] ; [[#Earl--2018|Earl and Simmonds, 2018]] ; [[#Forkel--2019|Forkel et al., 2019]] ). The human influence on fire ignition can be seen through the decrease documented on holy days (Sundays and Fridays) and traditional religious days of rest ( [[#Earl--2015|Earl et al., 2015]] ). Overall, human land use exerts an influence on wildfire trends for global terrestrial area as a whole that can be stronger than climate change ( ''medium confidence'' ).

<div id="2.4.4.2.3" class="h4-container"></div>

<span id="observed-changes-in-wildfire-in-individual-regions-with-complex-attribution"></span>
===== 2.4.4.2.3 Observed changes in wildfire in individual regions with complex attribution =====

<div id="h4-19-siblings" class="h4-siblings"></div>

While burned area has increased in parts of Asia, Australia, Europe and South America, published research has not yet attributed the increases to anthropogenic climate change ( ''medium evidence'' , ''high agreement'' ).

In the Amazon, deforestation for agricultural expansion and the degradation of forests adjacent to deforested areas cause wildfire in moist humid tropical forests not adapted to fire ( ''robust evidence'' , ''high agreement'' ) ( [[#Fonseca--2017|Fonseca et al., 2017]] ; [[#van%20Marle--2017|van Marle et al., 2017]] ; [[#da%20Silva--2018|da]] [[#Silva--2018|Silva et al., 2018]] ; [[#da%20Silva--2021|da Silva et al., 2021]] ; [[#dos%20Reis--2021|dos Reis et al., 2021]] ; [[#Libonati--2021|Libonati et al., 2021]] ). Roads facilitate deforestation, fragmenting the rainforest and increasing the dryness and flammability of vegetation ( [[#Alencar--2015|Alencar et al., 2015]] ). Extreme droughts that occur during warm phases of the ENSO and the Atlantic Multi-Decadal Oscillation combine with the degradation of vegetation to cause extreme fire events ( ''robust evidence'' , ''high agreement'' ) ( [[#Fonseca--2017|Fonseca et al., 2017]] ; [[#Aragao--2018|Aragao et al., 2018]] ; [[#da%20Silva--2018|da]] [[#Silva--2018|Silva et al., 2018]] ; [[#Burton--2020|Burton et al., 2020]] ; [[#dos%20Reis--2021|dos Reis et al., 2021]] ; [[#Libonati--2021|Libonati et al., 2021]] ). In the State of Roraima, Brazil, distance to roads and infrastructure that enable deforestation and ENSO were the factors most explaining fire occurrence in the extreme 2015–2016 fire season ( [[#Fonseca--2017|Fonseca et al., 2017]] ). From 1973 to 2014, burned area increased in the Amazon, coinciding with increased deforestation ( [[#van%20Marle--2017|van Marle et al., 2017]] ). In the State of Acre, Brazil, burned area increased 36-fold from 1984 to 2016, with 43% burned in agricultural and livestock settlement areas ( [[#da%20Silva--2018|da]] [[#Silva--2018|Silva et al., 2018]] ). In the extreme fire year 2019, 85% of the area burned in the Amazon occurred in areas deforested in 2018 ( [[#Cardil--2020|Cardil et al., 2020]] ). Even though relatively higher moisture in 2019 led to burning below the 2002–2019 average across most of South America, burning in areas of recent deforestation in the Amazon were above the 2002–2019 average, indicating that deforestation, not meteorological conditions, triggered the 2019 fires ( [[#Kelley--2021|Kelley et al., 2021]] ; [[#Libonati--2021|Libonati et al., 2021]] ). Furthermore, from 1981 to 2018, deforestation in the Amazon reduced moisture inputs to the lower atmosphere, increasing drought and fire in a self-reinforcing feedback ( [[#Xu--2020|Xu et al., 2020]] ). In the Amazon, deforestation exerts an influence on wildfire that can be stronger than climate change ( ''robust evidence'' , ''high agreement'' ).

In Australia, burned area increased significantly between the periods 1950–2002 and 2003–2020 in the southeast state of Victoria, with the area burned in the 2019–2020 bushfires being the highest on record ( [[#Lindenmayer--2020|Lindenmayer and Taylor, 2020]] ). In addition to the deaths of dozens of people and the destruction of thousands of houses, the 2019–2020 bushfires burned almost half of the area protected for conservation in Victoria, two-thirds of the forests allocated for timber harvesting ( [[#Lindenmayer--2020|Lindenmayer and Taylor, 2020]] ), wildlife and extensive areas of habitat for threatened plant and animal species ( [[#Geary--2021|Geary et al., 2021]] ). Generally, past timber harvesting did not lead to more severe fire canopy damage ( [[#Bowman--2021b|Bowman et al., 2021b]] ). Across southeastern Australia, the fraction of vegetated area that burned increased significantly in eight of the 32 bioregions from 1975 to 2009, but decreased significantly in three bioregions ( [[#Bradstock--2014|Bradstock et al., 2014]] ). Increases in four bioregions were correlated to increasing temperature and decreasing precipitation. Decreases in burned area occurred despite increased temperature and decreased precipitation. Analyses of climate across Australia from 1950 to 2017 ( [[#Dowdy--2018|Dowdy, 2018]] ; [[#Harris--2019|Harris and Lucas, 2019]] ) and during periods with extensive fires in 2017 in eastern Australia ( [[#Hope--2019|Hope et al., 2019]] ), in 2018 in northeastern Australia ( [[#Lewis--2020|Lewis et al., 2020]] ), and in period 2019–2020 in southeastern Australia ( [[#Abram--2021|Abram et al., 2021]] ; [[#van%20Oldenborgh--2021|van Oldenborgh et al., 2021]] ) indicate that temperature and drought extremes due to the ENSO, Southern Annular Mode and other natural inter-decadal cycles drive inter-annual variability of fire weather. While the effects of inter-decadal climate cycles on fire are superimposed on long-term climate change, the relative importance of anthropogenic climate change in explaining changes in burned area in Australia remains unquantified ( ''medium evidence'' , ''high agreement'' ).

In Africa, the rate of change of burned area on the continent as a whole ranged from a non-statistically significant −0.45% yr -1 in the period 2002–2016 ( [[#Zubkova--2019|Zubkova et al., 2019]] ) to a significant −1.9% yr -1 in the period 2001–2016 ( [[#Wei--2020|Wei et al., 2020]] ). These decreases coincided with areas of agricultural expansion or areas where drought reduced fuel loads ( [[#Zubkova--2019|Zubkova et al., 2019]] ; [[#Wei--2020|Wei et al., 2020]] ). It is possible, however, that the 500-m spatial resolution of Modis remote-sensing fire data underestimates the area burned in Africa by half, by missing small fires ( [[#Ramo--2021|Ramo et al., 2021]] ). In the Serengeti-Mara savanna of east Africa, burned area showed no significant change from 2001 to 2014, although an increase in domestic livestock would tend to reduce the grass cover that fuels savanna fires ( [[#Probert--2019|Probert et al., 2019]] ).

