Increasing contribution of the atmospheric vertical motion to precipitation in a warming climate

The study addresses how precipitation, especially extremes, is modulated by dynamic (vertical motion) and thermodynamic (moisture availability) processes in a warming climate. While the wet-get-wetter, dry-get-drier paradigm helps explain mean precipitation changes, extremes are influenced by interacting processes that vary by region. Prior analyses using time-averaged moisture budgets cannot reveal day-to-day interactions or the propagation of perturbations among components. The purpose here is to develop a structural, dynamically interactive model of atmospheric moisture that can track how shocks to specific processes (vertical motion, moisture content, horizontal advection, evaporation) propagate and jointly determine precipitation on daily timescales under present and future warming scenarios. Understanding these mechanisms is crucial for anticipating changes in heavy precipitation and associated societal impacts.

The atmospheric moisture budget decomposes precipitation into contributions from vertical moisture advection (split into dynamic and thermodynamic components), horizontal moisture advection, and evaporation. Previous studies quantified these components, often emphasizing the relative importance of one over others and applying linearized decompositions to separate vertical motion (dynamic) from moisture content (thermodynamic), sometimes linking thermodynamics to Clausius-Clapeyron scaling. However, these time-averaged approaches neglect interactions among components and their temporal evolution, despite evidence that increased moisture reduces the required ascent for a given precipitation intensity, implying dynamic–thermodynamic interplay. Horizontal moisture advection has been shown to source moisture for later precipitation. There is also mixed evidence across regions and models regarding whether extremes scale mainly with thermodynamics or are dominated by dynamics, motivating a temporally resolved, system-level analysis.

Data: Daily outputs from 10 CMIP5 coupled models (CanESM2, CMCC-CM, CNRM-CM5, FGOALS-g2, GFDL-CM3, GFDL-ESM2M, IPSL-CM5A-LR, MIROC5, MPI-ESM-MR, MRI-CGCM3) were used. Variables: precipitation, surface latent heat flux (evaporation), specific humidity, zonal and meridional wind, surface pressure, and omega (vertical velocity). Outputs were regridded to 2° × 2.5°. Periods: pre-industrial (20 years), present (2006–2025), and future (2081–2100) under RCP4.5. Moisture budget: P = −[q∇·V] − [V·∇q] − [∂q/∂t] + E, with storage term small and neglected. Vertical advection [ω ∂q/∂p] is linearly decomposed into: (i) dynamic component DY (changes in ω holding ∂q/∂p at climatology) and (ii) thermodynamic component TH (changes in ∂q/∂p holding ω at climatology). The remaining second-order terms are ignored. Thus P ≈ DY + TH + HA + E, where HA is horizontal advection and E is evaporation. Climatological means for the decomposition are computed over the PI 20 years, and over 2006–2100 for present and future climates. Dynamic system and identification: A structural vector autoregression (SVAR) is specified over the five variables (HA, DY, TH, P, E) at each grid cell: Ay_t = B_1 y_{t−1} + … + B_k y_{t−k} + ε_t, ε_t ~ N(0, I). Identification uses Cholesky orthogonalization with precipitation ordered last; all permutations of the remaining four variables are considered (4!). Optimal lag length is chosen using information criteria. Impulse-response functions (IRFs): For each grid and each valid ordering where the impulse precedes the response, IRFs are computed and averaged across models and orderings. Shocks are normalized to one standard deviation of the impulse variable. IRFs are evaluated up to seven days after the shock to quantify the timing and magnitude of impacts on precipitation and inter-component interactions. Forecast error variance decomposition (FEVD): The variance of precipitation forecast errors is decomposed into contributions from shocks to DY, TH, HA, E, and P itself. Spatial considerations and residuals: Neighbor-grid variables were tested in SVARs for 100 random locations and found to have an order-of-magnitude smaller influence than local variables; thus not included due to computational cost. Residual closure of the moisture budget was assessed; global residuals R are −0.24, −0.17, and −0.28 mm d−1 for PI, present, future, respectively (about 5–10% of global-mean P). Geographical and land–ocean analyses: IRFs and their distributions are aggregated over land, ocean, and globally; spatial patterns of impacts and their climate-state differences are mapped.

