Deforestation intensifies daily temperature variability in the northern extratropics

The study addresses how deforestation affects daily temperature variability, a largely unexplored aspect compared to well-documented effects on mean, diurnal cycle, and extremes of temperature. Deforestation alters surface albedo, evapotranspiration, and aerodynamic roughness, with cooling dominating boreal regions and warming dominating the tropics, while mid-latitude effects are uncertain. Prior work shows diurnal asymmetry (daytime warming, nighttime cooling) and impacts on extremes, yet the influence on variability—especially high-frequency day-to-day variability critical to human health, ecosystems, and economies—remains unclear. Observed historical changes in daily temperature variability have been mainly attributed to greenhouse gases, aerosols, urbanization, and internal variability, rarely to forest change. Given extensive historical deforestation and proposed large-scale afforestation as a climate solution, the authors aim to quantify the biogeophysical effects of deforestation and afforestation on daily temperature variability globally, elucidate mechanisms, and assess implications for policy.

• Biogeophysical impacts of deforestation: increased albedo induces cooling; reduced evapotranspiration, lower roughness, and rooting depth induce warming; overall cooling in boreal regions, warming in tropics, and uncertain effects in mid-latitudes.
• Deforestation yields daytime warming and nighttime cooling globally, amplifying diurnal temperature range; effects on hot extremes are mixed in literature but often indicate aggravation of heat extremes due to reduced evaporative cooling capacity of open land.
• Historical day-to-day temperature variability generally decreased in northern mid- and high-latitudes during cold seasons, with some summer increases; changes have been attributed to anthropogenic GHGs, aerosols, urbanization, and internal variability, with little focus on land cover change.
• Afforestation is proposed for climate mitigation via carbon sequestration and biogeophysical cooling, but effects on temperature variability are largely unassessed. The paper situates its contribution within these gaps, focusing on variability and mechanisms.

Multiple coordinated model experiments and observational datasets are used.

Idealized deforestation (biogeophysical effect isolation):

• Simulations: CMIP6 piControl (all forcings fixed at 1850 levels) and LUMIP deforest-globe (80-year, 20 million km² forest converted to grassland linearly over first 50 years on the top 30% tree-cover grid cells; vegetation dynamics off in deforested cells; atmospheric CO2 fixed to piControl so effects are purely biogeophysical). Analysis uses years 51–80 (equilibrated period); deforest-globe minus piControl isolates deforestation signal. Five ESMs with required daily outputs were used.

Historical deforestation effect:

• Simulations: CMIP6 historical (1850–2014, evolving natural and anthropogenic forcings) and LUMIP hist-noLu (same but land use/land cover fixed at 1850). Compare 1985–2014 means (historical minus hist-noLu) to isolate historical LULCC effects. Nine ESMs used.

Net anthropogenic forcings (context):

• DAMIP hist-nat (natural forcings only) versus historical. Compare 1985–2014 (historical minus hist-nat) to estimate net anthropogenic (GHGs, aerosols, LULCC) effect. Four ESMs with daily outputs used.

Future afforestation effect:

• ScenarioMIP ssp370 (2015–2100, SSP3-7.0 with substantial deforestation and high emissions) and LUMIP ssp370-ssp126Lu (same but land use from SSP1-2.6 with substantial afforestation/reforestation; atmospheric CO2 identical between the pair). Compare 2071–2100 (ssp370-ssp126Lu minus ssp370) to isolate biogeophysical afforestation effects. Six ESMs used.

Daily variability metrics:

• DTDT (day-to-day temperature variation): mean absolute difference of daily mean 2-m temperature between adjacent days over a period.
• SDT: standard deviation of daily mean 2-m temperature with annual cycle removed. For models providing only daily Tmax/Tmin, mean is approximated as average of Tmax and Tmin.

