Encoder–decoder-based image transformation approach for integrating multiple spatial forecasts
As the damage caused by heavy rainfall worsens, there is a growing demand for improved forecasts. One practical approach to address this issue is the linear integration of multiple existing forecasts, which allows for visualizing the contribution of each forecast at different locations. However, cur...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Elsevier
2023-06-01
|
Series: | Machine Learning with Applications |
Subjects: | |
Online Access: | http://www.sciencedirect.com/science/article/pii/S2666827023000269 |
_version_ | 1797796313194037248 |
---|---|
author | Hirotaka Hachiya Yusuke Masumoto Atsushi Kudo Naonori Ueda |
author_facet | Hirotaka Hachiya Yusuke Masumoto Atsushi Kudo Naonori Ueda |
author_sort | Hirotaka Hachiya |
collection | DOAJ |
description | As the damage caused by heavy rainfall worsens, there is a growing demand for improved forecasts. One practical approach to address this issue is the linear integration of multiple existing forecasts, which allows for visualizing the contribution of each forecast at different locations. However, current methods such as arithmetic and Bayesian averages utilize a single weight shared across the entire space, making it difficult to account for local variations in importance. Additionally, while U-Net-based spatial forecasts have been proposed, they are limited to short-term predictions and do not facilitate the visualization of individual forecast contributions due to their non-linear processes. To overcome these challenges, we propose a new integration framework based on U-Net image transformation. This framework generates weight images that dynamically integrate forecasts based on both time and location. To effectively handle large and imbalanced precipitation data, we introduce novel extensions to the U-Net model. These extensions address heavily imbalanced precipitation data and enable position and time-dependent integration. Experimental results using real precipitation forecast data in Japan demonstrate that our proposed method outperforms existing integration methods. |
first_indexed | 2024-03-13T03:31:11Z |
format | Article |
id | doaj.art-3f276bd665454884b59cac11a1c3dfaf |
institution | Directory Open Access Journal |
issn | 2666-8270 |
language | English |
last_indexed | 2024-03-13T03:31:11Z |
publishDate | 2023-06-01 |
publisher | Elsevier |
record_format | Article |
series | Machine Learning with Applications |
spelling | doaj.art-3f276bd665454884b59cac11a1c3dfaf2023-06-24T05:19:43ZengElsevierMachine Learning with Applications2666-82702023-06-0112100473Encoder–decoder-based image transformation approach for integrating multiple spatial forecastsHirotaka Hachiya0Yusuke Masumoto1Atsushi Kudo2Naonori Ueda3Wakayama University, 930, Sakaedani, Wakayama-city, Wakayama, 640-8510, Japan; Center for AIP, RIKEN, Nihonbashi 1-chome Mitsui Building, 15th floor, 1-4-1, Nihonbashi, Chuo-ku, Tokyo, 103-0027, Japan; Corresponding author at: Wakayama University, 930, Sakaedani, Wakayama-city, Wakayama, 640-8510, Japan.Wakayama University, 930, Sakaedani, Wakayama-city, Wakayama, 640-8510, Japan; Center for AIP, RIKEN, Nihonbashi 1-chome Mitsui Building, 15th floor, 1-4-1, Nihonbashi, Chuo-ku, Tokyo, 103-0027, JapanNumerical Prediction Development Center, Japan Meteorological Agency, 1-2, Nagamine, Tsukuba-city, Ibaraki, 305-0052, JapanCenter for AIP, RIKEN, Nihonbashi 1-chome Mitsui Building, 15th floor, 1-4-1, Nihonbashi, Chuo-ku, Tokyo, 103-0027, JapanAs the damage caused by heavy rainfall worsens, there is a growing demand for improved forecasts. One practical approach to address this issue is the linear integration of multiple existing forecasts, which allows for visualizing the contribution of each forecast at different locations. However, current methods such as arithmetic and Bayesian averages utilize a single weight shared across the entire space, making it difficult to account for local variations in importance. Additionally, while U-Net-based spatial forecasts have been proposed, they are limited to short-term predictions and do not facilitate the visualization of individual forecast contributions due to their non-linear processes. To overcome these challenges, we propose a new integration framework based on U-Net image transformation. This framework generates weight images that dynamically integrate forecasts based on both time and location. To effectively handle large and imbalanced precipitation data, we introduce novel extensions to the U-Net model. These extensions address heavily imbalanced precipitation data and enable position and time-dependent integration. Experimental results using real precipitation forecast data in Japan demonstrate that our proposed method outperforms existing integration methods.http://www.sciencedirect.com/science/article/pii/S2666827023000269Precipitation forecastImage-to-image transformationEncoder–decoder network |
spellingShingle | Hirotaka Hachiya Yusuke Masumoto Atsushi Kudo Naonori Ueda Encoder–decoder-based image transformation approach for integrating multiple spatial forecasts Machine Learning with Applications Precipitation forecast Image-to-image transformation Encoder–decoder network |
title | Encoder–decoder-based image transformation approach for integrating multiple spatial forecasts |
title_full | Encoder–decoder-based image transformation approach for integrating multiple spatial forecasts |
title_fullStr | Encoder–decoder-based image transformation approach for integrating multiple spatial forecasts |
title_full_unstemmed | Encoder–decoder-based image transformation approach for integrating multiple spatial forecasts |
title_short | Encoder–decoder-based image transformation approach for integrating multiple spatial forecasts |
title_sort | encoder decoder based image transformation approach for integrating multiple spatial forecasts |
topic | Precipitation forecast Image-to-image transformation Encoder–decoder network |
url | http://www.sciencedirect.com/science/article/pii/S2666827023000269 |
work_keys_str_mv | AT hirotakahachiya encoderdecoderbasedimagetransformationapproachforintegratingmultiplespatialforecasts AT yusukemasumoto encoderdecoderbasedimagetransformationapproachforintegratingmultiplespatialforecasts AT atsushikudo encoderdecoderbasedimagetransformationapproachforintegratingmultiplespatialforecasts AT naonoriueda encoderdecoderbasedimagetransformationapproachforintegratingmultiplespatialforecasts |