The annual phytoplankton bloom is an important marine event. Its annual variability can be easily recognized by ocean-color satellite sensors through the increase in surface Chlorophyll-a concentration, a key indicator to quantitatively characterize all phytoplankton groups. However, a common problem is that the satellites used to gather the data are obstructed by clouds and other artifacts. This means that time series data from satellites can suffer from spatial data loss. There are a number of algorithms that are able to reconstruct the missing parts of these images to varying degrees of accuracy, with Data INterpolating Empirical Orthogonal Functions (DINEOF) being the most popular. However, DINEOF has a high computational cost, taking both significant time and memory to generate reconstructions. We propose a machine learning approach to reconstruction of Chlorophyll-a data using a Variational Autoencoder (VAE). Our method is 3-5x times faster (50-200x if the method has already been run once in the area). Our method uses less memory and increasing the size of the data being reconstructed causes computational cost to grow at a significantly better rate than DINEOF. We show that our method's accuracy is within a margin of error but slightly less accurate than DINEOF, as found by our own experiments and similar experiments from other studies. Lastly, we discuss other potential benefits of our method that could be investigated in future work, including generating data under certain conditions or anomaly detection. / Graduate

[pt] A variabilidade das fontes de energia renovável, como energia eólica, apresenta um desafio significativo para o operador do sistema elétrico, em especial para o médio prazo (de horas a dias à frente). Isos porque é um período crítico para tomada de decisões do setor, sendo influenciado tanto por dados recentes quanto por padrões mais amplos. O atual estudo propõe a utilização de uma rede convolucional para gerar cenários para as componentes u- (latitudinal) e v- (longitudinal) do vento, utilizando o algoritmo Redes Adversárias Generativas Condicionais para treinamento. O modelo gerador proposto foi comparado com o estado da arte para previsão meteorológica, um sistema de previsão numérica. Os resultados mostram que o modelo - tendo um custo computacional inferior, menos informações de entrada e estabilidade de longo prazo similar - foi capaz de superar o benchmark em um quarto dos meses do conjunto de teste na previsão de duas semanas à frente (28 passos de 12 horas). Além disso, as medianas das séries geradas são estatisticamente iguais às previstas pelo estado da arte em 71.97 por cento dos casos. / [en] The variability of renewable energy sources, such as wind power, presents a significant challenge for grid operators in maintaining operational stability. This is specially true to the medium-term (from hours to days ahead), which is both influenced by recent past data and broader trends and heavily influences decision making. This research proposes a Convolutional Generator Network conditioned on the previous step of u- (latitudinal) and v- (longitudinal) wind speed components to generate wind speed scenarios using the Conditional Generative Adversarial Networks training algorithm. The model is compared to the state of the art in weather forecasting, Numerical Weather Prediction Systems. The proposed generator model outperforms the benchmark for a forth of the months in the test dataset when predicting over two weeks (28 12-hourly steps) starting from a single data point with much lower computational cost, less input data and similar long-term stability. Additionally, its forecasts are statistically equal to the state-of-the-art in 71.97 percent of series.

No description available.