Attribution framework (thermodynamic energy equation):

• Decompose change in DTDT into contributions from changes in near-surface horizontal temperature advection (TADV; −V·∇T), adiabatic/vertical terms (negligible), and diabatic heating variability approximated via day-to-day variation in surface sensible heat flux (DTDSHF). TADV requires daily 10-m wind and temperature fields; DTDSHF estimated from day-to-day differences in SHF. Units of Δterm3 are qualitative for contribution sign.

Validation and observations:

• Reanalyses: ERA5 and NCEP-DOE AMIP-II used to validate modeled spatial patterns and seasonality of DTDT and SDT (1985–2014 benchmarks). Models reproduce larger variability over higher-latitude continents and winter-dominant seasonality.
• Biogeophysical mean temperature response validation: satellite-based “space-for-time” dataset (global potential forest-to-openland temperature change) used to evaluate mean temperature response to deforestation; models broadly consistent, noting differences due to inclusion of atmospheric feedbacks in models vs local-only signal in observations.
• Paired in situ sites: 24 forest–openland neighboring site pairs from FLUXNET2015 and AmeriFlux in North America and Europe (mean separation ~24.6 km; mean elevation difference ~71.7 m) used to compare DTDT, wind speed, SHF variability, upward shortwave radiation, and albedo variability between covers; Kolmogorov–Smirnov two-sample test (95% level) applied to distributions of daily temperature differences (δT).

Statistical significance:

• Model agreement denoted where most/all models share change sign; Student’s t-test used where applicable; K–S test for δT distributions.

• Idealized large-scale deforestation increases daily temperature variability (DTDT and SDT) prominently in the northern extratropics (North America, Eurasia), strongest in winter:
  • Winter DTDT increases up to ~0.7 °C (≈20%); spring/autumn increases ~0.1–0.5 °C (≈10–15%); summer increases ~0.05–0.3 °C (≈5–10%). All five models agree on sign across affected regions.
  • Tropics and Southern Hemisphere show negligible DTDT changes (within ±0.05 °C or ±5%) with low inter-model consistency.
• Rapid warming and cooling events both become more frequent post-deforestation:
  • In North America (winter), frequency within −3 to 3 °C δT decreases by 5.3%, while δT < −3 °C and δT > 3 °C increase by 2.8% and 2.5%, respectively; similar symmetric tail increases across seasons and in Eurasia (except summer in Eurasia not significant).
  • Percentage increases grow with extremity: moderate (|δT| 5–10 °C) and extreme (|δT| >10 °C) winter δT frequencies rise by ~12.8% and ~34.1% in North America; larger local-scale increases noted.
• Mechanisms:
  • Enhanced near-surface horizontal temperature advection (TADV) over deforested areas due to:
    • Increased near-surface wind speed from reduced surface roughness (supported by multimodel means and paired-site observations: openland wind > forest).
    • Increased horizontal temperature gradients in midlatitudes in winter/spring due to albedo-driven boreal cooling, enhancing north–south contrast.
  • Decreased day-to-day variability in surface sensible heat flux (DTDSHF) following deforestation (up to about −9 W m−2, strongest in spring/summer) partially offsets increased DTDT. Reduced aerodynamic roughness suppresses turbulent heat exchange variability.
  • Increased day-to-day variability in upward shortwave radiation and albedo (especially in spring with snow) links to larger DTDSHF reduction then; observations support these radiation/albedo variability increases.
• Observational pairing results:
  • Winter: openland sites exhibit higher DTDT than adjacent forest sites (mean difference ~0.07 °C), consistent with simulations.
  • Spring and summer: paired sites show lower or similar DTDT at openland, attributed to atmospheric mixing reducing local TADV contrasts while SHF variability reduction dominates.
• Historical deforestation (1850–2014):
  • North America experienced coherent net tree cover loss (~−1.3 million km² multimodel mean). Seven of eight models simulate winter DTDT increases ~0.2–0.5 °C (≈2–12%); SDT increases ~0.2–0.6 °C (≈2–10%). Mechanisms mirror idealized case: higher TADV, lower DTDSHF.
  • Regional attribution: In deforested North American areas, positive deforestation contribution (+0.11 °C) largely offsets negative combined GHG+aerosol effect (−0.17 °C) on winter DTDT; globally, deforestation effect on DTDT (+0.01 °C) is smaller than combined GHGs+aerosols (−0.03 °C).
• Future afforestation (SSP1-2.6 land use vs SSP3-7.0):
  • Projected increases in tree cover (eastern America, western Europe, central China, central Africa, western Amazon).
  • Eastern America: consistent reductions in DTDT by ~0.06–0.1 °C in winter, spring, and autumn by 2071–2100; smaller or negligible DTDT responses in western Europe and central China (smaller tree cover changes) and in tropical regions (low DTDT sensitivity).
• Overall, deforestation increases day-to-day temperature variability and the frequency of rapid warming/cooling events in northern extratropics; afforestation can reduce variability regionally.

