To a large degree, Professor N. A. Phillips laid the basis for numerical weather prediction and climate modeling. In particular, he made seminal contributions that provided a physical basis through which inter-regional communication occurred in physical systems leading to the understanding of how real signals are transmitted and how errors originating in data poor regions (for example) selectively infect other regions. Much of this work is reviewed in the Sixth IMO Lecture embodied in the “Dispersion Processes in Large-Scale Weather Prediction” (World Meteorological Organization Technical Document No. 700: 1990, 126 pp). The study is based on the proposition that “… at the present time, the accuracy of initial analyses is the major limitation to the successful prediction of large-scale flow patterns..” from which Phillips posed the question “.. How rapidly do influences from regions of inadequate data spread to those regions for which an accurate forecast is desired?” The WMO document provides a detailed accounting of basic theory of wave dispersion in the presence of zonal flows and lays the basis for both an understanding of the propagation of real atmospheric signals as well as patterns of initial error growth and their subsequent remote degradation of forecasts.

We consider two extensions of Phillips work. First, we consider wave dispersion in flows that vary in three dimensions. The emphasis is on the dispersion of energy from “centers of action” in the tropics towards the extratropics. It is shown that a combination of latitudinal and vertical shears in the slowly varying basic state plus, and especially, zonal variations (stretching deformation) of the basic flow, lead to regions of “energy accumulation” where the initial signal is amplified. For example, the regions of upper tropospheric westerlies in the tropics form a natural region of accumulation and also a conduit between the tropics and extratropics on both weather and climate time scales (defining, respectively, “fast” and “slow” teleconnections) and possibly between the troposphere and the stratosphere. Second, we consider the propagation of events that appear to exist uniquely in spectral space and ask the question about how numerical weather prediction may be extended by emphasizing their unique nature. Specifically, we refer to intraseasonal variability (15-40 days) in the tropics and monsoon regions and how their influence is dispersed through a complex atmospheric state. We note that a physically based Bayesian statistical technique show substantial predictability on the 20-25 day time scales and wonder if these slow manifold physics can be translated into gains in numerical weather prediction beyond current multi-day limits.

We close with some personal remarks regarding Professor Phillips’s contributions to the field of numerical weather prediction and (in particular) his role as advisor and mentor.