In Mediterranean Europe, the area burned in the region as a whole decreased from 1985 to 2011 ( [[#Turco--2016|Turco et al., 2016]] ), although the burned area for Spain did not show a significant long-term increase from 1968 to 2010 ( [[#Moreno--2014|Moreno et al., 2014]] ) whereas that for Portugal in 2017 was the highest in the period 1980–2017 ( [[#Turco--2019|Turco et al., 2019]] ). Increased summer maximum temperature and decreased soil moisture explained most of the burned area observed, suggesting a contribution of climate change, but fire suppression, fire prevention, agricultural abandonment and reforestation as well as the reduction in forest area exerted even stronger influences on burned area than the climate across Mediterranean Europe ( ''robust evidence'' , ''high agreement'' ) ( [[#Moreno--2014|Moreno et al., 2014]] ; [[#Turco--2017|Turco et al., 2017]] ; [[#Viedma--2018|Viedma et al., 2018]] ; [[#Turco--2019|Turco et al., 2019]] ).

In the Arctic tundra and boreal forest, where wildfire has naturally been infrequent, burned area showed statistically significant increases of ~50% yr -1 across Siberia, Russia, from 1996 to 2015 ( [[#Ponomarev--2016|Ponomarev et al., 2016]] ) and 2% yr -1 across Canada from 1959 to 2015 ( [[#Hanes--2019|Hanes et al., 2019]] ). Wildfire burned ~6% of the area of four extensive Arctic permafrost regions in Alaska, USA, eastern Canada and Siberia from 1999 to 2014 ( [[#Nitze--2018|Nitze et al., 2018]] ). In boreal forest in the Northwest Territories, Canada and Alaska, USA, the area burned by wildfire increased at a statistically significant rate of 6.8% yr -1 in the period 1975–2015, ( [[#Veraverbeke--2017|Veraverbeke et al., 2017]] ), with smouldering below-ground fires that lasted through the winter covering ~1% of burned area in the period 2002–2016 ( [[#Scholten--2021|Scholten et al., 2021]] ). While burned area was correlated with temperature and reduced precipitation in Siberia ( [[#Ponomarev--2016|Ponomarev et al., 2016]] ; [[#Masrur--2018|Masrur et al., 2018]] ) and correlated with lightning, temperature and precipitation in the Northwest Territories and Alaska ( [[#Veraverbeke--2017|Veraverbeke et al., 2017]] ), no attribution analyses have examined relative influences of climate and non-climate factors.

In Indonesia, deforestation and draining of peat swamp forests dries out the peat, providing substantial fuel for fires ( [[#Page--2016|Page and Hooijer, 2016]] ). Extreme fire years in Indonesia, including 1997, 2006 and 2015, coincided with extreme heat and aridity during the warm phase of the ENSO ( [[#Field--2016|Field et al., 2016]] ). Fire-resistant forest in 2019 covered only 3% of peatlands and 4.5% of non-peatlands on Sumatra and Kalimantan ( [[#Nikonovas--2020|Nikonovas et al., 2020]] ).

In Chile, the area burned in the summer of 2016–2017 was 14 times the mean for the period 1985–2016 and the highest on record ( [[#Bowman--2019|Bowman et al., 2019]] ). While this extreme fire year coincided with the highest daily mean maximum temperature in the period 1979–2017 ( [[#Bowman--2019|Bowman et al., 2019]] ) in central Chile (the area of highest fire activity), burned area from 1976 to 2013 showed the highest correlation with the precipitation cycles of the ENSO and the temperature cycles of the Antarctic Oscillation ( [[#Urrutia-Jalabert--2018|Urrutia-Jalabert et al., 2018]] ).

Overall, burned area has increased in the Amazon, Arctic, Australia and parts of Africa and Asia, consistent with, but not formally attributed to anthropogenic climate change ( ''medium evidence'' , ''high agreement'' ). Deforestation, peat draining, agricultural expansion or abandonment, fire suppression and inter-decadal cycles such as the ENSO exert a stronger influence than climate change on wildfire trends in numerous regions outside of North America ( ''high confidence'' ).

<div id="2.4.4.2.4" class="h4-container"></div>

<span id="observed-changes-in-fire-seasons-globally"></span>
===== 2.4.4.2.4 Observed changes in fire seasons globally =====

<div id="h4-20-siblings" class="h4-siblings"></div>

The IPCC AR6 WGI assessed fire weather ( [[#Ranasinghe--2021|Ranasinghe et al., 2021]] ), while this chapter assesses the impacts of changes in fire weather: burned area and fire frequency. The global increases in temperature from anthropogenic climate change have increased aridity and drought, lengthening the fire weather season (the annual period with a heat and aridity index greater than half of its annual range) on one-quarter of global vegetated area and increasing the average fire season length by one-fifth from 1979 to 2013 ( [[#Jolly--2015|Jolly et al., 2015]] ). Climate change has contributed to increases in the fire weather season or the probability of fire weather conditions in the Amazon ( [[#Jolly--2015|Jolly et al., 2015]] ), Australia ( [[#Dowdy--2018|Dowdy, 2018]] ; [[#Abram--2021|Abram et al., 2021]] ; [[#van%20Oldenborgh--2021|van Oldenborgh et al., 2021]] ), Canada ( [[#Hanes--2019|Hanes et al., 2019]] ), central Asia ( [[#Jolly--2015|Jolly et al., 2015]] ), East Africa ( [[#Jolly--2015|Jolly et al., 2015]] ) and North America ( [[#Jain--2017|Jain et al., 2017]] ; [[#Williams--2019|Williams et al., 2019]] ; [[#Goss--2020|Goss et al., 2020]] ). In forest areas, the burned area correlates with fuel aridity, a function of temperature; in non-forest areas, the burned area correlates with high precipitation in the previous year, which can produce high grass fuel loads ( [[#Abatzoglou--2018|Abatzoglou et al., 2018]] ). Fire use in agriculture and raising livestock or other factors have generated a second fire season on approximately one-quarter of global land where fire is present, despite sub-optimal fire weather in the second fire season ( [[#Benali--2017|Benali et al., 2017]] ). In summary, anthropogenic climate change, through a 0.9°C surface temperature increase since the pre-industrial period, has lengthened or increased the frequency of periods with heat and aridity that favour wildfire on up to one-quarter of vegetated area since 1979 ( ''robust evidence, high agreement'' ).

<div id="2.4.4.2.5" class="h4-container"></div>

<span id="observed-changes-in-post-fire-vegetation"></span>
===== 2.4.4.2.5 Observed changes in post-fire vegetation =====

<div id="h4-21-siblings" class="h4-siblings"></div>

Globally, fire has contributed to biome shifts ( [[#2.4.3.2|Section 2.4.3.2]] ) and tree mortality (Sections 2.4.4.2, 2.4.4.3) attributed to anthropogenic climate change. Research since the AR5 has also found vegetation changes from wildfire due to climate change. Through increased temperature and aridity, anthropogenic climate change has driven post-fire changes in plant regeneration and species composition in South Africa ( [[#Slingsby--2017|Slingsby et al., 2017]] ), and tree regeneration in the western USA ( [[#Davis--2019b|Davis et al., 2019b]] ). In the fynbos vegetation of the Cape Floristic Region, South Africa, post-fire heat and drought and the legacy effects of exotic plant species reduced the regeneration of native plant species, decreasing species richness by 12% from 1966 to 2010 and shifting the average temperature tolerance of species communities upward by 0.5°C ( [[#Slingsby--2017|Slingsby et al., 2017]] ). In burned areas across the western USA, the increasing heat and aridity of anthropogenic climate change from 1979 to 2015 pushed low-elevation ponderosa pine ( ''Pinus ponderosa'' ) and Douglas fir ( ''Pseudotsuga menziesii'' ) forests across critical thresholds of heat and aridity that reduced the post-fire tree regeneration by half ( [[#Davis--2019b|Davis et al., 2019b]] ). In the southwestern USA, where anthropogenic climate change has caused drought ( [[#Williams--2019|Williams et al., 2019]] ) and increased wildfire ( [[#Abatzoglou--2016|Abatzoglou and Williams, 2016]] ), high-severity fires have converted some forest patches to shrublands ( [[#Barton--2018|Barton and Poulos, 2018]] ). Field evidence shows that anthropogenic climate change and wildfire, together, altered vegetation species composition in the southwestern USA and Cape floristic region, South Africa, reducing post-fire natural regeneration and species richness of tree and other plant species, between 1966 and 2015, at GMST increases of 0.3°C–0.9°C ( ''medium evidence'' , ''high agreement'' ).

