The cloud radiative kernel method is a popular approach to quantify contributions to the cloud feedback from changes in cloud amount, altitude, and optical depth, separately for low and non-low clouds.  However, because this method relies on cloud property histograms derived from passive satellite sensors (or produced by passive satellite simulators in models), changes in obscuration of low-level clouds by upper-level clouds can cause apparent low cloud feedbacks even in the absence of changes in low-level cloud properties. In this work, we quantify the impact of changing obscuration on cloud feedback and provide a recommended methodology for properly diagnosing the true low and non-low cloud feedbacks. Increases in upper-level cloud cover that obscure underlying low-level clouds cause an apparent positive low-cloud amount feedback that is unrelated to low cloud changes.  Because these radiative changes are solely due to changes in upper-level cloud cover, we recommend that this component be removed from the low cloud feedback and instead be included as part of the non-low cloud amount feedback. Averaged across a suite of CMIP5 and CMIP6 global climate models subjected to a uniform 4K warming of the sea surface temperature, true low cloud feedbacks – those occurring in the unobscured portion of the scene – tend to be smaller (less positive) than their original unadjusted counterparts. Conversely, true non-low cloud feedbacks tend to be larger (more positive) than their original unadjusted counterparts. Accounting for obscuration effects reduces the inter-model spread in both non-low and low cloud amount feedbacks while simultaneously removing a mostly artificial anti-correlation between these two feedbacks.

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. It is supported by the Regional and Global Model Analysis Program of the Office of Science at the DOE. IM Release # LLNL-ABS-861525.