- Time-averaged means show increased global precipitation of 6.7% (present) and 13.3% (future) versus pre-industrial, with warming of about 1.2 °C by 2025 and slightly over 3 °C by 2100. - Instantaneous dynamics: IRFs reveal that a shock to the dynamic component (vertical motion) produces the largest same-day increase in precipitation, decaying rapidly thereafter. A shock to the thermodynamic component yields a smaller immediate effect that persists for about two days (same day and next day). Horizontal moisture advection peaks in its effect on precipitation the following day, with minimal same-day impact. Evaporation’s impact is largest on the same day and drops substantially afterward. - Dominance of dynamics: FEVD attributes 47.6% of precipitation forecast error variance to dynamic shocks, exceeding precipitation’s own contribution (44.4%). Thermodynamic, HA, and evaporation shocks contribute 3.5%, 1.2%, and 3.2%, respectively. - Warming intensifies impacts: Across PI, present, and future climates, impacts of all components on precipitation increase. Dynamics’ same-day effect strengthens, while thermodynamic, HA, and evaporation impacts tend to persist up to four days in warmer climates. Despite stronger dynamic impacts, their duration does not lengthen, consistent with rapid moisture removal by vertical motion. - Increased spread and extremes: Probability densities of precipitation responses at the time of maximum impact show increased spread with warming for dynamics, thermodynamics, and HA, with the largest spread increase for dynamics. Probabilities of low/medium impacts decrease and extreme impacts increase, implying heavier precipitation becomes more likely; light precipitation becomes less likely in the future. - Land–ocean contrasts: The spread increase for thermodynamic impacts is larger over land; for HA it is larger over ocean. - Spatial patterns: As climate warms, same-day dynamic impacts strengthen in the tropics and storm-track regions and weaken in the subtropics, with indications of poleward shifts (e.g., North Pacific). Thermodynamic same-day impacts become less localized but still emphasize tropics and storm tracks with further warming. HA emphasizes increased moisture transport to extratropics for next-day precipitation, especially in the future. - Interactions among components: Dynamically induced rainfall reduces next-day atmospheric moisture and horizontal advection, and increases relative humidity that depresses same-day surface evaporation. Increased thermodynamic moisture boosts dynamics (upward motion), supporting more precipitation; this coupling strengthens in warmer climates. Evaporation increases are associated with reduced near-surface relative humidity states and tend to be linked with reduced dynamics and HA. In warmer climates, dynamic events more completely exhaust column moisture (higher precipitation efficiency), contributing to heavier rainfall for the same dynamic forcing.

The results demonstrate that vertical motion dynamically controls the timing and intensity of daily precipitation, especially extremes, with thermodynamics and horizontal advection acting primarily to supply moisture over one to several days. This helps reconcile discrepancies between Clausius–Clapeyron moisture increases and observed precipitation changes: CC scaling does not directly set daily precipitation without concurrent ascent. The event-based, temporally resolved analysis reveals contributors to heavy-rainfall days that differ from time-averaged moisture budgets, in which dynamic ascent regions coexist with divergence/drying. With warming, dynamics become more effective at depleting column moisture, and the distribution of dynamic impacts shifts toward heavier events and fewer light events, consistent with increased precipitation variability and poleward-shifting storm tracks. While some prior studies suggested thermodynamic dominance for extremes at mid-to-high latitudes, these results indicate a widespread and strengthening dynamic control across latitudes, highlighting the importance of vertical motion and its coupling with moisture availability in a warmer climate.

This study develops a dynamically interactive, daily-scale moisture budget framework using SVAR to quantify how shocks to vertical motion, moisture content, horizontal advection, and evaporation propagate to precipitation. Key contributions are: (i) establishing the dominant same-day role of dynamics in precipitation and extremes; (ii) showing that thermodynamics and horizontal advection predominantly act as multi-day moisture suppliers; (iii) demonstrating that warming increases the intensity and spread of dynamic impacts, leading to more heavy and fewer light rainfall events; and (iv) identifying spatial shifts emphasizing tropics and storm-track regions, with indications of poleward movement. Future work should validate the inferred mechanisms with observations, examine sub-daily convective processes where parameterizations limit direct dynamic diagnostics, and investigate microphysical and local surface controls (soil moisture, cloud cover, topography) on component interactions and precipitation efficiency.

- IRFs quantify responses to one-time, one-standard-deviation shocks; persistent multi-day events likely involve multiple, possibly interacting shocks. - The SVAR framework identifies contemporaneous structure via Cholesky ordering but does not diagnose the physical origin of perturbations (e.g., Rossby waves, atmospheric rivers). - Daily averaging may obscure sub-daily convective dynamics; many models use convective parameterizations without explicit vertical velocities at sub-daily scales. - Model dependence: only 10 CMIP5 models were used; differing convection schemes may affect extratropical results. RCP4.5 was the scenario; other scenarios could yield different regional changes. - Moisture budget residuals indicate 5–10% imbalance at the global mean, reflecting model and diagnostic errors. - Some models required regridding with sparse missing grid information; while tests suggest minimal bias for IRFs and aggregates, uncertainties remain. - Neighboring-grid influences were not included in SVAR due to computational cost, though tests suggested their effects are much smaller than local components.