The data used in this study are obtained from the European Centre for Medium-range Weather Forecasts (ECMWF) Re-Analysis Interim (ERA-Interim; Dee and Uppala 2009; Dee et al. 2011). The atmospheric variables include geopotential height, air temperature, specific humidity, ozone mixing ratio, cloud cover, and cloud liquid/ice water content. All atmospheric variables are defined at 37 pressure levels from 1000 to 1 hPa. We also consider the incoming solar energy flux at the top of the atmosphere (TOA), the surface skin temperature, the upward longwave radiative flux at the surface, surface albedo, and surface sensible/latent heat fluxes.
Following Hu et al. (2016), we use the data in the periods of the four major canonical El Niño winters (1982/1983, 1986/1987, 1997/1998, and 2006/2007) and eight ENSO-neutral winters (1980/1981, 1981/1982, 1985/1986, 1989/1990, 1992/1993, 1993/1994, 2001/2002, and 2003/2004) to construct the composite El Niño and neutral events, respectively. The differences (denoted by the symbol \(\Delta\)) between the composite El Niño and neutral events are referred to as El Niño anomalies. We focus on the El Niño anomalies over the tropical Pacific and part of the East Asia monsoon region (30°S–30°N, 80°E–80°W).
We use the Fu-Liou radiative transfer model (Fu and Liou 1992, 1993) to evaluate all terms in (2) at the original 37 levels of the ERA-interim on each grid point as the following:
$$\begin{gathered} \Delta ^{{(RAD)}} Q_{j} = Q_{j} (T_{{all}}^{E} ,WV^{E} ,C^{E} ,other^{E} ) - Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) \hfill \\ \Delta ^{{(T_{{below}} )}} Q_{j} = Q_{j} (T_{{below\_j}}^{E} ,T_{{others\_b}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) - Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) \hfill \\ \Delta ^{{(T_{{above}} )}} Q_{j} = Q_{j} (T_{{above\_j}}^{E} ,T_{{others\_a}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) - Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) \hfill \\ \Delta ^{{(WV)}} Q_{j} = Q_{j} (T_{{all}}^{N} ,WV^{E} ,C^{N} ,other^{N} ) - Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) \hfill \\ \Delta ^{{(C)}} Q_{j} = Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{E} ,other^{N} ) - Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) \hfill \\ \Delta ^{{(T_{{{\text{air}}}} )}} R_{j} = -\left[ {Q_{j} (T_{j}^{E} ,T_{{other\_j}}^{N} ,WV^{N} ,C^{N} ,other^{N} ) - Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{N} )} \right] \hfill \\ \Delta ^{{(other)}} Q_{j} = Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{E} ) - Q_{j} (T_{{all}}^{N} ,WV^{N} ,C^{N} ,other^{N} ), \hfill \\ \end{gathered}$$
(6)
where Q
j
is the net radiative energy flux convergence (in units of W/m2) in each atmospheric layer obtained from the radiative transfer model using the information of T, WV, and C, as well as other variables/parameters (denoted as “other”) such as the incoming solar radiative flux at the TOA, ozone, and surface albedo. The superscript “N” and “E” denote, respectively, the variables derived from the composite mean fields of the eight ENSO-neutral winters and the four major El Niño winters. In (6), subscript “all” denotes temperature in all layers including the surface level. Subscripts “below_j” and “other_b” represent the temperatures in the layers below j (which includes SST) and the temperatures in the remaining layers, respectively. Subscripts “above_j” and “other_a” denote the temperatures in the layer above j and the remaining layers respectively, and “j” and “other_j” denote the temperatures at the layer j and in other layers respectively. Note that the vertical profiles of C include cloud liquid and ice water as well as cloud area. We then use (2) to infer the term \({\Delta ^{(DYN)}}{Q_j}\) indirectly as
$${\Delta ^{(DYN)}}{Q_j}= - {\Delta ^{(RAD)}}{Q_j}.$$
(7)
Equations (6) and (7) enable us to obtain all of the terms in (4). It can be verified that the approximation (3) is indeed valid, implying that the perturbation of total net radiative heating rate in each layer can be linearly decomposed into individual terms given in (6).
Note that because the units of all terms in (6) have been converted from K/s (degree per second) to W/m2, they can be summed up vertically without changing their physical meanings. For example, the vertical summation of \({\Delta ^{(WV)}}{Q_j}\) from the lowest to the highest atmospheric layers corresponds to the perturbation in the net radiative heating rate by the atmosphere due to the changes in atmospheric water vapor. Therefore, we can reduce the number of layers in our discussions to a few selected layers by adding the terms vertically. We have divided the atmospheric column broadly into six layers in presenting the results of (3) and (4): one for the boundary layer (1000–925 hPa), two for the lower troposphere (925–800 hPa and 800–600 hPa), one for the middle troposphere (600–400 hPa), and two for the upper troposphere (400–250 hPa and 250–150 hPa). We will explain geopotential height anomalies at the interface levels (i.e., 925, 800, 600, 400, 250, and 150 hPa) from sea-level pressure anomalies and the layer temperature resulted according to (1).