<div id="2.4.4.3" class="h3-container"></div>

<span id="observed-changes-in-tree-mortality"></span>
==== 2.4.4.3 Observed Changes in Tree Mortality ====

<div id="h3-26-siblings" class="h3-siblings"></div>

<div id="2.4.4.3.1" class="h4-container"></div>

<span id="observed-tree-mortality-globally"></span>
===== 2.4.4.3.1 Observed tree mortality globally =====

<div id="h4-22-siblings" class="h4-siblings"></div>

Anthropogenic climate change can cause tree mortality directly via increased aridity or drought ( [[#2.4.4.3.3|Section 2.4.4.3.3]] ) or indirectly through wildfire ( [[#2.4.4.2.1|Section 2.4.4.2.1]] ) and insect pests ( [[#2.4.4.3.3|Section 2.4.4.3.3]] ). Catastrophic failure of the plant hydraulic system, in which a lack of water causes the xylem to lose hydraulic conductance, is the principal mechanism of drought-induced tree death ( [[#Anderegg--2016|Anderegg et al., 2016]] ; [[#Adams--2017|Adams et al., 2017]] ; [[#Anderegg--2018|Anderegg et al., 2018]] ; [[#Choat--2018|Choat et al., 2018]] ; [[#Menezes-Silva--2019|Menezes-Silva et al., 2019]] ; [[#Brodribb--2020|Brodribb et al., 2020]] ).

Up through the AR5 ( [[#Settele--2014|Settele et al., 2014]] ), detection and attribution analyses had found that anthropogenic climate change, with global temperature increases of 0.3°C–0.9°C above the pre-industrial period and the increases in aridity exceeding the effects of local non-climate change factors, caused three cases of drought-induced tree mortality of up to 20% in the period 1945–2007 in western North America ( [[#van%20Mantgem--2009|van Mantgem et al., 2009]] ), the African Sahel ( [[#Gonzalez--2012|Gonzalez et al., 2012]] ) and North Africa ( [[#le%20Polain%20de%20Waroux--2012|le Polain de Waroux and Lambin, 2012]] ). Increased wildfire and pest infestations, driven by climate change, also contributed to North American tree mortality ( [[#van%20Mantgem--2009|van Mantgem et al., 2009]] ). In addition, a meta-analysis of published cases found that drought consistent with, but not formally attributed to, climate change had caused tree mortality at 88 sites in boreal, temperate and tropical ecosystems ( [[#Allen--2010|Allen et al., 2010]] ), with 49 additional cases found by the AR5 ( [[#Settele--2014|Settele et al., 2014]] ).

Since the AR5 ( [[#Settele--2014|Settele et al., 2014]] ), global meta-analyses found at least 15 ( [[#Allen--2015|Allen et al., 2015]] ) and 25 ( [[#Hartmann--2018|Hartmann et al., 2018]] ) additional sites, respectively, of drought-induced tree mortality around the world. These and other global analyses found more rapid mortality than previously ( [[#Allen--2015|Allen et al., 2015]] ), rising background mortality ( [[#Allen--2015|Allen et al., 2015]] ), mortality increasing with drought severity ( [[#Greenwood--2017|Greenwood et al., 2017]] ), mortality of tropical trees increasing with temperature ( [[#Locosselli--2020|Locosselli et al., 2020]] ), mortality increasing with tree size for many species ( [[#Bennett--2015|Bennett et al., 2015]] ), mortality predominantly at the dry edge of species ranges ( [[#Anderegg--2019|Anderegg et al., 2019]] ) and three-quarters of drought-induced mortality cases leading to a change in the dominant species ( [[#Batllori--2020|Batllori et al., 2020]] ). Multiple non-climate factors contribute to tree mortality, including timber cutting, livestock grazing and air pollution ( [[#Martinez-Vilalta--2016|Martinez-Vilalta and Lloret, 2016]] ). Globally, tropical dry forests lost, from all causes, 95,000 km 2 , 8% of their total area, from 1982 to 2016, the most extensive area of mortality of any biome ( [[#Song--2018|Song et al., 2018]] ).

In summary, anthropogenic climate change caused drought-induced tree mortality of up to 20% in the period 1945–2007 in western North America, the African Sahel and North Africa, via global temperature increases of 0.3°C–0.9°C above the pre-industrial period and increases in aridity, and it contributed to over 100 other cases of drought-induced tree mortality in Africa, Asia, Australia, Europe and North and South America ( ''high confidence'' ). Field observations document accelerating mortality rates, rising background mortality and post-mortality vegetation shifts ( ''high confidence'' ). Water stress, leading to plant hydraulic failure, is the principal mechanism of drought-induced tree mortality. Timber cutting, agricultural expansion, air pollution and other non-climate factors also contribute to tree death.

<div id="2.4.4.3.2" class="h4-container"></div>

<span id="observed-tree-mortality-in-tropical-ecosystems"></span>
===== 2.4.4.3.2 Observed tree mortality in tropical ecosystems =====

<div id="h4-23-siblings" class="h4-siblings"></div>

In the Brazilian Amazon, deforestation to clear agricultural land comprises the principal cause of tree mortality, reducing forest cover by an average of 13,900 km 2 yr -1 from 1988 to 2020 ( [[#Assis--2019|Assis et al., 2019]] ). In addition, in a set of 310 Amazon field plots, an annual average temperature increase of 1.2°C from 1950 to 2018 ( [[#Marengo--2018|Marengo et al., 2018]] ) contributed to tree mortality of ~40% from 1983 to 2011 ( [[#Brienen--2015|Brienen et al., 2015]] ). In another set of plots, mortality among newly recruited trees of mesic genera increased and drought-tolerant genera became more abundant from 1985 to 2015 ( [[#Esquivel-Muelbert--2019|Esquivel-Muelbert et al., 2019]] ). In other plots, tree mortality did not show a statistically significant change from 1965 to 2016, but rose abruptly in severe drought years, mainly during warm phases of the ENSO ( [[#Aleixo--2019|Aleixo et al., 2019]] ). Nearly half the area of the Amazon has experienced extremely dry conditions during ENSO warm phases; this can cause extensive wildfire ( [[#2.4.4.2.3|Section 2.4.4.2.3]] ). Wildfires can increase tree mortality rates by &gt;600% above rates in non-burned areas, with the higher mortality persisting for up to a decade after a fire ( [[#Silva--2018|Silva et al., 2018]] ; [[#Berenguer--2021|Berenguer et al., 2021]] ). Climate change has contributed to tree mortality in the Amazon rainforest ( ''medium evidence'' , ''medium agreement'' ).

