A Multiscale Diagnostics Hierarchy for Detecting, Source-Tracking, Understanding, and Reducing Model Biases in the US Warm Season S2S Precipitation Variability

Precipitation processes are multi-scale in nature. A faithful representation of precipitation in a model relies on its capability to capture 1) large-scale atmospheric circulation patterns that trigger the development of a precipitating weather event such as cyclones and thunderstorms, and 2) local, smaller scale physical processes (including convection, radiation, cloud physics, air-sea interaction, etc.) that determine the lifecycle of a weather event through their interactions with large-scale flow. In the central United States, warm season (March-August) precipitation is mainly associated with Mesoscale Convective Systems (MCS), a form of “layered” overturning circulations that is often poorly resolved or parameterized in a global climate model. The failure of such parameterizations to realistically account for scale-interactions, together with model intrinsic biases in reproducing large-scale forcing of MCSs, poses a major challenge in our effort to simulate and predict warm season precipitation, particularly across the S2S timescales. In response to this challenge, here we propose a multi-scale diagnostics hierarchy for detecting, source-tracking, understanding and reducing model biases in the US warm season S2S precipitation variability. The cornerstone of this hierarchy is the partitioning of MCS processes into two components: large-scale forcing and local, smaller scale physics. Teasing out large-scale forcing from a myriad of interacting scales of an MCS allows one to potentially trace the origin of MCS variability and identify remote sources of predictability for MCS precipitation.
By integrating data diagnosis with numerical modeling, the PIs will develop the diagnostics hierarchy targeting processes of MCS initiation, growth and decay. Specific tasks to be carried out include 1) constructing new evaluation metrics to quantify the S2S variability in the U.S. warm season precipitation, 2) statistical mapping of MCS variability onto S2S precipitation variability, 3) partitioning the GFDL AM4’s MCS biases into components associated with large-scale forcing and model physics, 4) multi-scale diagnostics and idealized modeling to reveal the dynamical nature of model biases in MCS large-scale forcing, 5) experimenting with new packages of model physics to further understand the contribution of local processes to MCS biases, and 6) connecting model biases in MCS large-scale forcing with modes of climate variability and exploring remote sources of S2S predictability for MCSs with NOAA-funded field campaign observations.
The proposed project is a direct response to the joint competition to “advance process understanding and representation of precipitation in models”. Aiming at the longstanding problem of MCS simulation, we will develop, test, and deliver to the community an innovative multiscale diagnostic framework that encompasses process-level metrics development, scale-resolving diagnostics, error partitioning, source tracking, and generation of dynamics-based guidance for model optimization and update. This work contributes directly to the goal of “Focus Area A: Identifying and understanding key processes that influence model biases and systematic errors in the simulation of precipitation at the subseasonal to seasonal (S2S) timescale”. The insights gained from the scale-resolving bias attribution will also pave the way for formulating and testing (with NOAA field campaign observations) hypotheses regarding remote sources of S2S predictability of precipitation from the tropical Indo-Pacific and Atlantic. Given the significance of S2S precipitation forecasts for hazards mitigation and water resource management, the proposed project will ultimately contribute to the objective of the NOAA CPO – “advancing scientific understanding, monitoring, and prediction of climate and its impacts to enable effective decisions”.