Central Asia is experiencing an increase in the occurrence of extreme flood events while still featuring in scarce precipitation. These extreme events are governed by weather systems of various scales as well as complex terrain effects. It is therefore difficult to analyze and forecast precipitation with traditional methods. In this study, an approach is designed by decomposing water vapor flux divergence (Qvall) under model's terrain following coordinates following the Helmholtz theorem. The traditionally used but important variable for depicting anomalous precipitation, Qvall, is partitioned into three components, i.e., flow divergence (Qvdiv), moisture advections by large-scale motion (QvL) and by sub-synoptic motion (QvS). Applications to a blizzard event occurred over Ili Valley during 30 Nov to 1 Dec 2018 is performed as an example. The eastward water vapor flux brings moisture from Balkhash Lake to Ili Valley during the entire precipitation event, while Qvall fails to indicate the initiation and weakening of precipitation. Temporal and spatial evolutions of the three decomposed components are analyzed with comparison to precipitation on both windward slope and lee side. Qvdiv with dominant magnitude always converges no matter precipitation intensifies or weakens, and therefore leads to false moisture aggregation signals in Qvall. The other two components, QvL and QvS, converge prior to precipitation and diverge or weaken before the event ends. Besides, QvL performs better than QvS on windward slope and shows moisture transport from upper levels while QvS achieves better on lee side. Future applications of this flow decomposition approach may thus extend to improving precipitation predictions with negligible computational costs over complex topography in other places around the world.