Self-organizing Maps Nodes Research Articles

Development of an Evaluation Method for Precipitation Particle Types by Using Disdrometer Data

Abstract A method for evaluating precipitation particle types from the size–fall velocity data observed by disdrometers by combining an expectation–maximization (EM) algorithm and a self-organizing map (SOM) was developed. An EM algorithm was used to estimate the particle size–fall velocity relationships according to previous work, and the SOM map was used to classify the relationships into graupel or snow categories. The method was applied for snowfall data observed by the volume scanning video disdrometer at a site in Sapporo, Japan, and the daily graupel-to-solid precipitation ratio (G/S ratio) from 1 December 2017 to 28 February 2018 was obtained. The sensitivity of the results to different SOM configurations, SOM node groupings, and training datasets was also tested. The G/S ratio obtained from the simulation data of a meteorological model was compared as an example of the application of the product created by the new method. Significance Statement A method using machine learning for evaluating the precipitation particle types from size and fall velocity data observed by a disdrometer was developed. This method was applied to classify data observed in Sapporo, Japan, in a single winter into graupel and snow, and the daily graupel-to-solid precipitation ratio was obtained. The daily graupel-to-solid precipitation ratio was compared with that simulated by a meteorological model as an example of the application of disdrometer data.

Near‐surface wind variability in Prydz Bay and Amery Ice Shelf region, East Antarctica: A four‐decade SOM analysis

AbstractNear‐surface wind fields in the Prydz Bay and Amery Ice Shelf region of East Antarctica play a crucial role in the formation and variability of Antarctic Bottom Water, a cold, dense water mass that sinks and spreads across the deep ocean basins influencing ocean circulation and modulating earth's climate system. This study investigates the primary modes of variability of these wind fields using the self‐organizing map (SOM) method and data from the latest version of the European Centre for Medium‐Range Weather Forecasts (ECMWF) ReAnalyses (ERA5), spanning four decades from 1979 to 2020. While the wind field climatology, characterized by small seasonal variation, is dominated by katabatic and large‐scale forcing, the spatial patterns of the primary variability modes are mainly influenced by synoptic system activities. The overall trend in annual wind speed anomalies is positive across the study region, with the exception of the southwestern part and central Prydz Bay. However, significant trends are observed in only two out of nine SOM nodes (nodes 4 and 9), which collectively explain less than 30% of the averaged trends over the region. The interannual variability in the seasonal occurrences of certain nodes is linked to several well‐known climate modes, including the El Niño–Southern Oscillation, the Southern Annular Mode and Zonal Wavenumber 3. Our results provide a reference for forecasting the occurrence frequency of specific patterns, which could help mitigate the impact of extreme wind events through improved forecasting.

Characterizing Mesoscale Cellular Convection in Marine Cold Air Outbreaks With a Machine Learning Approach

AbstractDuring marine cold‐air outbreaks (MCAOs), when cold polar air moves over warmer ocean, a well‐recognized cloud pattern develops, with open or closed mesoscale cellular convection (MCC) at larger fetch over open water. The Cold‐Air Outbreaks in the Marine Boundary Layer Experiment provided a comprehensive set of ground‐based in situ and remote sensing observations of MCAOs at a coastal location in northern Norway. MCAO periods that unambiguously exhibit open or closed MCC are determined. Individual cells observed with a profiling Ka‐band radar are identified using a watershed segmentation method. Using self‐organizing maps (SOMs), these cells are then objectively classified based on the variability in their vertical structure. The SOM nodes contain some information about the location of the cell transect relative to the center of the MCC. This adds classification noise, requiring numerous cell transects to isolate cell dynamical information. The SOM‐based classification shows that comparatively intense convection occurs only in open MCC. This convection undergoes an apparent lifecycle. Developing cells are associated with stronger updrafts, large spectrum width, larger amounts of liquid water, lower surface precipitation rates, and lower cloud tops than mature and weakening cells. The weakening of these cells is associated with the development of precipitation‐induced cold pools. The SOM classification also reveals less intense convection, with a similar lifecycle. More stratiform vertical cloud structures with weak vertical motions are common during closed MCC periods and are separated into precipitating and non‐precipitating stratiform cores. Convection is observed only occasionally in the closed MCC environment.