In the African Sahel, field research has continued to detect tree mortality, ranging from 20 to 90% in the period 1965–2018 ( [[#Kusserow--2017|Kusserow, 2017]] ; [[#Trichon--2018|Trichon et al., 2018]] ; [[#Dendoncker--2020|Dendoncker et al., 2020]] ), and declines in tree biodiversity, with up to 80% local losses of tree species in the period 1970–2014 ( [[#Hanke--2016|Hanke et al., 2016]] ; [[#Kusserow--2017|Kusserow, 2017]] ; [[#Ibrahim--2018|Ibrahim et al., 2018]] ; [[#Dendoncker--2020|Dendoncker et al., 2020]] ), consistent with, but not formally attributed to, climate change. In Algeria, mortality of the Atlas cedar ( ''Cedrus atlantica'' ) increased from 1980 to 2006, coinciding with a ~1°C spring temperature increase, but non-climate factors were not examined ( [[#Navarro-Cerrillo--2019|Navarro-Cerrillo et al., 2019]] ). Across southern Africa, nine of the 13 oldest known (1100–2500 years old) baobab trees ( ''Adansonia digitata'' ) have died since 2005, although the causes are unknown ( [[#Patrut--2018|Patrut et al., 2018]] ). In South Africa, savanna trees experienced an order of magnitude increase in mortality, related, but not formally attributed to, decreased rainfall ( [[#Case--2019|Case et al., 2019]] ). In Tunisia, insect infestations related, but not formally attributed to, hotter temperatures led to mortality of cork oaks ( ''Quercus suber'' ) ( [[#Bellahirech--2019|Bellahirech et al., 2019]] ).

<div id="2.4.4.3.3" class="h4-container"></div>

<span id="observed-tree-mortality-in-boreal-and-temperate-ecosystems"></span>
===== 2.4.4.3.3 Observed tree mortality in boreal and temperate ecosystems =====

<div id="h4-24-siblings" class="h4-siblings"></div>

The most extensive research into tree mortality since the AR5 has been in the western USA, where anthropogenic climate change accounted for half the magnitude of a drought in the period 2000–2020 that has been the most severe since the 1500s, ( [[#Williams--2020|Williams et al., 2020]] ) and for one-tenth to one-quarter of the magnitude of the 2012–2014 period of th e severe drought in California that lasted from 2012 to 2016 ( [[#Williams--2015a|Williams et al., 2015a]] ). Across the western USA, anthropogenic climate change doubled tree mortality between 1955 and 2007 ( [[#van%20Mantgem--2009|van Mantgem et al., 2009]] ). Lodgepole pine ( ''Pinus contorta'' ) mortality increased 700% from 2000 to 2013 ( [[#Anderegg--2015|Anderegg et al., 2015]] ) and piñon pine ( ''P. edulis'' ) experienced &gt;50% mortality from 2002 to 2014 ( [[#Redmond--2018|Redmond et al., 2018]] ). In montane conifer forest in California, anthropogenic climate change has increased tree mortality by one-quarter ( [[#Goulden--2019|Goulden and Bales, 2019]] ). One-quarter of the trees died in some areas, with mortality rates of ponderosa pine ( ''P. ponderosa'' ) and sugar pine ( ''P. lambertiana'' ) increasing to up to 700% of pre-drought rates ( [[#Stephenson--2019|Stephenson et al., 2019]] ; [[#Stovall--2019|Stovall et al., 2019]] ). Substantial field evidence shows that anthropogenic climate change has caused extensive tree mortality in North America ( ''robust evidence'' , ''high agreement'' ).

In western North America, increased infestations of bark beetles and other tree-feeding insects that benefit from higher winter temperatures (section 3.3.1.1 in ( [[#IPCC--2021a|IPCC, 2021a]] )) and longer growing seasons (section 2.3.4.3.1 in ( [[#IPCC--2021a|IPCC, 2021a]] )) have killed drought-stressed trees ( [[#2.4.2.1|Section 2.4.2.1]] ) ( [[#Anderegg--2015|Anderegg et al., 2015]] ; [[#Kolb--2016|Kolb et al., 2016]] ; [[#Lloret--2018|Lloret and Kitzberger, 2018]] ; [[#Redmond--2018|Redmond et al., 2018]] ; [[#Stephens--2018|Stephens et al., 2018]] ; [[#Fettig--2019|Fettig et al., 2019]] ; [[#Restaino--2019|Restaino et al., 2019]] ; [[#Stephenson--2019|Stephenson et al., 2019]] ). Increasing temperatures have allowed bark beetles to move further north and to higher elevations, survive through the winter at sites where they would previously have died and reproduce more often ( [[#Raffa--2008|Raffa et al., 2008]] ; [[#Bentz--2010|Bentz et al., 2010]] ; [[#Jewett--2011|Jewett et al., 2011]] ; [[#Macfarlane--2013|Macfarlane et al., 2013]] ; [[#Raffa--2013|Raffa et al., 2013]] ; [[#Hart--2017|Hart et al., 2017]] ; [[#Stephenson--2019|Stephenson et al., 2019]] ; [[#Teshome--2020|Teshome et al., 2020]] ; [[#Koontz--2021|Koontz et al., 2021]] ). Under warmer conditions, some insects that were previously innocuous have become important agents of tree mortality ( [[#Stephenson--2019|Stephenson et al., 2019]] ; [[#Trugman--2021|Trugman et al., 2021]] ). Field observations show mixed effects of bark beetle-induced tree mortality on subsequent fire-caused tree mortality ( [[#Andrus--2016|Andrus et al., 2016]] ; [[#Meigs--2016|Meigs et al., 2016]] ; [[#Candau--2018|Candau et al., 2018]] ; [[#Lucash--2018|Lucash et al., 2018]] ; [[#Talucci--2019|Talucci and Krawchuk, 2019]] ; [[#Wayman--2021|Wayman and Safford, 2021]] ). From 1997 to 2018, ~5% of the forest area in the western USA died from bark beetle infestations ( [[#Hicke--2020|Hicke et al., 2020]] ). Under most circumstances, trees that have been weakened by drought are more vulnerable to being killed by bark beetles ( [[#Anderegg--2015|Anderegg et al., 2015]] ; [[#Kolb--2016|Kolb et al., 2016]] ; [[#Lloret--2018|Lloret and Kitzberger, 2018]] ; [[#Redmond--2018|Redmond et al., 2018]] ; [[#Stephens--2018|Stephens et al., 2018]] ; [[#Fettig--2019|Fettig et al., 2019]] ; [[#Restaino--2019|Restaino et al., 2019]] ; [[#Stephenson--2019|Stephenson et al., 2019]] ; [[#Koontz--2021|Koontz et al., 2021]] ). In summary, climate change has contributed to bark beetle infestations that have caused much of the tree mortality in North America ( ''robust evidence'' , ''high agreement'' ) (see also [[#2.4.2.1|Section 2.4.2.1]] ).

