Uncertainty quantification based cloud parameterization sensitivity analysis in the NCAR community atmosphere model

Abstract

Using uncertainty quantification techniques, we carry out a sensitivity analysis of a large number (17) of parameters used in the NCAR CAM5 cloud parameterization schemes. The LLNL PSUADE software is used to identify the most sensitive parameters by performing sensitivity analysis. Using Morris One-At-a-Time (MOAT) method, we find that the simulations of global annual mean total precipitation, convective, large-scale precipitation, cloud fractions (total, low, mid, and high), shortwave cloud forcing, longwave cloud forcing, sensible heat flux, and latent heat flux are very sensitive to the threshold-relative-humidity-for-stratiform-low-clouds ($$rhminl)$$ r h m i n l ) and the auto-conversion-size-threshold-for-ice-to-snow $$\left( {dcs} \right).$$ dcs . The seasonal and regime specific dependence of some parameters in the simulation of precipitation is also found for the global monsoons and storm track regions. Through sensitivity analysis, we find that the Somali jet strength and the tropical easterly jet associated with the south Asian summer monsoon (SASM) show a systematic dependence on $$dcs$$ dcs and $$rhminl$$ rhminl . The timing of the withdrawal of SASM over India shows a monotonic increase (delayed withdrawal) with an increase in $$dcs$$ dcs . Overall, we find that $$rhminl$$ rhminl , $$dcs$$ dcs , $$ai,$$ a i , and $$as$$ as are the most sensitive cloud parameters and thus are of high priority in the model tuning process, in order to reduce uncertainty in the simulation of past, present, and future climate.

Uncertainty Quantification and Bayesian Inference of Cloud Parameterization in the NCAR Single Column Community Atmosphere Model (SCAM6)

“…A decrease in PRECT (∼3.7 to 3.1 mm/day) in response to an increase in dcs is also noted (Figure 9), which indirectly alters (decreases) the stability of the atmosphere (CAPE; Figure 7) and thus the convective precipitation (PRECC; Figure 9), since changes occur in CMP but not in convection parameterizations. This phenomenon was also reported for different CMP parameters by Lin et al (2016) and Pathak et al (2020). In addition, the increase in ai caused more ice particles to fall, thereby decreasing CLDHGH (∼85 to 78%), LWCF (∼30 to 19 W/m 2 ) and SWCF (−35 to −29 W/m 2 ), and IWP (0.1 to 0.06 kg/m 2 ) (Mitchell et al, 2008), while causing an increase in LWP from 0.044 to 0.049 kg/m 2 (Figures 7, 8).…”

“…While, the second-and higher-order sensitivity of dcs, ai, rhmini, rhmaxi, and eii together contribute about 7 and 20% of the total variance in different QoIs. Some previous studies using the predecessor version of this model have also argued that dcs and ai are the most sensitive parameters for the simulation of total precipitation over the tropical region (He and Posselt, 2015;Qian et al, 2015;Zhang, 2015;Pathak et al, 2020). Other studies that employed different models have also suggested dcs and ai to be highly sensitive to cloud distribution (e.g., Bony and Dufresne, 2005;Sanderson et al, 2008;Golaz et al, 2011;Gettelman et al, 2012).…”

“…We selected eight parameters to quantify the uncertainties in cloud microphysics and macrophysics (hereafter CMP) as outlined in Table 1. Although many uncertain parameters are involved in the CMP scheme, we considered only few parameters that have been reported as being sensitive in the previous version of this model (e.g., Covey et al, 2013;Yang et al, 2013;Qian et al, 2015Qian et al, , 2018Pathak et al, 2020). The ice cloud fraction (CF i ) is calculated from relative humidity (RH) using the total ice water mixing ratio [i.e., the ice mass mixing ratio (q i ) plus water vapor mixing ratio (q v )] and the saturated vapor mixing ratio over ice (q sat ), as follows:…”

“…In our setting, posterior parameters distributions were inferred from data collected in subtropics at the Southern Great Plains (36 • N and 97 • W; Zhang et al, 2016), and as such the resulting parameters are calibrated for such regions, which also explains the improvement we obtained in the subtropics and tropics. Recent results (Pathak et al, 2020) suggest a spatial dependency of some of the considered parameters despite being prescribed as constant values across the globe. Our results over the polar regions also support the use of spatially varying cloud parameters in climate models, which will be explored in depth in our future work.…”

“…Uncertainties in climate model simulations are mainly due to cloud parameterizations such as cloud distribution, cloudaerosol interactions, cloud feedback, and the convective activity of clouds (e.g., Albrecht et al, 1988;Bony and Dufresne, 2005;Lee et al, 2011Lee et al, , 2012Carslaw et al, 2013;Zelinka et al, 2013Zelinka et al, , 2020Zhao et al, 2013;Bony et al, 2015;Pathak et al, 2020). Significant improvements in representing convection, clouds, and cloud-aerosol interactions in climate models have been achieved in the last two decades (e.g., Sanderson et al, 2008;Gettelman et al, 2010;Golaz et al, 2011;He and Posselt, 2015;Anand et al, 2018;Pathak et al, 2020). However, work is still required to minimize the uncertainties associated with cloud representation (e.g., Schwartz, 2004;Lohmann et al, 2007;Gettelman et al, 2012;Lee et al, 2012;Hazra et al, 2015).…”