This dissertation aims to better understand the role of moist convection (clouds and turbulence) in the complex dynamic coupling that occurs between components (land and atmosphere) as well as between subregions (Andes and Amazon) of the Earth system, using modern numerical modeling approaches. First, climate model representation of land-atmosphere coupling is compared between standard techniques that use conventionally parameterized convection versus prototype next-generation simulations that can include explicit convection. That is, numerical experiments are designed to isolate the global hydrologic land-atmosphere coupling on weekly-to-subseasonal and seasonal timescales using both approaches to representing convection in climate simulation. Initial analysis focuses on hydrologic coupling dynamics, and then the effects of explicit convection on thermal land-atmosphere coupling are highlighted over the southwestern U.S. and the Arabian Peninsula. Implications on global land surface Bowen ratio and its climate sensitivity are also discovered and discussed. Finally, a separate topic is analyzed, related to clarifying ongoing issues in capturing realistic regional hydroclimate over the Amazon. In this case the coupling dynamics are not between atmosphere and land, but between convective heating over a mountain chain (the Andes) and associated modulations of the atmospheric thermodynamic environment over an upstream rainforest (the Amazon). These dynamics are elucidated using mechanism denial techniques within an ensemble hindcast experiment, which sheds light on a previously underappreciated Andean control on Amazonian rainfall. Key results from these analyses are summarized below.

The first theme (Chapter 2) is to investigate the effects of explicit convection on global hydrologic land-atmosphere coupling across multiple timescales. Over weekly-to-subseasonal timescales, based on experiments following the Global Land-Atmosphere Coupling Experiment (GLACE) protocol, explicit convection is found to systematically mute the soil moisture–precipitation coupling strength on a global scale which is primarily linked to the atmospheric segment (evapotranspiration–precipitation) of the land-atmosphere feedback. Whereas this sensitivity is pronounced on short timescales relevant to synoptically driven drought and hydrologic weather extremes, over longer (seasonal) timescales, based on simulations following the Atmospheric Model Intercomparison Project (AMIP) protocol, the effects of explicit convection on seasonal hydrologic land-atmosphere coupling are undetectable.

The second topic (Chapter 3) is to uncover the effects of explicit convection on global thermal land-atmosphere coupling. In both the GLACE- and AMIP-type simulations, a range of complementary metrics all suggest that explicit convection robustly changes the thermal coupling within specific subregions of the globe. Over the Arabian Peninsula, explicit convection produces less but more realistic time-mean rainfall, pushing local soil moisture regime away from the optimal transitional regime, which is required for strong coupling. As a result, local thermal coupling strength is significantly reduced. Over the southwestern U.S. and northern Mexico, explicit convection substantially enhances the thermal coupling strength independent of soil moisture or precipitation. Instead, such enhancement is related to increases in the local surface Bowen ratio. Based on a variety of simulations across multiple model version pairs of using parameterized vs. explicit convection, I conclude that the use of explicit convection has a systematic and robust effect of increasing the land surface Bowen ratio. Furthermore, under climate change scenarios, comparing to parameterized convection, explicit convection exhibits a remarkably stronger amplification of the land surface Bowen ratio. This opens up new questions in the era of explicitly simulated convection in climate models pertaining to surface climate simulation.

The third subject matter (Chapter 4) of the thesis is motivated by the ubiquitous “Wet Andes – Dry Amazon” (WADA) dipole pattern of rainfall biases seen in the Coupled Model Intercomparison Project Phase 5 (CMIP5) archive. Using the Community Earth System Model version 1 (CESM v1) and focusing on the wet season, mechanism denial experiments are conducted to investigate the remote influence of Andean convection on Amazonian rainfall. It is shown that Andean convective heating reduces rainfall over the Amazon with a maximum of -1 mm/day, and that this response sets up rapidly, within 15 hours. Column moisture budget analysis shows that the total advection of moisture over the Amazon due to Andean convective forcing is controlled by the vertical advective term in ways that can be traced back to a mountain-forced component of the vertical velocity field. Meanwhile, other causative pathways that could control Amazon precipitation through the surface moist static energy, gross moist stability, or column energetics are found to be either irrelevant or secondary.