Across Europe, rates of tree mortality in field inventories from 2000 to 2012 were highest in Spain, Bulgaria, Sweden and Finland, positively correlated to maximum winter temperature and inversely correlated to spring precipitation ( [[#Neumann--2017|Neumann et al., 2017]] ). Tree mortality in Austria, the Czech Republic, Germany, Poland, Slovakia and Switzerland doubled from 1984 to 2016, correlated with intensified logging and increased temperatures ( [[#Senf--2018|Senf et al., 2018]] ). Drought-related tree mortality rates from 1987 to 2016 were highest in the Ukraine, Moldova, southern France and Spain ( [[#Senf--2020|Senf et al., 2020]] ). Climate contributed to tree mortality across Europe from 1958 to 2001 ( [[#Seidl--2011|Seidl et al., 2011]] ). In addition, insect infestations related to higher temperatures ( [[#Okland--2019|Okland et al., 2019]] ) have caused the extensive mortality of Norway spruce ( ''Picea abies'' ) across nine European countries ( [[#Marini--2017|Marini et al., 2017]] ; [[#Mezei--2017|Mezei et al., 2017]] ). Across the Mediterranean Basin, a combination of drought, wildfire, pest infestations and livestock grazing ( [[#Peñuelas--2021|Peñuelas and Sardans, 2021]] ) has driven tree mortality. In summary, climate change has contributed to tree mortality in Europe ( ''high confidence'' ) (see also [[#2.4.2.1|Section 2.4.2.1]] ).

<div id="2.4.4.3.4" class="h4-container"></div>

<span id="tree-mortality-and-fauna"></span>
===== 2.4.4.3.4 Tree mortality and fauna =====

<div id="h4-25-siblings" class="h4-siblings"></div>

A global meta-analysis of 59 studies encompassing 631 cases of animal abundance changes in areas of tree mortality over the past 7–59 years, primarly in North America and Australia, with a few sites in other regions (e.g. Europe). Overall, in areas with documented high tree mortality, bird abundances increased (n=186 bird species), there was no significant trend for mammals (n=33 species), a slight trend towards declines in invertebrates (n=28 species), and insufficient information to categorize the responses of reptiles (n=20 species). However, within groups, significant differences appeared. Mammals that use trees as refugia showed declines with tree mortality ''(high confidence)'' , but flying mammals (e.g. bats) increased ''(medium confidence)'' . Ground-nesting, ground-foraging, tree-hole nesting and bark-foraging birds increased most, but nectar-feeding and foliage-gleaning birds declined ''(high confidence)'' . Within invertebrates, declines were strongest in ground-foraging predators and detritivores ''(medium confidence)'' ( [[#Fleming--2021|Fleming et al., 2021]] ).

<div id="2.4.4.4" class="h3-container"></div>

<span id="observed-terrestrial-ecosystem-carbon"></span>
==== 2.4.4.4 Observed Terrestrial Ecosystem Carbon ====

<div id="h3-27-siblings" class="h3-siblings"></div>

<div id="2.4.4.4.1" class="h4-container"></div>

<span id="observed-terrestrial-ecosystem-carbon-globally"></span>
===== 2.4.4.4.1 Observed terrestrial ecosystem carbon globally =====

<div id="h4-26-siblings" class="h4-siblings"></div>

Terrestrial ecosystems contain carbon stocks: 450 GtC (range 380–540 GtC) in vegetation, 1700 ± 250 GtC in soils that are not permanently frozen and 1400 ± 200 GtC in permafrost ( [[#Hugelius--2014|Hugelius et al., 2014]] ; [[#Batjes--2016|Batjes, 2016]] ; [[#Jackson--2017|Jackson et al., 2017]] ; [[#Strauss--2017|Strauss et al., 2017]] ; [[#Erb--2018a|Erb et al., 2018a]] ; [[#Xu--2021a|Xu et al., 2021a]] ). Ecosystem carbon stocks, totalling 3000–4000 GtC (from the lowest and highest estimates above), substantially exceed the ~900 GtC carbon in unextracted fossil fuels (see( [[#Canadell--2021|Canadell et al., 2021]] )).

Deforestation, draining of peatlands and the expansion of agricultural fields, livestock pastures and human settlements and other LULCCs emitted carbon at a rate of 1.6 ± 0.7 Gt yr -1 from 2010 to 2019, ( [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ), of which wildfires and peat burning emitted 0.4 ± 0.2 Gt yr -1 from 1997 to 2016 ( [[#van%20der%20Werf--2017|van der Werf et al., 2017]] ). Anthropogenic climate change has caused some of these emissions through increases in wildfire ( [[#2.4.4.2.1|Section 2.4.4.2.1]] ) and tree mortality ( [[#2.4.4.3.1|Section 2.4.4.3.1]] ), but the fraction of the total remains unquantified. LUC produced ~15% of global anthropogenic emissions, from fossil fuels and land ( [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ). Terrestrial ecosystems removed carbon from the atmosphere through plant growth at a rate of -3.4 ± 0.9 Gt yr -1 from 2010 to 2019 ( [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ).

Tropical deforestation and the draining and burning of peatlands produce almost all of the carbon emissions from LUC ( [[#Houghton--2017|Houghton and Nassikas, 2017]] ; [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ), while forest growth accounts for two-thirds of ecosystem carbon removals from the atmosphere ( [[#Pugh--2019b|Pugh et al., 2019b]] ). Global terrestrial ecosystems comprised a net sink of -1.9 ± 1.1 Gt yr -1 from 2010 to 2019 ( [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ), mainly due to growth in forests ( [[#Harris--2021|Harris et al., 2021]] ; [[#Xu--2021a|Xu et al., 2021a]] ), mitigating ~31% of global emissions from the burning of fossil fuels and LUC ( [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ).

In summary, terrestrial ecosystems contain 3000–4000 GtC in vegetation, permafrost and soils, three to five times the amount of carbon in unextracted fossil fuels and 4.4 times the carbon currently in the atmosphere ( ''robust evidence'' , ''high agreement'' ). Tropical deforestation, the draining and burning of peatlands and other LULCCs emit 0.9–2.3 GtC yr -1 , ~15% of the global emissions from fossil fuels and ecosystems ( ''robust evidence'' , ''high agreement'' ). Terrestrial ecosystems currently remove more carbon from the atmosphere (-3.4±0.9 Gt yr -1 ) than they emit (+1.6±0.7 Gt yr -1 ), a net sink of -1.9±1.1 Gt yr -1 ( [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ) . Thus, tropical rainforests, Arctic permafrost and other ecosystems provide the global ecosystem service of naturally preventing carbon from contributing to climate change ( ''high confidence'' ).

<div id="2.4.4.4.2" class="h4-container"></div>

<span id="observed-stocks-in-high-carbon-terrestrial-ecosystems"></span>
===== 2.4.4.4.2 Observed stocks in high-carbon terrestrial ecosystems =====

<div id="h4-27-siblings" class="h4-siblings"></div>

The ecosystem that attains the highest above-ground carbon density in the world is the coast redwood ( ''Sequoia sempervirens'' ) forest in California, USA, with 2600 ± 100 tonnes ha -1 carbon ( [[#Van%20Pelt--2016|Van Pelt et al., 2016]] ). The ecosystem with the second highest documented carbon density in the world is the mountain ash ( ''Eucalyptus regnans'' ) forest in Victoria, Australia, with ~1900 tonnes ha -1 ( [[#Keith--2009|Keith et al., 2009]] ). In the Tropics, tropical evergreen broadleaf forests (rainforests) in the Amazon, the Congo and Indonesia attain the highest carbon densities, reaching a maximum of 230 tonnes ha -1 in the Amazon ( [[#Mitchard--2014|Mitchard et al., 2014]] ) and the Congo ( [[#Xu--2017|Xu et al., 2017]] ). Temperature increases reduce the tropical rainforest above-ground carbon density 9.1 tonnes ha -1 per degree Celsius, through reduced growth and increased tree mortality ( [[#Sullivan--2020|Sullivan et al., 2020]] ).

