Deep 3D Scattering of Solar Radiation in the Atmosphere Due to Clouds-D3D ^†

Abstract

The three-dimensional (3D) structure of clouds is a key factor in atmospheric processes, profoundly influencing solar radiation transfer, weather patterns, and climate dynamics. However, accurately representing this complex structure in radiative transfer models remains a significant challenge. As part of the Deep 3D Scattering of Solar Radiation in the Atmosphere due to Clouds (D3D) project, we conducted a comprehensive study on the role of all-sky imagers (ASIs) in reconstructing observational 3D cloud fields and integrating them into advanced 3D cloud modeling. Since November 2022, a network of four ASIs has been operating across the broader Patras region in Greece, continuously capturing atmospheric measurements over an area of approximately 50 km². Using simultaneously captured images from the ASIs within the network, a 3D cloud reconstruction was performed utilizing advanced image processing techniques, with a primary focus on cumulus cloud scenarios. The Structure from Motion (SfM) technique was employed to reconstruct the 3D structural characteristics of clouds from two-dimensional images. The resulting 3D cloud fields were then integrated into the MYSTIC three-dimensional radiative transfer model to simulate and reconstruct solar irradiance fields.

1. Introduction

Aerosols and clouds are among the most critical atmospheric constituents influencing the variability of solar irradiance reaching the Earth’s surface [1]. Their direct and indirect radiative effects significantly affect the magnitude and temporal variability of both global horizontal and direct normal irradiance [2]. The high spatial and temporal variability of both aerosols and clouds presents a major challenge in accurately estimating solar resources, particularly in regions with limited observational data or complex atmospheric conditions [3]. This variability leads to substantial uncertainty in solar irradiance modeling, forecasting, and climate projections, underscoring the importance of incorporating high-resolution, real-time observations and advanced modeling techniques to improve solar energy resource and climate change assessments.

Uncertainties in solar irradiance modeling largely stem from the inadequate representation of 3D cloud structures. Traditional approaches often rely on the Plane-Parallel Approximation (PPA), which assumes horizontally homogeneous cloud layers. However, this simplification introduces significant errors—up to 20% in thermal radiative fluxes and over 100% in shortwave fluxes—due to the neglect of cloud heterogeneity, spatial variability, and complex cloud-edge effects [4]. In this study, we demonstrate the methodology of reproducing the 3D structure of clouds using advance image processing techniques for sky images captured by one or more ASIs. This information can serve as a valuable input for improving the estimation and nowcasting of solar irradiance resources as well as the role of clouds on climate in future applications.

2. All-Sky Imagers Network

An ASI network was deployed in the Rio area of Patras, Greece. A total of four Mobotix Q24/Q26 ASIs were installed, sourced from MOBOTIX AG, headquartered in Langmeil, Germany. Their spatial arrangement was designed to provide an optimal geometric configuration for 3D cloud reconstruction. Each camera is equipped with a high-resolution 3/6 megapixel color sensor and a wide-angle lens, enabling the capture of high-quality video data.

3. Three-Dimensional Cloud Reconstruction Based on ASIs

The reconstruction of 3D cloud structures was performed using the Structure from Motion (SfM) technique [5,6]. SfM is a widely adopted method in computer vision that generates three-dimensional models from a sequence of two-dimensional images captured from different viewpoints. The fundamental principle of this technique is the exploitation of parallax—the apparent shift in object position relative to the background as the viewing angle changes—in order to estimate the spatial configuration of objects within a scene.

The SfM workflow consists of several essential stages. Initially, features or key points of interest are detected across multiple images (see Figure 1). These features are then matched between image pairs to establish correspondences. Using these correspondences along with information about the camera geometry, the algorithm estimates both the relative camera positions and the 3D coordinates of the matched points through a process known as triangulation. The output is a 3D point cloud, which serves as a spatial representation of the scene’s structure (Figure 1).

In this study, a sequence of multiple 2D images from two All Sky Imagers was used to reconstruct the 3D structure of a cumulus cloud scene over the area of Rio, Patras, Greece. By processing images captured simultaneously from different viewpoints, we were able to infer the volumetric distribution and morphology of the cloud field. This 3D reconstruction provided valuable insights into cloud geometry, layering, and development—factors that are essential for understanding atmospheric processes and hold significant potential for applications in weather modification and radiative transfer modeling.

4. Three-Dimensional Radiative Transfer Modeling

Artificial atmospheric conditions, including various cumulus cloud scenes, were simulated using the libRadtran software packageversion 2.0.5 [4,7]. This package employs the uvspec radiative transfer (RTE) model to compute radiances and irradiances for a given atmospheric state. For the purposes of this study, we utilized a three-dimensional RTE solver—Monte Carlo code for the physically correct tracing of photons In Cloudy atmospheres (MYSTIC) [4,7]. MYSTIC is one of the most widely used solvers for calculating both solar and thermal radiances and irradiances, based on Monte Carlo simulations. It offers physically accurate modeling of photon transport in complex cloudy atmospheres.

Figure 2a illustrates an example of an artificial 3D cumulus cloud field used in the MYSTIC radiative transfer model to simulate solar radiation reaching the Earth’s surface. Cumulus clouds, which are commonly observed in the broader region of Rio, Patras, were selected for this scenario due to their relevance and frequency. Figure 2b displays the corresponding cloud optical depth, derived from the volume extinction coefficient. Figure 2c presents the simulated values of direct, diffuse, and global solar irradiance as calculated by MYSTIC for a reference solar zenith angle (SZA) of 30°, using the 3D cloud structure depicted in Figure 2a,b.

It is evident that direct solar irradiance is attenuated in regions where clouds are present. The degree of attenuation depends on the cloud optical depth, as shown in Figure 2b. As the SZA increases, a reduction in direct solar irradiance is also observed under clear-sky conditions, decreasing from approximately 1100 Wm⁻² (SZA = 10°) to about 700 Wm⁻² (SZA = 50°). Regarding diffuse irradiance, high values are observed near cloud edges due to Mie scattering. Specifically, diffuse irradiance can reach up to 500 Wm⁻² in areas adjacent to clouds, especially at lower SZA values.

Of particular interest is the enhancement of solar irradiance near cloud edges, as observed in the global irradiance values. This increase, compared to clear-sky conditions, is attributed to elevated diffuse irradiance caused by Mie scattering from cloud particles. Incorporating 3D cloud information allows for a more accurate estimation of the solar resource and offers detailed insight into cloud-induced effects, such as irradiance enhancement at cloud boundaries. In a future step, the 3D cloud reconstructions obtained from the ASI network will be integrated into the MYSTIC model.

5. Conclusions

In this study, we present a methodology for reconstructing the 3D structure of cumulus cloud fields using a combination of ASI observations and the SfM technique. We demonstrate the potential of using the reconstructed 3D cloud data in radiative transfer simulations with the MYSTIC model to investigate the impact of cloud geometry on solar irradiance components. The findings highlight the effectiveness of integrating multi-angle imaging with physically based radiative transfer modeling to enhance our understanding of cloud-induced variability in surface solar radiation. This approach offers a promising framework for future research in atmospheric science and supports applications in solar energy forecasting and climate modeling.