Identifying wind regimes near Kuwait using self-organizing maps

Optimization of wind energy integration requires knowing the relationship between weather patterns and winds they cause. For a region with less-studied weather such as the Middle East, climatology becomes more vital. The Shagaya Renewable Energy Park in development in Kuwait experiences regional wind regimes that affect wind power production. Weather Research and Forecasting (WRF) model output allowed investigation into the weather regimes most likely to impact Shagaya Park. The self-organizing maps (SOMs) machine-learning method clustered the WRF output into six primary weather regimes experienced by the Middle East. According to the wind regimes mapped by the SOM, two of the six regimes have average wind speeds of approximately 9.9 and 8.6 m s−1 at 80 m near Shagaya Park, as well as wind speed and estimated wind power distributions that are more favorable to wind power production in Kuwait. One regime depicts a strong northwesterly wind called the summer shamal, and the second is associated with strong westerlies. Regimes less favorable for Kuwaiti wind power production are represented by the remaining four SOM nodes: local weak southeasterlies, an African nocturnal low-level jet, a daytime planetary boundary layer, and local northwesterlies from autumn to spring. The remaining four SOM nodes have average wind speeds of 5.7–7.2 m s−1 and wind speed and estimated wind power distributions which indicate regimes less favorable for wind power production in Kuwait.

Analyzing Wintertime Extreme Winds over Türkiye and Their Relationships with Synoptic Patterns Using Cluster Analysis

This study investigates the patterns of extreme winds and the correlation between synoptic patterns in Türkiye throughout the winter season, using the cluster analysis technique. We utilized the k-means algorithm to detect the surface patterns of extreme winds. Additionally, we deployed the Self-Organizing Map (SOM) technique to identify clusters of geopotential height at the 500 hPa level, average temperature at the 850 hPa level, and mean sea level pressure. We adopted the dataset from the New European Wind Atlas (NEWA) project for analyzing surface-level weather conditions and the ERA5 datasets for studying upper-level weather conditions. The k-means algorithm identifies six distinct clusters when applied to the ground-level data in Türkiye. These clusters are predominantly located around the Taurus Mountain ranges, which stretch in an east-west and northeastern direction along the Black Sea coast. The formation of these clusters is controlled by the characteristics of the land and its physical features. The higher-level clusters, consisting of nine SOM nodes, are unaffected by terrain and weather systems, which are characteristic of the macro-Mediterranean climate. These clusters are detected in the Eastern Mediterranean, Black Sea, and inner Aegean areas, emphasizing the impact of topography on surface patterns.

Domain Incremental Object Detection Based on Feature Space Topology Preserving Strategy

Object detection with the capacity to incrementally adapt to new domains is a crucial yet relatively under-explored research topic. The <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">catastrophic forgetting</i> problem presents a significant challenge to achieve this goal, where the model’s performance improves quickly in new conditions but deteriorates sharply in old ones after several incremental learning sessions. Drawing on recent discoveries in visual memories of the human brain, we introduce the Topology-Preserving Domain Incremental Object Detection (TP-DIOD) approach, which aims to address the catastrophic forgetting problem by extracting the topological structure of the feature space learned by the Convolutional Neural Network (CNN) model and preserving this topology during the subsequent incremental learning sessions. Specifically, we model the feature space topology using the self-organizing map (SOM) and construct an anchor image set based on the centroid vectors of the SOM nodes to memorize the feature space topology. We then develop the anchor loss function to penalize the topological changes of the feature space during the subsequent incremental learning sessions. Experimental evaluations on two sets of datasets demonstrate the effectiveness of the proposed TP-DIOD method in mitigating the catastrophic forgetting problem and achieving high accuracy on both old and new domain datasets.