Tropical forests contain the largest vegetation carbon stocks in the world, with 180–250 GtC above and below ground ( [[#Saatchi--2011|Saatchi et al., 2011]] ; [[#Baccini--2012|Baccini et al., 2012]] ; [[#Avitabile--2016|Avitabile et al., 2016]] ). The Amazon contains a stock of 45–60 GtC ( [[#Baccini--2012|Baccini et al., 2012]] ; [[#Mitchard--2014|Mitchard et al., 2014]] ; [[#Englund--2017|Englund et al., 2017]] ).

Ecosystems with high soil carbon densities include the peat bogs in Ireland with up to 3000 tonnes ha -1 ( [[#Tomlinson--2005|Tomlinson, 2005]] ), the Cuvette Centrale swamp forest peatlands in Congo with an average of ~2200 tonnes ha -1 ( [[#Dargie--2017|Dargie et al., 2017]] ), the Arctic tundra with an average of ~900 tonnes ha -1 ( [[#Tarnocai--2009|Tarnocai et al., 2009]] ) and the mangrove peatlands in Kalimantan, Indonesia, with an average of 850 ± 320 tonnes ha -1 ( [[#Murdiyarso--2015|Murdiyarso et al., 2015]] ). Arctic permafrost contains 1400 ± 200 GtC to a depth of 3 m, the largest soil carbon stock in the world ( [[#Hugelius--2014|Hugelius et al., 2014]] ). Globally, peatlands contain 470–620 GtC ( [[#Page--2011|Page et al., 2011]] ; [[#Hodgkins--2018|Hodgkins et al., 2018]] ), of which boreal and temperate peatlands contain 415 ± 150 GtC ( [[#Hugelius--2020|Hugelius et al., 2020]] ) and tropical peatlands contain 80–350 GtC ( [[#Page--2011|Page et al., 2011]] ; [[#Dargie--2017|Dargie et al., 2017]] ; [[#Gumbricht--2017|Gumbricht et al., 2017]] ; [[#Ribeiro--2021|Ribeiro et al., 2021]] ). Other analyses increase the upper estimates for boreal and temperate peatlands to 800–1200 GtC ( [[#Nichols--2019|Nichols and Peteet, 2019]] ; [[#Mishra--2021b|Mishra et al., 2021b]] ).

Tropical forests and Arctic permafrost contain the highest ecosystem carbon stocks in above-ground vegetation and soil, respectively, in the world ( ''robust evidence'' , ''high agreement'' ). These ecosystems form natural sinks that prevent the emission to the atmosphere of 1400–1800 GtC that would otherwise increase the magnitude of climate change ( ''high confidence'' ).

<div id="2.4.4.4.3" class="h4-container"></div>

<span id="biodiversity-and-observed-terrestrial-ecosystem-carbon"></span>
===== 2.4.4.4.3 Biodiversity and observed terrestrial ecosystem carbon =====

<div id="h4-28-siblings" class="h4-siblings"></div>

High biodiversity and ecosystem carbon generally occur together, with rainforests in the Amazon, Congo and Indonesia containing the largest above-ground vegetation carbon stocks ( [[#Saatchi--2011|Saatchi et al., 2011]] ; [[#Baccini--2012|Baccini et al., 2012]] ; [[#Avitabile--2016|Avitabile et al., 2016]] ) and the highest vascular plant species richness ( [[#Kreft--2007|Kreft and Jetz, 2007]] ) in the world. Above-ground ecosystem carbon and animal species richness show high correlation but also high spatial variability ( [[#Strassburg--2010|Strassburg et al., 2010]] ). Above-ground carbon is correlated to genus richness globally ( [[#Cavanaugh--2014|Cavanaugh et al., 2014]] ), but to species richness only in local areas ( [[#Poorter--2015|Poorter et al., 2015]] ; [[#Sullivan--2017|Sullivan et al., 2017]] ). Species richness generally increases vegetation productivity in the humid tropics while tree abundance increases productivity in drier conditions ( [[#Madrigal-Gonzalez--2020|Madrigal-Gonzalez et al., 2020]] ). Across the Amazon, ~1% of tree species contain 50% of the above-ground carbon, due to abundance and maximum height ( [[#Fauset--2015|Fauset et al., 2015]] ). Above-ground carbon in tropical forests shows positive correlations to vertebrate species richness (P values not reported) ( [[#Deere--2018|Deere et al., 2018]] ; [[#Di%20Marco--2018|Di Marco et al., 2018]] ). In logged and burned tropical forest in Brazil, species richness of plants, birds and beetles increased with carbon density up to ~100 tonnes ha -1 ( [[#Ferreira--2018|Ferreira et al., 2018]] ).

National parks and other protected areas which, in June 2021, covered 15.7% of global terrestrial area (UNEP-WCMC et al., 2021) contain ~90 GtC in vegetation and ~150 GtC in soil (one-fifth and one-tenth, respectively, of global stocks) and remove carbon from the atmosphere at a rate of ~0.5 Gt yr -1 (one-sixth of global removals) ( [[#Melillo--2016|Melillo et al., 2016]] ). The most strictly protected areas contain carbon at higher densities, but illegal deforestation and fires in some protected areas emit 38 ± 17 Mt yr -1 globally ( [[#Collins--2017|Collins and Mitchard, 2017]] ). In the Amazon, protected areas store more than half of the above-ground vegetation carbon stocks of the region, but account for only one-tenth of net emissions ( [[#Walker--2020|Walker et al., 2020]] ). Conservation of high biodiversity areas, particularly in protected areas, protects ecosystem carbon, prevents emissions to the atmosphere and reduces the magnitude of climate change ( ''high confidence'' ).

<div id="2.4.4.4.4" class="h4-container"></div>

<span id="observed-emissions-and-removals-from-high-carbon-terrestrial-ecosystems"></span>
===== 2.4.4.4.4 Observed emissions and removals from high-carbon terrestrial ecosystems =====

<div id="h4-29-siblings" class="h4-siblings"></div>

Most global deforestation is occurring in tropical forests ( [[#Pan--2011|Pan et al., 2011]] ; [[#Liu--2015|Liu et al., 2015]] ; [[#Houghton--2017|Houghton and Nassikas, 2017]] ; [[#Erb--2018a|Erb et al., 2018a]] ; [[#Li--2018|Li et al., 2018]] ; [[#Harris--2021|Harris et al., 2021]] ), primarily as a result of clearing for agricultural land ( [[#Hong--2021|Hong et al., 2021]] ), causing primary tropical forest to comprise a net source of carbon from 2001 to 2019: emissions to the atmosphere 0.6 GtC yr -1 , removals from the atmosphere -0.5 GtC yr -1 and net 0.1 GtC yr -1 ( [[#Harris--2021|Harris et al., 2021]] ). While wildfires emitted an average of 0.4 ± 0.2 GtC yr -1 from 1997 to 2016 ( [[#van%20der%20Werf--2017|van der Werf et al., 2017]] ), individual fire seasons can emit the same magnitude, such as the 0.4 GtC from the Amazon fires of 2007 ( [[#Aragao--2018|Aragao et al., 2018]] ), the 0.5 GtC from the Amazon fires of 2015–2016 ( [[#Berenguer--2021|Berenguer et al., 2021]] ) and the 0.2 Gt from the Australia fires of 2019–2020 ( [[#Shiraishi--2021|Shiraishi and Hirata, 2021]] ). Wildfires thus account for up to one-third of annual average ecosystem carbon emissions, while major fire seasons can emit up to two-thirds of global ecosystem carbon ( ''medium evidence'' , ''medium agreement'' ).

