Exploring how land surfaces impact convection using cluster analysis



Fast, Jerome D — Pacific Northwest National Laboratory

Area of research

Surface Properties

Journal Reference

Chen J, S Hagos, H Xiao, J Fast, and Z Feng. 2020. "Characterization of Surface Heterogeneity Induced Convection using Cluster Analysis." Journal of Geophysical Research: Atmospheres, 125(20), e2020JD032550, 10.1029/2020JD032550.


Accurately predicting the timing, location, and intensity of convection over highly populous land areas is extremely important. However, simulating these complex, highly turbulent, small-scale processes takes significant computing power. By combining real land properties, a regional numerical model, and an unsupervised machine-learning algorithm (i.e., cluster analysis) researchers identified six distinct daytime cycles of clouds and precipitation. They first used cluster analysis to simplify the complexity introduced by the inclusion of realistic soil moisture, land use, and land type feature data. They found that the earliest convection occurs in two clusters over dry land and the cooling effects from the evaporation of the corresponding precipitation triggers additional convection in neighboring clusters. These findings enhance our current understanding of land-atmosphere coupling.


Precipitation influences humans’ everyday lives. Accurately predicting the timing, magnitude, and location of atmospheric convection is essential to estimate the global energy budget that affects climate and provide precise precipitation forecasts to the public. The results of this study clarify how land properties influence convective initiation and emphasize a specific mechanism (i.e., cold pool lifting) that spreads the convection. Importantly, this study clearly shows the initiation and the spreading of convection over complex and highly variable land properties for the first time.


Researchers characterized the influences of soil moisture, land use type, and soil type on the timing and location of shallow cloud formation by conducting large-eddy simulations using the Weather Research and Forecasting model that includes an interactive land model. The case study comes from data collected during the Holistic Interactions of Shallow Clouds, Aerosols, and Land-Ecosystems field campaign, which took place over the Southern Great Plains. Using information about surface moisture levels and land use available at the selected site, scientists designed control and sensitivity experiments to isolate the effects of the local surface properties. The comparisons of control and “no-large-scale advection,” which disregard the impacts of the surrounding region on the movement of air within the studied area, simulations show that large-scale advection moves clouds from over dry soil to over wet soil. Applying cluster analysis in the latter simulation revealed key features of land-atmosphere interactions. Specifically, convective clouds that first form over regions can transfer more conductive heat to the atmosphere. The precipitation from those convective clouds triggers new, nearby updrafts about two hours after those convective clouds. The cluster analysis also shows that in addition to the spatial pattern of soil moisture, land use and soil texture in western Oklahoma also influence where convective clouds begin forming.