Synoptic circulation patterns of urban floods for the city of Hyderabad

AbstractSynoptic fingerprinting of flood driving rainfall events over the Hyderabad city, India were assessed through an unsupervised machine learning technique—self‐organizing maps (SOM) for the period 1979–2020. Flood dates were identified using nonscientific data sources, and then large‐scale climate variables, that is, integrated vapour transport (IVT) and geopotential height at 500 hPa (Z500) along with rainfall at daily scale were considered. SOM was applied on IVT and results indicated large spatial scale‐ and season‐specific mechanisms in addition to local circulation systems—this highlights SOM's ability to cluster circulation mechanisms as per the relevance. The rainy seasons, that is, monsoon and postmonsoon showed well‐known moisture pathways. While synoptic systems can be easily linked with monsoon and postmonsoon mechanisms, it is interesting to note that the local systems exist even during the monsoon and postmonsoon season in addition to the summer season. Spatial patterns of Z500 corresponding to SOM nodes of IVT showed the presence of low‐pressure systems (LPSs), tropical cyclones and trough systems. Further analysis was performed, and a total of 35 atmospheric rivers (ARs) were identified. Plots of back trajectories for the selected AR events indicated that Arabian Sea is the main source of moisture in the monsoon season followed by Bay of Bengal for postmonsoon and tropical cyclones in summer season events. Additionally, moisture from land is also observed as another source of moisture. Back‐trajectory analysis further indicated AR‐LPS interaction during the monsoon and ARs feeding moisture to LPSs. Overall, well‐known and season‐specific synoptic systems that brought significant rainfall to the study region were identified by the SOM. The proposed framework is adaptable for different locations, and as floods in cities across the globe are on the rise, these findings will have a prominent role in urban flood forecasting and management.

Impact of synoptic circulation patterns on renewable energy-related variables over China

This study investigates the influence of atmospheric circulation patterns on the day-to-day variabilities of renewable energy- and demand-related variables (temperature, wind speed, and sunshine duration) over China in winter. Circulation patterns are classified by self-organizing maps (SOMs), and six categories are obtained, including two opposite northwest-southeast dipole modes, two opposite widespread monopoles, and two modes with cyclonic or anticyclonic circulation over northeastern China.It is found that the position of temperature anomalies has a clear association with the anomalous circulation, the high wind speed is generally situated in the south or west sector of cyclonic circulations, while the long sunshine duration is associated with northerly airflow and small cloud cover on the west of cyclonic circulations. Additionally, the cyclonic flow over northeastern China offers the optimum weather condition for wind and solar resources, whereas the Yunnan-Guizhou Plateau under the control of a positive-negative northwest-southeast dipole experiences the worst condition when the temperature anomaly is negative. Finally, the variables are simulated by circulation patterns through a multi-linear regression (MLR) model for 42 winters during 1979–2021. Specifically, the MLR model built by 56 SOM nodes well reproduces the day-to-day variabilities and trends of these variables.

Adaptive Self-Organizing Map Using Optimal Control

The self-organizing map (SOM), which is a type of artificial neural network (ANN), was formulated as an optimal control problem. Its objective function is to minimize the mean quantization error, and the state equation is the weight updating equation of SOM. Based on the objective function and the state equations, the Hamiltonian equation based on Pontryagin’s minimum principle (PMP) was formed. This study presents two models of SOM formulated as an optimal control problem. In the first model, called SOMOC1, the design is based on the state equation representing the weight updating equation of the best matching units of the SOM nodes in each iteration, whereas in the second model, called SOMOC2, it considers the weight updating equation of all the nodes in the SOM as the state updating equation. The learning rate is treated as the control variable. Based on the solution of the switching function, a bang-bang control was applied with a high and low learning rate. The proposed SOMOC2 model performs better than the SOMOC1 model and conventional SOM as it considers all the nodes in the Hamiltonian equation, and the switching function obtained from it is influenced by all the states, which provides one costate variable for each. The costate determines the marginal cost of violating the constraint by the state equations, and the switching function is influenced by this, hence producing a greater improvement in terms of the mean quantization error at the final iteration. It was found that the solution leads to an infinite order singular arc. The possible solutions for the suitable learning rates during the singular arc period are discussed in this study.