Tropical precipitation in modern global climate models (GCMs) still shows large errors in both spatial and temporal patterns, despite notable improvement of GCMs over past decades. As tropical precipitation biases are linked to large scale circulation errors and in turn have global impacts in simulated climate, tropical precipitation bias is one of the primary issues to be tackled to improve the accuracy of GCM output. This dissertation addresses key issues in the underlying model dynamics and physics responsible for tropical precipitation biases in modern, state-of-art GCMs.

The first of the three research projects presented in this dissertation systematically tests the time step sensitivity of a superparameterized (SP) GCM to better understand the physical essence of superparameterized convection and its impact on global climatology. The time step in an SPGCM dictates the frequency of information exchange between two scales (be- tween the GCM and the cloud resolving model (CRM) embedded in each grid cell), and, accordingly, a different time step artificially imposes a constraint on model behavior. In an SP version of the Community Atmosphere Model 3.0, several important sensitivities in the tropical climatology are identified, e.g. a faster scale coupling causes weaker cloud forcing,temperature gradients. These sensitivities are distinct from the time step sensitivities of conventionally parameterized GCMs and have implications for understanding emergent be- haviors of multiscale deep convective organization in superparameterized GCMs.

The second research project examines the impact of superparameterization on the initial growth of the double ITCZ (Intertropical Convergence Zone) bias in a dynamic ocean coupled GCM, with the initial intention to test if superparameterization can be a useful tool to mitigate this long-standing model problem by avoiding convective parameterizations that have been known as one of major double ITCZ bias sources. In an SP version of Community Earth System Model (CESM) 1.1, a fast bias growth, e.g. days for precipitation biases and weeks for sea surface temperature biases (also known as the cold tongue SST bias), is observed and validates our novel approach using short, ensemble hindcast simulations. The SP simulations show weaker double ITCZ biases than non-SP simulations, but this improvement is due to non-physical causes. A key discovery is that the historical neglect of convective momentum transport in SP-CESM causes unrealistically strong zonal wind shear near the surface that effectively suppresses the increase of zonal surface wind stress even when stronger overlying easterlies are present. The results highlight a less-appreciated role of convective momentum transport as a potential double ITCZ bias source, providing a new perspective to ITCZ dynamics and also suggesting future model development strategies.

The third research project focuses on a marquee issue of modern tropical climate dynam- ics by exploring how oceanic circulation response buffers forced ITCZ shift when a top-of- atmosphere solar flux perturbation is present at the different latitudinal bands. Particular attention is paid to heat transport associated with Atlantic meridional overturning circula- tion (AMOC) that, unlike subtropical cell, does not have mechanical constraints coupled to Hadley circulation allowing potentially more efficient ITCZ shift buffering. A set of 200-year simulations using CESM 1.2 shows that the heat transport partitioning becomes more ocean centric as the solar forcing is located at higher latitudes and that such sensitivity is indeed driven by the AMOC responses. On the other hand, a previously identified STC (subtrop- ical cell) damping mechanism does not respond sensitively to the solar forcing locations, as a result of compensating circulation strength and depth responses. The findings demand the inclusion of the previously underappreciated AMOC heat transport to the current ITCZ migration framework to fully capture the atmosphere-ocean coupling. Besides, the results have some practical implication in GCM development strategy suggesting fixing tropical bias would be more effective to alleviate tropical precipitation biases in GCMs than fixing extratropical biases.

Ecosystem dynamics are heavily dependent on atmospheric inputs such as rainfall, and are in turn an integral part of land-atmosphere coupling and the global carbon cycle. These global interactions and cycles are commonly modeled by Earth System Models (ESMs), and one of the largest sources of uncertainty in current models is cloud simulation techniques.

