P2 – Cloud Reconstruction and 3D Radiative Transfer

This project targets an improved understanding of magnitude and impact of cloud-related 3D radiative effects by means of forward radiation modelling. While 3D radiative transfer models are available nowadays as mature tools in the scientific community, a comprehensive understanding of the magnitude of 3D radiative effects, their dependencies on cloud properties and structure, and their relevance for climate remains limited. In particular, the relative importance of cloud microphysics versus morphology for determining the magnitude of 3D radiative effect remains unclear. While it is evident that current kilometer-resolution satellite imagers and atmospheric models cannot fully resolve cloud variability as needed to realistically represent 3D effects, it is not yet clear which spatial scales have to be considered to resolve bias and small-scale variability introduced by 3D radiative transfer effects. Also, most studies investigating 3D effects have considered scenes with idealized cloud types, and the dependence of 3D effects on cloud type and/or regime remains elusive.

Project 2 aims at resolving cloud-induced bias and small-scale variability in radiation. These will be related to characteristic 3D cloud structural parameters obtained from realistic cloud reconstructions as well as cloud model simulations. On the one hand, the cloud reconstruction will utilize a combination of vertically resolved cloud properties from ground-based active remote sensing and horizontally highly-resolved cloud property fields obtained from passive satellite imagers. Also, the corresponding tools will be refined within Project 2. On the other hand, the reconstructed cloud properties will be used as an input for the versatile 3D radiative transfer model MYSTIC, the Monte Carlo code for the physically correct tracing of photons in cloudy atmospheres.

Thus, Project 2 strives for a better understanding of the sensitivity of 3D radiative effects to cloud structure. It will provide the regime-resolved relationships between cloud structure and radiation quantities. Furthermore, we will study the resolution-dependence of 3D radiative transfer effects.

Principal Investigators:

Project Scientists:

Research Questions

The primary research questions of this project are:

  1. How large are systematic biases and small-scale fluctuations in the radiation budget resulting from 3D effects? Can they be meaningfully partitioned by cloud regime?
  2. What structural and physical cloud parameters determine 3D bias and variability? How do systematic and random 3D effects affect satellite cloud retrievals (jointly with P4)?
  3. How well are the different instrumental perspectives able to resolve and constrain relevant structural parameters of clouds?
  4. What model and observational resolution is required to quantify the complete 3D effect and how does that depend on cloud regime (jointly with P1)?

Work Programme

A central part of this project is the development of a purely observation-based scene reconstruction of 3D cloud properties as input for the 3D radiative transfer simulations, as contrast and reference for the simulations carried out in project 1. Thereby, vertical cloud profiles from ground-based remote sensing will be combined with passive satellite images. For this purpose, also the observations performed during the Small-Scale Variability of Solar Radiation campaign (S2VSR, June – August 2023) at the ARM Southern Great Plains site will be valuable.

From the ground perspective of a human observer, the International Cloud Atlas defines the notion of cloud types, which is the basis of synoptic cloud observations. Approaches to standardize the classification of clouds from space-borne observations are much less mature. The most widely used methods come from the International Cloud Climatology Project (ISCCP). Its cloud type definition considers values of cloud optical depth and cloud top pressure from passive satellite observations as basis. The ISCCP cloud regime classification will be established in Project 2. The sensitivity of the resulting cloud regimes to the domain size used for generation of the 2D histograms will be investigated, with the aim to select an optimal size for characterizing local-scale cloudiness.

dumm text From the ground perspective of a human observer, the International Cloud Atlas defines the notion of cloud types, which is the basis of synoptic cloud observations. Approaches to standardize dummyy text
European Cloud regimes, clustering of 2D joint histograms of cloud optical depth and top pressure (from Tzallas et al., 2022). Methodology based on ISCCP, Jakob&Tselioudis (2003).

Furthermore, the cloud regime classification will be extended to include additional aspects such as meteorological variables in the clustering process. Moreover, composite vertical cloud profiles and correlations between cloud properties will be established for different cloud regimes, and used to refine the reconstruction. Thereby, the classification will be designed in such a way, that it is also applicable to the model results of Project 1.

Within C3SAR, radiative transfer calculations will be performed with the Monte Carlo code for the physically correct tracing of photons in cloudy atmospheres (MYSTIC). The code was developed from Bernhard Mayer, who is a PI in this project. At the International Intercomparison of 3D Radiative Transfer Codes (I3RC, Cahalan et al., 2005), which was a coordinated effort to intercompare and improve available radiative transfer codes, MYSTIC was found to be one of the most versatile codes. The code was continuously extended and improved since then.

The radiative transfer simulations with MYSTIC will build upon a three-fold dataset. The first dataset constitutes the reconstructed data from this project. These data will have the highest resolution in this study. However, they will probably not cover all aspects relevant as input to a radiative transfer model, like e.g. detailed description of the ice microphysics and will be complemented by realistic assumptions from the literature. Furthermore, the simulations from Project 1 will provide cloud data with high horizontal resolution of down to 75m. Thereby, the use of the new ICON model version guarantees best possible model physics. For the study of the resolution-dependence of the 3D effects we will need even higher resolution where we will use LES simulations with resolutions of 10 meters and smaller.

Within this project, we also aim at further enhancing the computational speed of MYSTIC for the demands of C3SAR. The bias caused by 3D effects is quickly calculated since the bias only requires comparison of domain averages and those are not expensive to calculate. The extra variability introduced by 3D effects, however, is demanding since it requires calculating spatially resolved radiance or irradiance fields. We will focus on two major improvements: On the one hand, we will implement the option to define the desired Monte Carlo standard deviation rather than the number of photons to calculate. Especially in a broken cloud scene the standard deviation will vary considerably from pixel to pixel. With this improvement, many photon-calculations could be saved at pixels which quickly reach a low noise level. On the other hand, another improvement would be to quantify and subtract Monte Carlo noise from irradiance or radiance histograms to reduce computational time. This is based on the first mentioned improvement, requiring that the standard deviation of all output pixels is identical. For the application in this project we are mainly interested in bias and variability introduced by 3D effects. While the bias for a scene is quickly determined with high accuracy as the average over all output grid boxes, the variability is composed of the real variability and the Monte Carlo noise. While the latter can be determined as a by-product of the Monte Carlo computation, it can be subtracted from irradiance or radiance histogram assuming Gaussian distribution of the Monte Carlo noise. Both would be highly innovative additions to the MYSTIC model, improving model performance without introducing approximations.

In cooperation with Project 1 and Project 4, differences between modelled cloud fields and the observation-based reconstructions will be evaluated in detail, with focus on differences in cloud micro- and macrophysical properties and cloud morphology. This work shall in particular determine their relevance for the magnitude of 3D radiative effects. Additionally, we will identify suitable metrics to describe cloud structures. Thereby, we will focus on how different cloud properties change over time, across space, and under different weather conditions. Here, the ultimate goal is to assess their importance and how they interact to modulate the 3D radiative effects of clouds.