A Self-Organizing Maps Analysis of Wintertime North Pacific Jet Stream Variability

Abstract Previous research regarding the intraseasonal variability of the wintertime Pacific jet has employed empirical orthogonal function (EOF)/principal component (PC) analysis to characterize two leading modes of variability: a zonal extension or retraction and a ∼20° meridional shift of the jet exit region. These leading modes are intimately tied to the large-scale structure, sensible weather phenomena, and forecast skill in and around the vast North Pacific basin. However, variability within the wintertime Pacific jet and the relative importance of tropical and extratropical processes in driving such variability, is poorly understood. Here, a self-organizing maps (SOM) analysis is applied to 73 Northern Hemisphere cold seasons of 250-hPa zonal winds from the NCEP–NCAR reanalysis data to identify 12 characteristic physical jet states, some of which resemble the leading EOF Pacific jet patterns and combinations of them. Examination of teleconnection patterns such as El Niño–Southern Oscillation (ENSO) and the Madden–Julian oscillation (MJO) provide insight into the varying nature of the 12 SOM nodes at inter- and intraseasonal time scales. These relationships suggest that the hitherto more common EOF/PC analysis of jet variability obscures important subtleties of jet structure, revealed by the SOM analysis, which bear on the underlying physical processes associated with Pacific jet variability as well as the nature of its downstream impacts.

Linking Synoptic Patterns to Cloud Properties and Local Circulations Over Southeastern Texas

AbstractThis study classifies meteorological regimes in the southeastern Texas region to identify environmental conditions that favor sea‐breeze induced convection. The classification is accomplished using a Self‐Organizing Map (SOM) approach. We applied SOM to 10 years of 700‐hPa geopotential height anomalies during the summer months from reanalysis data to distinguish three dominant synoptic regimes, with a continuum of transitional states between those. The primary regimes include: (a) a pre‐trough regime associated with a synoptic trough, (b) a post‐trough regime with upper‐level northerly flow, and (c) an anticyclonic regime within the westward extent of the Bermuda High. We project the data from the Geostationary Operational Environmental Satellite and the Next Generation Weather Radar system onto each SOM node to investigate the characteristics of cloud and precipitation properties in different regimes. When southeastern Texas is positioned to the southwest quadrant of a maritime high pressure system, an increased cloud frequency is observed over the region during the afternoon hours due to significant moisture advection. A confluence of synoptic southerly flow and sea‐breeze circulation commonly occurs in this regime. When a high pressure system is over southeastern Texas, the area is dominated by large‐scale subsidence with weak pressure gradients and moderate precipitable water vapor. This weak synoptic forcing is favorable for the formation of a sea‐breeze circulation. This is confirmed by an enhanced onshore flow and a decreased temperature at the surface in the early afternoon, as well as a sharp increase in radar echo top height.

Extreme Precipitation in the Great Lakes Region: Trend Estimation and Relation With Large-Scale Circulation and Humidity

Extreme precipitation contributes to widespread impacts in the U.S. Great Lakes region, ranging from agricultural losses to urban floods and associated infrastructure costs. Previous studies have reported historical increases in the frequency of extreme precipitation in the region and downscaled model projections indicate further changes as the climate system continues to warm. Here, we conduct trend analysis on the 5 km NOAA NClimDiv data for the U.S. Great Lakes region using both parametric (Ordinary Least Squares) and non-parametric methods (Theil-Sen/Mann-Kendall) and accounting for temporal autocorrelation and field significance to produce robust estimates of extreme precipitation frequency trends in the region. The approaches provide similar overall results and reflect an increase in extreme precipitation frequency in parts of the U.S. Great Lakes region. To relate the identified trends to large scale drivers, a bivariate self-organizing map (SOM) is constructed using standardized values of 500 hPa geo-potential height and 850 hPa specific humidity obtained from the ECMWF ERA-5 reanalysis. Using a Monte Carlo approach, we identify six SOM nodes that account for only 25.4% of all days, but 50.5% of extreme precipitation days. Composites of days with and without extreme precipitation for each node indicate that extreme events are associated with stronger features (height gradient and background humidity) than their non-extreme counterparts. The analysis also identifies a significant increase in the frequency of one SOM node often associated with extreme precipitation (accounting for 8.5% of all extreme precipitation days) and a significant increase in the frequency of extreme precipitation days relative to all days across the six extreme precipitation nodes collectively. Our results suggest that changes in atmospheric circulation and related moisture transport and convergence are major contributors to changes in extreme precipitation in the U.S. Great Lakes region.

