Evaluation of Machine Learning Classifiers for Predicting Deep Convection

Ukkonen, Peter; Mäkelä, Antti

doi:10.1029/2018ms001561

Cited by 53 publications

(52 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is increasingly common in meteorology to use machine learning approaches for identifying patterns in the atmosphere using large amounts of historical data (Dueben & Bauer, ; Scher & Messori, ; Ukkonen & Mäkelä, ; Weyn et al, ). This approach, of extracting the underlying physical relationships in the atmosphere from data, opens an opportunity to explore new algorithms that optimize the output based on different verification metrics.…”

Section: Introductionmentioning

confidence: 99%

Optimization of Deep Learning Precipitation Models Using Categorical Binary Metrics

Larraondo

Renzullo

Dijk

et al. 2020

J Adv Model Earth Syst

View full text Add to dashboard Cite

This work introduces a methodology for optimizing neural network models using a combination of continuous and categorical binary indices in the context of precipitation forecasting. Probability of detection and false alarm rate are popular metrics used in the verification of precipitation models. However, machine learning models trained using gradient descent cannot be optimized based on these metrics, as they are not differentiable. We propose an alternative formulation for these categorical indices that are differentiable and we demonstrate how they can be used to optimize the skill of precipitation neural network models defined as a multiobjective optimization problem. To our knowledge, this is the first proposal of a methodology for optimizing weather neural network models based on categorical indices. Plain Language Summary Deep neural networks have recently demonstrated great versatilityand an unprecedented capacity to model complex problems. In weather modeling, these algorithms have been applied to solve different problems. This is a promising area of research, given the availability of large volumes of weather data and increasingly powerful computers. Neural network models can learn to solve problems based on a metric, which the model tries to optimize. However, the quality of weather models is measured using a large variety of metrics, which can be a challenge when choosing which metric the model should optimize. In the case of precipitation, categorical binary metrics are a popular choice to assess the quality of a model. These metrics reduce precipitation to a "yes" or "no" event, and the results of the predicting model can be compared with the actual observations. This method is simple, yet powerful and a large number of indices and statistics have been developed to assess different aspects of the quality of precipitation models. As precipitation models are commonly assessed using these categorical binary metrics, it would be very convenient to optimize models based on them. Unfortunately, the mathematical nature of these metrics makes them unsuitable for optimizing deep learning models. In this work we present an alternative formulation for these categorical binary indices which can be used to train models. We demonstrate how a deep learning model can be trained to generate better quality precipitation data.

show abstract

Section: Introductionmentioning

confidence: 99%

Optimization of Deep Learning Precipitation Models Using Categorical Binary Metrics

Larraondo

Renzullo

Dijk

et al. 2020

J Adv Model Earth Syst

View full text Add to dashboard Cite

show abstract

“…Multilayer perceptrons (MLP) (Goodfellow et al, 2016) are the most basic form of an artificial neural network. Good results achieved by MLP in predicting storms (Ukkonen and Mäkelä, 2019), they are a natural choice to experiment also in this work. The downside of the method is a large number of hyperparameters including the correct network topology.…”

Section: Classifying Storm Objectsmentioning

confidence: 88%

“…Hanewinkel (2005) conducted a similar study in Germany using artificial neural networks. Artificial neural networks have been used to predict extreme weather in Finland (Ukkonen et al (2017), Ukkonen and Mäkelä (2019)). To summarise, according to various sources, for example, the framework of IPCC (Masson-Delmotte et al, 2018), the impacts of the extreme weather risks can be analyzed by estimating the hazard, vulnerability, and exposure while machine learning techniques are becoming more popular in the task of connecting the natural hazards with the societal impact forecasts (Chen et al, 2008).…”

