TY - JOUR
T1 - Bias correction for spatially interpolated daily mean air temperature during winter in eastern Hokkaido using multimodal machine learning
AU - Murakami, Keach
AU - Hirota, Tomoyoshi
AU - Shimoda, Seiji
AU - Yazaki, Tomotsugu
N1 - Funding Information:
Meteorological data collected in the present study were kindly provided by Tokachi Federation of Agricultural Cooperatives, JA Okhotsk Abashiri, Shari town, Town of Ozora, Bihoro Town, JA Kitamirai and Obihiro City. We thank Shinji Yokoyama (AGW), Mutsuko Suzuki (AGW), and Iwao Oshima (JWA) for supporting data migration; Hirokazu Fukushima (JMA) for helpful discussion; Shigefumi Hatakeyama (JA Kitamirai) and Yuya Ito (JA Kitamirai) for validating meteorological observation data; Dr. Manabu Nemoto (HARC/NARO) and Dr. Satoshi Inoue (HARC/NARO) for comments on the manuscript. This work was financially supported by JSPS KAKENHI Grant Number JP19H00963.
Publisher Copyright:
© 2020, Society of Agricultural Meteorology of Japan. All rights reserved.
PY - 2020/10/10
Y1 - 2020/10/10
N2 - Interactions between boundary layer wind and topography form non-uniform air temperature distributions in cold and snow-covered regions. Because of this heterogeneity, spatially interpolated air temperatures sometimes deviate from observed values. To evaluate the quality of spatially interpolated daily mean temperatures (Tint) provided by a 1 km gridded meteorological data service (Ohno et al., 2016), we collected observed temperatures (Tobs) obtained at meteorological observation sites located near farmland in the Tokachi and Okhotsk regions̶in eastern Hokkaido, Japan̶in winter (October-March) and revisited the bias in the interpolated temperatures (dT). The root-mean-square error (RMSE) of Tint obtained at 88 sites was 1.16°C, and the absolute median dT values were greater than 1°C at 14 sites. The variance of dT was greater on cold and calm days, suggesting the involvement of radiative cooling and the accumulation of cold air parcels. To correct Tint by estimating dT at a given site by considering the formation mechanisms of the temperature distributions, we attempted to develop a multimodal machine learning model that had four predictors: surface and boundary layer meteorological data and topographical and geographical features around each site. To analyze the influence of the spatial extent of the topography and geography around each site, we compared models having these predictors with various sizes of the region of interest (ROI). By training the models and applying them to an independent test dataset, it has been shown that bias correction using models with a small topographical ROI (30×30 km or smaller) reduced the RMSE. The RMSE of the test dataset decreased by ~0.1°C via the application of a nested model, suggesting the potential usefulness of the presented approach for locally confined meteorological events. However, the biases were increased at several sites by application of the models, thus implying that further improvement is essential for practical use.
AB - Interactions between boundary layer wind and topography form non-uniform air temperature distributions in cold and snow-covered regions. Because of this heterogeneity, spatially interpolated air temperatures sometimes deviate from observed values. To evaluate the quality of spatially interpolated daily mean temperatures (Tint) provided by a 1 km gridded meteorological data service (Ohno et al., 2016), we collected observed temperatures (Tobs) obtained at meteorological observation sites located near farmland in the Tokachi and Okhotsk regions̶in eastern Hokkaido, Japan̶in winter (October-March) and revisited the bias in the interpolated temperatures (dT). The root-mean-square error (RMSE) of Tint obtained at 88 sites was 1.16°C, and the absolute median dT values were greater than 1°C at 14 sites. The variance of dT was greater on cold and calm days, suggesting the involvement of radiative cooling and the accumulation of cold air parcels. To correct Tint by estimating dT at a given site by considering the formation mechanisms of the temperature distributions, we attempted to develop a multimodal machine learning model that had four predictors: surface and boundary layer meteorological data and topographical and geographical features around each site. To analyze the influence of the spatial extent of the topography and geography around each site, we compared models having these predictors with various sizes of the region of interest (ROI). By training the models and applying them to an independent test dataset, it has been shown that bias correction using models with a small topographical ROI (30×30 km or smaller) reduced the RMSE. The RMSE of the test dataset decreased by ~0.1°C via the application of a nested model, suggesting the potential usefulness of the presented approach for locally confined meteorological events. However, the biases were increased at several sites by application of the models, thus implying that further improvement is essential for practical use.
UR - http://www.scopus.com/inward/record.url?scp=85092339991&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85092339991&partnerID=8YFLogxK
U2 - 10.2480/agrmet.D-20-00027
DO - 10.2480/agrmet.D-20-00027
M3 - Article
AN - SCOPUS:85092339991
SN - 0021-8588
VL - 76
SP - 164
EP - 173
JO - J. AGRICULTURAL METEOROLOGY
JF - J. AGRICULTURAL METEOROLOGY
IS - 4
ER -