Unsupervised classification of simulated magnetospheric regions

Abstract. In magnetospheric missions, burst-mode data sampling should be triggered in the presence of processes of scientific or operational interest. We present an unsupervised classification method for magnetospheric regions that could constitute the first step of a multistep method for the automatic identification of magnetospheric processes of interest. Our method is based on self-organizing maps (SOMs), and we test it preliminarily on data points from global magnetospheric simulations obtained with the OpenGGCM-CTIM-RCM code. The dimensionality of the data is reduced with principal component analysis before classification. The classification relies exclusively on local plasma properties at the selected data points, without information on their neighborhood or on their temporal evolution. We classify the SOM nodes into an automatically selected number of classes, and we obtain clusters that map to well-defined magnetospheric regions. We validate our classification results by plotting the classified data in the simulated space and by comparing with k-means classification. For the sake of result interpretability, we examine the SOM feature maps (magnetospheric variables are called features in the context of classification), and we use them to unlock information on the clusters. We repeat the classification experiments using different sets of features, we quantitatively compare different classification results, and we obtain insights on which magnetospheric variables make more effective features for unsupervised classification.

Spectrum of Near-Storm Environments for Significant Severe Right-Moving Supercells in the Contiguous United States

AbstractProximity soundings have long been used to explore how the vertical structure of temperature, humidity, and winds influence convective storms and their associated hazards. In severe thunderstorm research and forecasting, convective parameters are often used to summarize certain characteristics of the sounding. While extremely useful, these parameters are unable to describe the rich complexity that is readily apparent in hodographs and skew T–logp diagrams. Motivated by a desire to retain more of these details, the present study uses self-organizing maps (SOMs) to group soundings based on their full vertical structure. The analysis makes use of a sample of more than 10 000 model proximity soundings for right-moving supercells associated with tornadoes and significant severe hail and straight-line winds in the contiguous United States (CONUS). Separate SOMs are developed for the wind and thermodynamic profiles, each with 3 × 3 nodes, resulting in a set of nine hodographs and nine skew T–logp diagrams that broadly represent the spectrum of near-storm environments for significant severe right-moving supercells in the CONUS. Both SOMs are shown to provide a good representation of the variability in key convective parameters, although, for the thermodynamic SOM, variations in LCL heights and midlevel lapse rates are somewhat limited. Based on the soundings assigned to them, the SOM nodes are characterized in terms of their associated hazards, their relationship with storm mode and mesocyclone strength, and their spatial and temporal variability. Potential applications of the SOMs in severe weather forecasting and idealized numerical simulations are also highlighted.

An unsupervised learning approach to identifying blocking events: the case of European summer

