Interannual rainfall variability and SOM-based circulation classification

Abstract

synoptic variables over a particular location; even though they are used in the development of the SOM their influence, however, diminishes with the size of the SOM spatial domain. The influence of the SOM domain size, the choice of SOM atmospheric variables and grid-point explanatory variables on the levels of explained variance, is consistent with the general understanding of the dominant processes and atmospheric variables that affect rainfall variability at a particular location. Keywords SOM · Synoptic circulation · Rainfall variability · Southern AfricaThe relationship between synoptic circulation and local expressions of climate is the foundation of synoptic climatology and traditionally underpins the process of interpretative weather forecasting. The basic premise of synoptic climatology is that the large-scale atmospheric circulation exerts some control over weather and related environmental phenomena at the Earth's surface, thereby providing a level of deterministic forcing which influences the local weather. This directly leads to the use of atmospheric circulation patterns as a means to express drivers of weather (and climate) variability. Self-Organizing Maps (SOM, Kohonen 2001) is a pattern clustering method that is used as one of the methods to derive synoptic circulation types in this and other contexts, including statistical downscaling (e.g. Hewitson and Crane 2006;Yin et al. 2011;Ohba et al. 2016) and process-based validation of GCMs (e.g. Brown et al. 2010;Finnis et al. 2009;Higgins and Cassano 2010).In the application of SOM to study climate variability, the evolution of synoptic drivers of weather is typically represented by progression through classes of daily synoptic states identified as SOM nodes. That progression Abstract Self-Organizing Maps (SOM) based classifications of synoptic circulation patterns are increasingly being used to interpret large-scale drivers of local climate variability, and as part of statistical downscaling methodologies. These applications rely on a basic premise of synoptic climatology, i.e. that local weather is conditioned by the large-scale circulation. While it is clear that this relationship holds in principle, the implications of its implementation through SOM-based classification, particularly at interannual and longer time scales, are not well recognized. Here we use a SOM to understand the interannual synoptic drivers of climate variability at two locations in the winter and summer rainfall regimes of South Africa. We quantify the portion of variance in seasonal rainfall totals that is explained by year to year differences in the synoptic circulation, as schematized by a SOM. We furthermore test how different spatial domain sizes and synoptic variables affect the ability of the SOM to capture the dominant synoptic drivers of interannual rainfall variability. Additionally, we identify systematic synoptic forcing that is not captured by the SOM classification. The results indicate that the frequency of synoptic states, as schematized by a relatively disa...

Process‐based model evaluation and projections over southern Africa from Coordinated Regional Climate Downscaling Experiment and Coupled Model Intercomparison Project Phase 5 models

“…The second caveat is that the SOM node mean precipitation anomalies in Figure have not been analysed for statistical significance in any way. Other work (e.g., Wolski et al, ) shows that there can be high variance in the local‐scale precipitation response within each SOM node cluster. This would suggest that anomalies as calculated in Figure may indeed be insignificant relative to the underlying variance in precipitation response.…”

A SOM‐based analysis of the drivers of the 2015–2017 Western Cape drought in South Africa

“…It is similar to the domain used by Wolski et al . (2018) to assess rainfall variability over Cape Town based on SOMs. 850 hPa geopotential height is considered the best single predictor field characterizing rainfall variability over Southern Africa (Landman and Goddard, 2002).…”

“…To identify the synoptic circulation patterns over Southern Africa associated with rainfall over the SWC in South Africa, we applied self-organizing maps (SOMs) analysis to ERAINT daily 850 hPa geopotential height datasets using all days from March to October for the period 1979-2017 over a small Southern African domain extending from 0 to 34.5 E and 19.5 to 45 S. This domain is large enough to capture the major synoptic features such as the subtropical anticyclones and the westerly winds that influence rainfall variability over the SWC. It is similar to the domain used by Wolski et al (2018) to assess rainfall variability over Cape Town based on SOMs. 850 hPa geopotential height is considered the best single predictor field characterizing rainfall variability over Southern Africa (Landman and Goddard, 2002).…”

“…While most of these studies focused on the relationship between specific synoptic drivers and rainfall variability in the Western Cape region, Wolski et al . (2018) used circulation classification with self‐organizing maps (SOMs) to quantify rainfall variability that can be explained by the frequency of circulation types over a winter (Cape Town) and summer (Johannesburg) rainfall regime in South Africa. Over Cape Town, the synoptic states represented in the SOM classification of pressure field‐related variables explained between 40 and 60% of the winter (JJA) rainfall interannual variability at a local scale.…”

Understanding global teleconnections to surface air temperatures in Japan based on a new climate classification

“…To overcome the sensitivity to outliers and to preserve the topology of the data, the self-organizing map (SOM) became a popular algorithm among climatologists and meteorologists (e.g., Hewitson and Crane, 2002;Leloup et al, 2007Leloup et al, , 2008Tozuka et al, 2008;Morioka et al, 2010;Liu and Weisberg, 2011;Chattopadhyay et al, 2013;Oettli et al, 2014;Wolski et al, 2018), with remarkable success. SOM is a type of artificial neural network, with an unsupervised learning for the training (Kohonen, 1982(Kohonen, , 2001).…”