2015/04/27
题目：Observed changes of global and western Pacific precipitation associated with global warming SST mode and mega‑ENSO SST mode 原文：Kim-CD-2015-fv75_fn67_fp53.pdf
笔记：本文对近几十年来全球尤其是西太平洋降水的变化进行了研究，并试图分离由于人类活动及自然变率对这一变化的相对贡献。在过去的几十年里（1980年以来），研究表明无论是陆地、海洋，以及具体到全球季风的降水，总体上都是在增加的；而且，气候模式普遍预估未来全球降水也是增加的，当然这存在很大的不确定性；最近有研究指出，过去几十年降水的增加，不仅与人类活动造成的全球变暖有关，也与自然变率有关。作者的研究就是为了定量描述自然变率和人类活动对降水变化的贡献。
作者首先对3年滑动平均降水和SST做SVD分解，发现与降水场相关的海温第一模态表现出decadal ENSO的pattern(Wang et al. 2013).但是这并不能够区分出人类活动和自然变率。随后作者对Pr、SST以及垂直积分的水汽通量散度(convergence of vertical integrated moisture flux, CVIMF)进行MVEOF，发现第一模态对应全球变暖的长期趋势信号，第二模态对应类似PDO或mega-ENSO的信号，作者认为这样能够很好的对人类活动影响和自然变率进行区分。
随后作者对CVIMF动力作用项和热力作用项进行诊断，发现无论是人类活动作用还是自然变化作用，通过动力作用（辐散辐合）去影响CVIMF都是主要的，进一步地，作者仿照风场散度对势函数和辐散风的定义，对CVIMF的动力项求势函数和辐散风通量，以求锁定不同变率对应的水汽通量散度的源汇区：结果发现人类活动倾向于在太平洋出现三极型响应，西太和东太是汇区，而中太是源区；自然变率则倾向于形成西太汇而东太源的情况。
接下来，作者又用雷诺平均的方法，诊断定常波（stationary part）和瞬变波（eddy part）对势函数和辐散风通量的贡献，发现定常波在其中起到主要贡献。具体到西北太平洋近几十年降水的变化，作者认为是decadal ENSO的La Nina位相和人类强迫引起的印度洋增强的西风输送所引起。
总之，作者认为利用MVEOF是一种很好的区分人类强迫和自然变率的方法并深入分析了伴随这两种模态的水汽输送通量散度的变化，并试图对西太地区降水的增多进行解释。
亮点和借鉴：
1.这是一篇非常好的利用观测资料进行深入的动力学诊断的文章，利用MVEOF工具巧妙区分了人类强迫和自然变率的影响，进而深入利用高超的动力学分析工具去考察人类强迫和自然变率的对降水变化的贡献作用。环环相扣，渐入佳境。
2.MVEOF的方法，非同时SVD的方法在区分信号、联系因果的作用
3.CVIMF的诊断分析方法，很好的动力学诊断工具。
问题（2015.4.29邮件作者-2015.5.8收到回复）：
1.MVEOF的区分终归是线性的，人类活动对ENSO的影响又该如何表现出来？或者说，这一影响存在于第二个模态中么？
2.为什么要用三年滑动平均来做MVEOF？
3.定常扰动和瞬变扰动是如何区分的？
作者解答
Q1. I am impressed to see the power of MVEOF method to separate the effects of anthropogenic forcing and natural variability. However, since the anthropogenic forcing may also influence the natural variability such as ENSO (Vecchi et al., 2008), is that possible to see this kind of effect in the MVEOF2? * We tried to separate natural variability and anthropogenic forcing, eventually we used the third variable, moisture flux covergence, which may be linked to the dynamic and thermodynamic aspects for climate variability, in using MVEOF. As you know, MVEOFS qurantees independency with orthogonality of base functions. However, as Vecchi et al mentioned, we have assumption: although these two modes are independent, actually ENSO mode has a low-frequency variability such as Mege-ENSO, IPO-related one, which may be associated with anthropogenic component. Q2. I don’t know why 3-year running mean is used to analyze this phenomenon. What’s the advantage than using primitive annual mean or 5-year or 7-year running mean? * We attempted various cases in term of averaging time period, to remove interannual variability such as ENSO. Three-year running average is sufficient to show this and loss of data is relatively small. The advantage of running mean seems that we can show a long-term or low-frequency variabilities. Q3. How to separate stationary and eddy parts of potential function and divergent component of moisture flux? Has a 3-7 day band-pass filter method been used to acquire the eddy parts? * Actually eddy part was estimated with perturbation from annual mean. It has many sub-annual components such as seasonal cycle, sub-seasonal oscillation, high-frequency synoptic eddy, and so on. So my student Byung-Hee Kim (see CC email address), as a co-author has analyzed with separarting various components including eddy component only.