Abstract. Atmospheric blocking events are mid-latitude weather patterns, which obstruct the usual path of the polar jet streams. They are often associated with heat waves in summer and cold snaps in winter. Despite being central features of mid-latitude synoptic-scale weather, there is no well-defined historical dataset of blocking events. Various blocking indices (BIs) have thus been suggested for automatically identifying blocking events in observational and in climate model data. However, BIs show significant regional and seasonal differences so that several indices are typically applied in combination to ensure scientific robustness. Here, we introduce a new BI using self-organizing maps (SOMs), an unsupervised machine learning approach, and compare its detection skill to some of the most widely applied BIs. To enable this intercomparison, we first create a new ground truth time series classification of European blocking based on expert judgement. We then demonstrate that our method (SOM-BI) has several key advantages over previous BIs because it exploits all of the spatial information provided in the input data and reduces the dependence on arbitrary thresholds. Using ERA5 reanalysis data (1979–2019), we find that the SOM-BI identifies blocking events with a higher precision and recall than other BIs. In particular, SOM-BI already performs well using only around 20 years of training data so that observational records are long enough to train our new method. We present case studies of the 2003 and 2019 European heat waves and highlight that well-defined groups of SOM nodes can be an effective tool to diagnose such weather events, although the domain-based approach can still lead to errors in the identification of certain events in a fashion similar to the other BIs. We further test the red blocking detection skill of SOM-BI depending on the meteorological variable used to study blocking, including geopotential height, sea level pressure and four variables related to potential vorticity, and the 500 hPa geopotential height anomaly field provides the best results with our new approach. We also demonstrate how SOM-BI can be used to identify different types of blocking events and their associated trends. Finally, we evaluate the SOM-BI performance on around 100 years of climate model data from a pre-industrial simulation with the new UK Earth System Model (UKESM1-0-LL). For the model data, all blocking detection methods have lower skill than for the ERA5 reanalysis, but SOM-BI performs noticeably better than the conventional indices. Overall, our results demonstrate the significant potential for unsupervised learning to complement the study of blocking events in both reanalysis and climate modelling contexts.

Low-level Updraft Intensification in Response to Environmental Wind Profiles

AbstractSupercell storms can develop a “dynamical response” whereby upward accelerations in the lower troposphere amplify as a result of rotationally induced pressure falls aloft. These upward accelerations likely modulate a supercell’s ability to stretch near-surface vertical vorticity to achieve tornadogenesis. This study quantifies such a dynamical response as a function of environmental wind profiles commonly found near supercells. Self-organizing maps (SOMs) were used to identify recurring low-level wind profile patterns from 20,194 model-analyzed, near-supercell soundings. The SOM nodes with larger 0–500 m storm-relative helicity (SRH) and streamwise vorticity (ωs) corresponded to higher observed tornado probabilities. The distilled wind profiles from the SOMs were used to initialize idealized numerical simulations of updrafts. In environments with large 0–500 m SRH and large ωs, a rotationally induced pressure deficit, increased dynamic lifting, and a strengthened updraft resulted. The resulting upward-directed accelerations were an order of magnitude stronger than typical buoyant accelerations. At 500 m AGL, this dynamical response increased the vertical velocity by up to 25 m s–1, vertical vorticity by up to 0.2 s–1, and pressure deficit by up to 5 hPa. This response specifically augments the near-ground updraft (the midlevel updraft properties are almost identical across the simulations). However, dynamical responses only occurred in environments where 0–500 m SRH and ωs exceeded 110 m2 s–2 and 0.015 s–1, respectively. The presence vs. absence of this dynamical response may explain why environments with higher 0–500 m SRH and ωs correspond to greater tornado probabilities.

Kohonen map-wise regression applied to interval data

Kohonen maps, also known as self-organizing maps, is a powerful clustering method which groups data using multiple nodes that converge to clusters. Therefore, it is not necessary to know a priori the exact number of clusters in the data. This paper proposes a clusterwise regression method that combines self-organizing maps and a parametrized linear regression approach for interval data in the framework of Symbolic Data Analysis. The linear regression approach adapts itself to use the best set of points inside intervals to build the regression model, generalizing all approaches that select points of interest, such as center, range, minimum and maximum, to provide linear regression for interval data. Here, self-organizing map nodes are responsible for detecting local linear regression structures and the parametrized linear regression builds local regression models, one for each node. Finally, outputs are given as weighted averages of the local model predictions. Since our model fits the linear models using the map nodes, it avoids weak performances that might arise when clusters are not easily separable. As another contribution, we provide a way to extract predictions from an existing clusterwise regression method. Experiments with synthetic and real data sets show the usefulness of the proposed method.