Primary boreal and temperate forests also comprised net sources in the period 2001–2019; however, when including all tree age classes, boreal, temperate and tropical forests were net sinks (boreal -1.6 ± 1.1 Gt yr -1 , temperate -3.6 ± 48 Gt yr -1 ), as growth exceeded permanent forest cover losses ( [[#Harris--2021|Harris et al., 2021]] ), with boreal and temperate forests being much stronger sinks ( [[#Pan--2011|Pan et al., 2011]] ; [[#Liu--2015|Liu et al., 2015]] ; [[#Houghton--2017|Houghton and Nassikas, 2017]] ). Estimates of carbon removals from remote sensing may provide more accurate estimates of boreal forest carbon balances than ESMs which overestimate regrowth after timber harvesting and other disturbance ( [[#Wang--2021a|Wang et al., 2021a]] ). Mortality of the boreal forest in British Columbia from mountain pine beetle infestations converted 374,000 km 2 from a net carbon sink to a net carbon source ( [[#Kurz--2008|Kurz et al., 2008]] ). Modelling suggests that a potential increase in water-use efficiency and regrowth could offset the losses in part of the forest mortality area ( [[#Giles-Hansen--2021|Giles-Hansen et al., 2021]] ).

The Amazon as a whole was a net carbon emitter in the period 2003–2008 ( [[#Exbrayat--2015|Exbrayat and Williams, 2015]] ; [[#Yang--2018b|Yang et al., 2018b]] ), primarily due to the expansion of agricultural and livestock areas, which caused over two-thirds of deforestation from 1990 to 2005 ( [[#De%20Sy--2015|De Sy et al., 2015]] ; [[#De%20Sy--2019|De Sy et al., 2019]] ). Four sites in the Amazon also showed net carbon emissions in the period 2010–2018, from deforestation and fire ( [[#Gatti--2021|Gatti et al., 2021]] ). In the Amazon, deforestation emitted 0.17 ± 0.05 GtC yr -1 from 2001 to 2015 ( [[#Silva%20Junior--2020|Silva Junior et al., 2020]] ) while fires emitted 0.12 ± 0.14 GtC yr -1 from 2003 to 2015 ( [[#Aragao--2018|Aragao et al., 2018]] ). An analysis of the Amazon carbon loss from deforestation and degradation estimated a loss of 0.5 Gt yr -1 in the period 2010 -2019, with degradation accounting for three-quarters ( [[#Qin--2021|Qin et al., 2021]] ). Intact old-growth Amazon rainforest has been a net carbon sink from 2000 to 2010 (-0.45 Gt yr -1 , min. 0.31, max. 0.57) ( [[#Hubau--2020|Hubau et al., 2020]] ) but may have become a net carbon source in 2010–2019 (0.67 Gt, for the entire period, uncertainty not reported) ( [[#Qin--2021|Qin et al., 2021]] ). These factors combined—recent impacts of climate change on undisturbed forest, coupled with deforestation and agricultural expansion, along with associated intentional burning—have caused Amazon rainforest to become an overall net carbon emitter ''(medium confidence).''

In Indonesia and Malaysia, draining and burning of peat swamp forests for oil palm plantations emitted 60–260 MtC yr -1 from 1990 to 2015, converting peatlands in that period from a carbon sink to a carbon source ( [[#Miettinen--2017|Miettinen et al., 2017]] ; [[#Wijedasa--2018|Wijedasa et al., 2018]] ; [[#Cooper--2020|Cooper et al., 2020]] ). Deforestation of mangrove forests caused 10–30% of deforestation emissions in Indonesia from 1980 to 2005 ( [[#Donato--2011|Donato et al., 2011]] ; [[#Murdiyarso--2015|Murdiyarso et al., 2015]] ), even though mangroves comprised only 3% of Indonesia primary forest area in 2000 ( [[#Margono--2014|Margono et al., 2014]] ; [[#Murdiyarso--2015|Murdiyarso et al., 2015]] ).

In North America, wildfire emitted 0.1 ± 0.02 GtC yr -1 from in the period 1990–2012, but regrowth was slightly greater, producing a net sink ( [[#Chen--2017|Chen et al., 2017]] ). In California, USA, two-thirds of the 70 MtC emitted from natural ecosystems in 2001–2010 came from the 6% of the area that burned ( [[#Gonzalez--2015|Gonzalez et al., 2015]] ). Anthropogenic climate change caused up to half of the burned area ( [[#2.4.4.2.1|Section 2.4.4.2.1]] ).

In the Arctic, anthropogenic climate change has thawed permafrost ( [[#Guo--2020|Guo et al., 2020]] ), leading to emissions of 1.7 ± 0.8 GtC yr -1 in winter in the period 2003–2017 ( [[#Natali--2019|Natali et al., 2019]] ). Wildfires in the Arctic tundra in Alaska from ~1930 to 2010 caused up to a depth of 0.5 m of permafrost thaw ( [[#Brown--2015|Brown et al., 2015]] ), exposing peatland carbon ( [[#Brown--2015|Brown et al., 2015]] ; [[#Gibson--2018|Gibson et al., 2018]] ) including soil carbon deposits up to 1600 years old (Walker et al., 2019).

Tropical deforestation, the draining and burning of peatlands and the thawing of Arctic permafrost due to climate change have caused these ecosystems to emit more carbon to the atmosphere than they naturally remove through vegetation growth ( ''high confidence'' ).

<div id="2.4.4.5" class="h3-container"></div>

<span id="observed-changes-in-primary-productivity"></span>
==== 2.4.4.5 Observed Changes in Primary Productivity ====

<div id="h3-28-siblings" class="h3-siblings"></div>

<div id="2.4.4.5.1" class="h4-container"></div>

<span id="observed-changes-in-terrestrial-primary-productivity"></span>
===== 2.4.4.5.1 Observed changes in terrestrial primary productivity =====

<div id="h4-30-siblings" class="h4-siblings"></div>

The difference between photosynthesis by plants (gross primary productivity, GPP) and plant energy use through respiration is the net growth of plants (NPP), which removes CO 2 from the atmosphere and mitigates emissions from deforestation and other LUCs ( [[#2.4.4.4|Section 2.4.4.4]] ). Global terrestrial NPP has exceeded emissions due to land use since the early 2000s, making terrestrial ecosystems a net carbon sink ( [[#Friedlingstein--2020|Friedlingstein et al., 2020]] ).

