Environmental Science and Engineering Seminar
Carbon-water coupling: processes, model uncertainties and observational constraints
Future climate and hydrological projections in response to rising greenhouse gases, in particular carbon dioxyde, are highly uncertain.
Two of the most important issues in climate prediction are due
1) to the representation of clouds and (moist) convection and
2) to (very) large uncertainties in future continental carbon uptake.
In this presentation I will discuss some of those uncertainties and show potential paths forward, emphasizing how coupled the carbon and water cycles are.
1) Clouds and convection: clouds and convection largely control climate sensitivity and thus their representation in climate models is crucial for accurate global but also regional calumet and hydrological prediction. Cloud and convection parameterization has been a longstanding issue for more than 40 years, with still large, systematic, biases that are fundamentally limiting our prediction capacity. I will show how machine learning and high-resolution atmospheric modeling can nonetheless to a large extent break that cloud/convection parameterization deadlock, improving in particular hydrological prediction at the global and regional scale.
2) Continental carbon uptake: A second issue is related to uncertainties in the capacity of contents to act as global carbon sinks in the future. I will first show that most of this uncertainty is due to changes in soil moisture (both variability and trend) and its model representation. Thus the hydrologic cycle is a primary regulator of the carbon cycle, but we will see that vegetation CO2 response also strongly regulates the continental hydrologic cycle. Due to these interactions, predicting the future water and carbon cycles cannot be done in isolation without accounting for carbon and climate feedbacks, so that the water-carbon-climate system has to be studied as an interconnected system. Recent developments in remote sensing (e.g. solar induced fluorescence) along with machine learning techniques can be used to inform and refine ecosystem and land-surface models so they can better capture some of those first order effects/interactions.
Contact: Bronagh Glaser at 626-395-8732 firstname.lastname@example.org