mentioning

confidence: 99%

Predicting power outages caused by extratropical storms

Tervo¹,

Láng²,

Jung³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. Strong winds induced by extratropical storms cause a large number of power outages especially in highly forested countries such as Finland. Thus, predicting the impact of the storms is one of the key challenges for power grid operators. This article introduces a novel method to predict the storm severity for the power grid employing ERA5 reanalysis data combined with forest inventory. We start by identifying storm objects from wind gust and pressure fields by using contour lines of 15 m s−1 and 1000 hPa respectively. The storm objects are then tracked and characterized with features derived from surface weather parameters and forest vegetation information. Finally, objects are classified with a supervised machine learning method based on how much damage to the power grid they are expected to cause. Random Forest Classifier, Support Vector Classifier, Naive Bayes, Gaussian Processes, and Multilayer Perceptron were evaluated for the classification task, Support Vector Classifier providing the best results.

show abstract

“…This reanalysis has high temporal (every 1 h) and spatial (0.25° of latitude × longitude and 37 vertical pressure levels) resolutions and is available from 1979 to near real‐time. Due to the fine horizontal resolution, it is expected that ERA5 can solve mesoscale structures of the atmospheric systems (Chen et al., 2020; Maranan et al., 2019; Ukkonen et al., 2019). ERA5 provides variables at surface (e.g., SST and precipitation) and pressure levels.…”

Section: Methodsmentioning

confidence: 99%

Iba: The First Pure Tropical Cyclogenesis Over the Western South Atlantic Ocean

et al. 2021

View full text Add to dashboard Cite

Since the beginning of the satellite era, the only documented tropical cyclone over the western South Atlantic Ocean (SAO) has been Catarina, which occurred in March 2004. However, this system resulted from a tropical transition (TT), when a dipole blocking-pattern at the middle-upper levels of the atmosphere contributed to decreasing the vertical wind shear over an extratropical cyclone (McTaggart-Cowan et al., 2006; Pezza & Simmonds, 2005). Fifteen years after Catarina, in March 2019, the Brazilian Navy registered Iba, the first pure tropical cyclogenesis over tropical western SAO (∼16.5°S). Iba is an indigenous name previously established by the Brazilian Navy (NORMAN, 2018). As this system has not yet been described in the literature, it is the purpose of this study. The low frequency of tropical systems over the SAO has been explained by the unfavorable climate conditions for tropical cyclogenesis such as environmental vertical wind shear stronger than 10 m s −1 in the 850-200 hPa layer and sea surface temperature (SST) colder than 26°C (Gray, 1968; Pezza & Simmonds, 2005). On the other hand, the combination of the main environmental predictors for tropical cyclogenesis (absolute vorticity at 850 hPa, relative humidity at 700 hPa, potential intensity in terms of maximum wind speed, and the vertical wind shear at 850-200 hPa layer) in the Genesis Potential Index (GPI; K. A. Emanuel & Nolan, 2004) show some favorable conditions for tropical cyclogenesis near southeastern Brazil (Walsh et al., 2013). McTaggart-Cowan et al. (2013, 2015) also suggest that the SAO climate conditions are most likely to favor weak and strong TT than pure tropical cyclogenesis. The definition of weak and strong TT was introduced by Davis and Bosart (2003, 2004). The weak TT occurs in an environment with a weak upper-level disturbance and moderate lower-level thermal gradients, while the strong TT takes place in the presence of strong upper-level disturbance and strong lower-level thermal gradients (McTaggart-Cowan et al., 2013). Hodges et al. (2017), comparing six reanalysis datasets and IBTrACS (observed data) from 1979 to 2012, showed one tropical cyclone per year over the SAO in the observations and ∼7 cyclones per year in each

show abstract

Evaluation of Machine Learning Classifiers for Predicting Deep Convection

Cited by 53 publications

References 60 publications

Optimization of Deep Learning Precipitation Models Using Categorical Binary Metrics

Optimization of Deep Learning Precipitation Models Using Categorical Binary Metrics

Predicting power outages caused by extratropical storms

Iba: The First Pure Tropical Cyclogenesis Over the Western South Atlantic Ocean

Contact Info

Product

Resources

About