Global terrestrial NPP increased by 6% from 1982 to 1999 through increased temperature and increased solar radiation in the Amazon from decreased cloud cover ( [[#Nemani--2003|Nemani et al., 2003]] ), and then decreased 1% from 2000 to 2009, because of extensive droughts in the Southern Hemisphere ( [[#Zhao--2010|Zhao and Running, 2010]] ). From 1999 to 2015, increased aridity caused extensive declines in the NDVI globally, particularly semiarid ecosystems ( [[#Huang--2016|Huang et al., 2016]] ), indicating widespread decreases in NPP ( [[#Yuan--2019|Yuan et al., 2019]] ).

Global terrestrial GPP increased 2% from 1951 to 2010 and continued increasing at least until 2016, with increased atmospheric CO 2 showing a greater influence than natural factors ( [[#Li--2017|Li et al., 2017]] ; [[#Fernandez-Martinez--2019|Fernandez-Martinez et al., 2019]] ; [[#Liu--2019a|Liu et al., 2019a]] ; [[#Cai--2020|Cai and Prentice, 2020]] ; [[#Melnikova--2020|Melnikova and Sasai, 2020]] ). Global forest area increased 7% from 1982 to 2016, mainly from forest plantations and regrowth in boreal and temperate forests in Asia and Europe ( [[#Song--2018|Song et al., 2018]] ); regrowth in secondary forests &gt;20 years old, mainly in boreal, temperate and subtropical regions, generated a net removal of 7.7 Gt yr -1 CO 2 from the atmosphere from 2001 to 2019 ( [[#Harris--2021|Harris et al., 2021]] ). Vegetation growth that exceeds the modelled CO 2 fertilisation, gaps in field data and incomplete knowledge of plant mortality and soil carbon responses introduce uncertainties into quantifying the magnitude of CO 2 fertilisation ( [[#Walker--2021|Walker et al., 2021]] ). A combination of CO 2 fertilisation of global vegetation and secondary forest regrowth has increased global vegetation productivity ( ''medium evidence'' , ''medium agreement'' ).

The relative increase in GPP per unit of increased atmospheric CO 2 declined from 1982 to 2015, indicating a weakening of any CO 2 fertilisation effect ( [[#Wang--2020c|Wang et al., 2020c]] ). Increased growth from CO 2 fertilisation has begun to shorten the lifespan of trees due to a trade-off between growth rate and longevity, based on analyses of tree rings of 110 species around the world ( [[#Brienen--2020|Brienen et al., 2020]] ). Furthermore, water availability controls the magnitude of NPP ( [[#Beer--2010|Beer et al., 2010]] ; [[#Jung--2017|Jung et al., 2017]] ; [[#Yu--2017|Yu et al., 2017]] ), including water from precipitation ( [[#Beer--2010|Beer et al., 2010]] ), soil moisture ( [[#Stocker--2019|Stocker et al., 2019]] ), groundwater storage ( [[#Humphrey--2018|Humphrey et al., 2018]] ; [[#Madani--2020a|Madani et al., 2020a]] ) and atmospheric vapour ( [[#Novick--2016|Novick et al., 2016]] ; [[#Madani--2020b|Madani et al., 2020b]] ). Drought stress reduced NPP across tropical forests from 2000 to 2015 ( [[#Zhang--2019b|Zhang et al., 2019b]] ) and GPP in the tropics from 1982 to 2016 ( [[#Madani--2020b|Madani et al., 2020b]] ). Drought stress has also reduced GPP in some semiarid and arid lands ( [[#Huang--2016|Huang et al., 2016]] ; [[#Liu--2019a|Liu et al., 2019a]] ). In addition, nitrogen and phosphorus constrain CO 2 fertilisation ( [[#Terrer--2019|Terrer et al., 2019]] ), although phosphorus limitation of tropical tree growth is species-specific ( [[#Alvarez-Clare--2013|Alvarez-Clare et al., 2013]] ; [[#Thompson--2019|Thompson et al., 2019]] ). NPP has decreased during some time periods and in some regions where drought stress has exerted a greater influence than increased atmospheric CO 2 ( ''medium evidence'' , ''high agreement'' ).

<div id="2.4.4.5.2 " class="h4-container"></div>

<span id="observed-changes-in-freshwater-ecosystem-productivity"></span>
===== 2.4.4.5.2 Observed changes in freshwater ecosystem productivity =====

<div id="h4-31-siblings" class="h4-siblings"></div>

Temperature affects primary productivity by moderating phytoplankton growth rates, ice cover, thermal stratification and the length of growing seasons ( [[#Rühland--2015|Rühland et al., 2015]] ; [[#Richardson--2018|Richardson et al., 2018]] ). Global warming has reinforced eutrophication, especially cyanobacteria blooms ( [[#Wagner--2009|Wagner and Adrian, 2009]] ; [[#Kosten--2012|Kosten et al., 2012]] ; [[#O’Neil--2012|O’Neil et al., 2012]] ; [[#De%20Senerpont%20Domis--2013|De Senerpont Domis et al., 2013]] ; [[#Adrian--2016|Adrian et al., 2016]] ; [[#Visser--2016|Visser et al., 2016]] ; [[#Huisman--2018|Huisman et al., 2018]] ) ( ''very high confidence'' ) ''.'' Conversely '','' warming can reduce cyanobacteria in hypertrophic lakes ( [[#Richardson--2019|Richardson et al., 2019]] ). Freshwater cyanobacteria may benefit directly from elevated CO 2 concentrations ( [[#Visser--2016|Visser et al., 2016]] ; [[#Ji--2017|Ji et al., 2017]] ; [[#Huisman--2018|Huisman et al., 2018]] ; [[#Richardson--2019|Richardson et al., 2019]] ).

Macrophyte growth in freshwaters is likely to increase with rising water temperatures, atmospheric CO 2 and precipitation ( ''robust evidence'' , ''high agreement'' ) ( [[#Dhir--2015|Dhir, 2015]] ; [[#Hossain--2016|Hossain et al., 2016]] ; [[#Short--2016|Short et al., 2016]] ; [[#Reitsema--2018|Reitsema et al., 2018]] ). Nonetheless, primary productivity in rivers is variable and unpredictable ( [[#Bernhardt--2018|Bernhardt et al., 2018]] ) because seasonal variations in temperature and light are uncorrelated, frequent high-flow events reduce biomass of autotrophs and droughts can strand and desiccate autotrophs.

In large, nutrient-poor lakes, warming-induced prolonged thermal stratification can reduce primary production ( ''medium confidence'' ) ( [[#Kraemer--2017|Kraemer et al., 2017]] ). Warming may reduce phytoplankton concentrations when temperature-induced increases in consumption of phytoplankton outpace increases in phytoplankton production ( [[#De%20Senerpont%20Domis--2013|De Senerpont Domis et al., 2013]] ). These decreases in productivity may be under-recognised responses to climate change.

Summary: There is ''robust'' evidence of an increase in primary production along with warming trends. However, increases or declines of algae cannot entirely be attributed to climate change; they are lake-specific and modulated through weather conditions, lake morphology, salinity, land use and restoration and biotic interactions ( ''medium confidence'' ) ( [[#O’Beirne--2017|O’Beirne et al., 2017]] ; [[#Velthuis--2017|Velthuis et al., 2017]] ; [[#Rusak--2018|Rusak et al., 2018]] ; [[#Ho--2019|Ho et al., 2019]] ).

<div id="FAQ 2.3" class="h2-container"></div>

<span id="faq-2.3-is-climate-change-increasing-wildfire"></span>