Superparameterization (SP) is a proven method to better resolve cloud and rainfall processes in ESMs, but it is prohibitively computationally expensive for long-term climate simulations. The first part of this thesis suggests that – although the latest advances in manycore supercomputing do not provide a promising solution to this problem – a neural network (NN) can successfully be trained on an SP version of the Community Atmosphere Model (CAM) to emulate SP behavior at a fraction of the cost. Incorporating the NN into CAM under idealized aquaplanet conditions results in the Neural Network Community Atmosphere Model (NNCAM).

For NNCAM to work in fully comprehensive models that include interactive vegetation, deeper tests are needed. The second and most major contribution of this thesis is thus to analyze the one-way coupling of NNCAM onto the Community Land Model (CLM). Gross primary productivity (GPP) and net ecosystem exchange (NEE) from ensembles of one-way atmospheric forcing onto a CLM grid cell in the Amazon Basin for NNCAM and SP-CAM are shown to be statistically different from CAM but not each other. Additionally, results suggest that mean precipitation is the largest contributing factor to GPP and NEE in the Amazon Basin.

The water cycle is one of the most fundamental building blocks of the earth system; without it, life would cease to exist, but its extremes pose a threat to both economies and ecosystems. It is thus especially important to understand the hydrologic cycle in as much depth as possible, including how human actions are already shaping it and how it could change in the future. In that vein, this dissertation addresses three distinct topics – how irrigation has altered precipitation; how plant-physiological changes in response to rising CO2 can alter future flooding and streamflow; and how the Madden-Julian Oscillation (MJO) modulates tropical cyclones (TCs) in the West Pacific.

The first main chapter of this dissertation (Chapter 2) assesses non-local hydroclimate responses to irrigation in India, using an ensemble hindcast approach aimed at clarifying the ongoing debate in the literature about the robustness of forced water cycle responses relative to high levels of atmospheric internal variability. The results suggest a strong sensitivity to the initial synoptic condition, with separate non-local hotspots responding to irrigation differently (but robustly) under different initial conditions. This argues that chaos plays a major role such that even heavy irrigation such as that over India has difficulty manifesting as strong, robust non-local water cycle responses. On longer time scales and across ensembles, the Meiyu-Baiu rainband region is highlighted as having a potentially robust non-local irrigation-induced precipitation signal, opening new questions.

In my second study (Chapter 3), I investigate streamflow changes as driven by (1) atmospheric responses and (2) plant-physiological responses to rising CO2. A series of four modeling experiments help isolate these two response pathways and their combined effects, revealing that the plant physiological driver is actually of first order importance to projections of future flooding and streamflow. This is especially true in the tropics, where river discharge increases are controlled almost exclusively by the plant response in the Amazon, Parana, Congo, and Yangtze. This adds to a growing recognition in our field that profound changes in regional water cycles can occur even without warming-induced changes of precipitation as a result of how the terrestrial biosphere adapts to increased CO2.

The final project (Chapter 4) seeks to better understand the mechanisms through which the MJO, a slow-moving tropical weather pattern, modulates West Pacific TCs. How, where, and even if the MJO modulates tropical cyclones is unresolved in this region, despite the fact that it is especially prone to human vulnerability from both high levels of current TC activity and future MJO amplification. Through a novel downscaling strategy that creates thousands of synthetic cyclone tracks for each phase of the MJO, I reveal two previously un-emphasized but distinct stationary modes in opposite portions of the West Pacific basin that are modulated out of phase with one another. The South China Sea region is particularly responsive to the oscillation, driven by a transient combination of dynamic and thermodynamic factors. This adds to a decades long debate about what mechanisms mediate MJO-TC modulation in nature and identifies new subregions that will be particularly important to focus on in the coming decade, towards advancing understanding of how MJO amplification may affect TC hazards in a future climate.