Observational Constraints on Cloud Feedbacks: The Role of Active Satellite Sensors

Abstract

Cloud profiling from active lidar and radar in the A-train satellite constellation has significantly advanced our understanding of clouds and their role in the climate system. Nevertheless, the response of clouds to a warming climate remains one of the largest uncertainties in predicting climate change and for the development of adaptions to change. Both observation of long-term changes and observational constraints on the processes responsible for those changes are necessary. We review recent progress in our understanding of the cloud feedback problem. Capabilities and advantages of active sensors for observing clouds are discussed, along with the importance of active sensors for deriving constraints on cloud feedbacks as an essential component of a global climate observing system.

Statistical downscaling of water vapour satellite measurements from profiles of tropical ice clouds

“…ε represents a zero-mean Gaussian noise. Here, RH l is assumed to follow a beta distribution, which is the usual choice for continuous proportion data, and its canonical link function, the logit g(x) = log x 1−x , is used (Wood, 2011), which ensures that all values are in the (0,1) interval. To estimate each f , we can represent it as a weighted sum of known basis functions z k (·),…”

“…(2), we used the approach of Wood (2011): the appropriate degree of smoothness of each spline is determined by setting a maximal set of evenly spaced knots (i.e. bias(f ) var(f )) and then controlling the fit by regularization, by adding a "wiggliness"…”

“…These models allow us to account not only for the non-linear effects of the predictors (under the assumption of additivity), but also for their spatial distribution: two SR profiles, and therefore the corresponding water vapour structures, are more likely to be dependent if they are close by some metric. Given a set of geographical locations s, a (bivariate) smooth term f (s) can be represented as the random effect f (s) = (1, s T )γ + j w j C(s, s j ), with w ∼ N(0, (λC) −1 ), γ a vector of parameters and C(s, s j ) = c(||x − x j ||) a non-negative function such that c(0) = 1 and lim d→∞ c(d) = 0, which is interpretable as the correlation function of the smooth f (Wood, 2011). By adding this term to the model in Eq.…”

“…By adding this term to the model in Eq. (2), we explicitly include the spatial autocorrelation in the SR data without changing the mathematical structure of the minimization problem, and we can still use the GAM basis-penalty representation (Wood, 2011). Here, we assumed an isotropic exponential correlation function C(s, s j ) = exp(− s−s j /r) with the range r chosen equal to the size of SAPHIR pixels (10 km).…”

Improved Hydrometeor Detection Method: An Application to CloudSat

“…Narrowing the range in estimates of cloud feedback requires accurate long-term observations to understand and constrain physical mechanisms responsible for those changes (Bony et al, 2006;Kollias et al, 2005). In particular, satellite observations over decades are necessary to directly detect expected changes and separate them from natural variability (Geer et al, 2017;Winker et al, 2017).…”