Synoptic mode of Antarctic summer sea ice superimposed on interannual and decadal variability

In contrast to decreased Arctic sea ice extent, Antarctic sea ice extent shows a somewhat increased trend. There is a large interannual variability of Antarctic sea ice, especially in the Pacific sector of the Southern Ocean. The change and variability of Antarctic sea ice in synoptic timescales in the recent decades remain unclear. We identify synoptic modes of variability of Antarctic summer sea ice by applying the Self Organizing Map (SOM) technique to daily sea ice concentration data for the period 1979–2018. Nearly 40% of the variability is characterized by opposite changes between sea ice cover in the Bellingshausen, Amundsen and western Ross Seas and in the rest of the Antarctic seas, and another 30% by meridional asymmetry in the Weddell, Amundsen, and Ross Seas. Most of these spatial patterns may be explained by the dynamics and thermodynamic processes associated with anomalous atmospheric circulations related to the Southern Annular Mode (SAM) with a structure of strong zonal asymmetry. The interannual variability of the sea ice modes appears to have little connection to SAM, and only a weak relation to ENSO. The annual frequencies of SOM node occurrences also show a great decadal variability. Node 9 appears mainly prior to 1990; while node 1 occurs mainly after 1990. The decadal variability of nodes 1 and 9 is associated with the asymmetrical SAM, which results from two wavetrains excited over northern Australia and the southeastern Indian Ocean. These results further highlight the importance of understanding the role of southern mid-to-high latitude atmospheric intrinsic variability in predicting Antarctic summer sea ice variations from synoptic to decadal timescales.

Projection and possible causes of summer precipitation in eastern China using self-organizing map

Self-organizing map (SOM) is used to simulate summer daily precipitation over the Yangtze–Huaihe river basin in Eastern China, including future projections. SOM shows good behaviors in terms of probability distribution of daily rainfall and spatial distribution of rainfall indices, as well as consistency of multi-model simulations. Under RCP4.5 Scenario, daily rainfall at most sites (63%) is projected to shift towards larger values. For the early 21st century (2016–2035), precipitation in the central basin increases, yet decreases occur over the middle reaches of the Yangtze River as well as a part of its southeast area. For the late 21st century (2081–2100), the mean precipitation and extreme indices experience an overall increase except for a few southeast stations. The total precipitation in the lower reaches of the Yangtze River and in its south area is projected to increase from 7% at 1.5 °C global warming to 11% at 2 °C, while the intensity enhancement is more significant in southern and western sites of the domain. A clustering allows to regroup all SOM nodes into four distinct regimes. Such regional synoptic regimes show remarkable stability for future climate. The overall intensification of precipitation in future climate is linked to the occurrence-frequency rise of a wet regime which brings longitudinally closer the South Asia High (eastward extended) and the Western Pacific Subtropical High (westward extended), as well as the reduction of a dry pattern which makes the two atmospheric centers of action move away from each other.

Improved Groundwater System Characterization and Mapping Using Hydrogeophysical Data and Machine-Learning Workflows

The Australian machine-learning workflows apply fusion, clustering, and estimation operations to hydrogeophysical data for deriving hydrostratigraphic units (HSUs). Data fusion is performed by training a self-organizing map (SOM) with these data. The application of Davies-Bouldin criteria to K-means clustering of SOM nodes determines the number and location of HSUs. Estimation is handled by iterative least-squares minimization of the SOM quantization and topographical errors. Two workflows provide 3D characterization of HSUs (and related attributes) from different hydrogeophysical data (measured, derived, interpolated, and estimated values) sets. In Workflow 1, the SOM learns to recognize relationships among a subset of borehole geophysical and hydrogeologic data. Using the data-fusion approach described above, the missing hydrological data are estimated using these learned relationships and HSUs determined at borehole sample locations resulting in a low lateral density and high vertical density spatial distribution. Variogram modeling of the regional field data and HSU estimates is undertaken to evaluate the spatial statistical structure of selected attributes. In workflow 2, the learned relationships between borehole data and the more spatially extensive AEM conductivity model are used to estimate the key attributes and HSUs at a number of locations away from the borehole. The AEM conductivity profile at a number of random locations are mapped to the SOM network and estimation performed to arrive at a set of continuous HSUs with high lateral density and medium vertical density (based on m-layer modelled structure). Performance metrics and validation are used to test each step of both workflows.