The study directly addresses the previously underexplored question of how deforestation affects daily temperature variability. It shows that large-scale deforestation substantially intensifies variability in northern extratropics, particularly in winter, enhancing both rapid warming and cooling event frequencies, with stronger relative effects for more extreme day-to-day changes. Mechanistically, increased wind speeds due to reduced surface roughness and enhanced horizontal temperature gradients amplify near-surface temperature advection; reduced SHF variability moderates but does not negate the increase in variability. Observations from paired sites corroborate key aspects (wind increase, SHF variability decrease, winter DTDT rise), and differences in local vs nonlocal effects explain seasonal discrepancies in paired-site DTDT. At regional scales, historical deforestation can offset the net impacts of other anthropogenic forcings on daily variability (e.g., in North America), underscoring that land cover change is a non-negligible contributor to variability trends and should be included in detection and attribution studies. Looking forward, afforestation has potential co-benefits by reducing daily temperature variability in northern extratropics, complementing benefits for extremes and carbon sequestration. The findings are relevant for climate risk, public health, ecosystems, and economic impacts that are sensitive to temperature variability, highlighting the policy significance of land-based climate strategies.

This work demonstrates that deforestation increases daily temperature variability in the northern extratropics, especially in winter, leading to more frequent rapid warming and cooling events, and that afforestation can reduce variability in some regions. It identifies enhanced near-surface wind and horizontal temperature advection as primary mechanisms, partially counteracted by reduced variability in sensible heat flux. The results are robust across variability metrics (DTDT and SDT), multiple earth system models, and are supported by reanalysis and paired-site observations. Historical analysis reveals detectable increases in winter variability over deforested North America and shows that deforestation can offset other anthropogenic effects regionally. Future research directions include: conducting high-resolution (convection-permitting) simulations to resolve mesoscale circulations and small-scale deforestation effects; expanding the availability of daily model outputs (e.g., winds) across CMIP/LUMIP models to better quantify TADV-related mechanisms; refining observational strategies to capture both local and nonlocal effects; and integrating variability considerations into afforestation/reforestation policy planning.

• Mechanistic attribution constrained by data availability: daily 10-m wind and other fields needed for TADV were only available for a subset of models (e.g., CanESM5 for idealized, ACCESS-ESM1-5 for historical), limiting multi-model robustness of TADV diagnostics.
• Model resolution: analyses rely on coarse-resolution ESMs (~10–100 km), which may not capture mesoscale circulations induced by small-scale deforestation; responses can be resolution dependent and differ from convection-permitting simulations.
• Observational pairing (space-for-time) primarily reflects local effects and is influenced by atmospheric mixing, potentially under-representing nonlocal feedbacks captured in coupled models; seasonal discrepancies (spring/summer DTDT) arise from this methodological difference.
• Intermodel differences in implementing land-use forcing (LUH2) and vegetation dynamics introduce heterogeneity in tree cover changes and responses, especially in historical experiments.
• Afforestation projections: regional DTDT responses depend on the magnitude and spatial coherence of tree cover change; limited sensitivity in tropics reduces detectability.
• The approximation of diabatic heating contribution via DTDSHF is qualitative and not unit-consistent with DTDT, providing sign rather than magnitude of contribution.