Accurate simulation of summertime convection associated with the Asian monsoon trough over the subtropical western Pacific is important but diffcult to achieve in many general circulation models (GCMs). This study reports a case in which bias could be reduced by introducing a higher-resolution regional atmospheric model (RAM), two-way nested in an atmospheric GCM over the western Pacific. Additional partial-coupling experiments revealed that GCM bias correction was insensitive to the coupling domain. The two-way nesting effect was similar to one phase of a leading mode of natural variability in the system. This is indicative that the two-way nesting model provides more realistic tropical heating that effectively excites a correct phase of the intrinsic dynamical mode to reduce GCM bias.
We investigated a warming trend in the Kanto-Koshin area during a 30-year period (1976-2006). The warming trends at AMeDAS stations were estimated to average a little less than 1.3°C/30 years in both summer and winter. These warming trends were considered to include the trends of large-scale and local-scale warming effects. Because a regional climate model with 20-km resolution without any urban parameterization could not well express the observed warming trends and their daily variations, we investigated whether a mesoscale atmospheric model with an urban canopy scheme could express them. To make the simulations realistic, we used 3 sets of real data: National Land Numerical Information datasets for the estimation of the land use area fractions, anthropogenic heat datasets varying in space and time, and GIS datasets of building shapes in the Tokyo Metropolis for the setting of building aspect ratios. The time integrations over 2 months were executed for both summer and winter. A certain level of correlation was found between the simulated temperature rises and the observed warming trends at the AMeDAS stations. The daily variation of the temperature rises in urban grids was higher at night than in the daytime, and its range was larger in winter than in summer. Such tendencies were consistent with the observational results. From factor analyses, we figured out the classic and some unexpected features of urban warming, as follows: (1) Land use distribution change (mainly caused by the decrease of vegetation cover) had the largest daytime warming effect, and the effect was larger in summer than in winter; (2) anthropogenic heat had a warming effect with 2 small peaks owing to the daily variation of the released heat and the timing of stable atmospheric layer formation; and (3) increased building height was the largest factor contributing to the temperature rises, with a single peak in early morning.
The present study applies the WRF model involving the single-layer urban canopy model (hereafter, WRF_UCM) to urban climate simulation of the Tokyo metropolitan area for August (2004-2007) and compare results to (a) observations, and (b) the WRF model involving the slab urban model (hereafter, WRF_SLAB). In this urban area, WRF_UCM accurately captures the observed monthly mean daytime and nocturnal UHI, whereas WRF_SLAB does not show a nocturnal UHI. Moreover, the observed diurnal variations of the surface air temperature for central Tokyo and Kumagaya, a nearby inland city, are reproduced well by WRF_UCM. However, WRF_SLAB exhibits both a 1-hr phase shift and a 6.2°C excess oscillation magnitude over observations. In addition, WRF_UCM accurately reproduces the frequency distribution of surface air temperatures, showing a maximum at 27°C, whereas WRF_SLAB produce a bimodal distribution, with double peaks at 23 and 33°C. Finally, WRF_UCM does a much better job than WRF_SLAB at modeling the relative humidity.
This study presents the projected urban climate for the 2070s’ August in the three largest urban areas, Tokyo, Osaka, and Nagoya in Japan. To accurately evaluate the urban climate, the simulations use the Weather Research and Forecast (WRF) model with 3-km grid increment coupled to an urban canopy model (UCM). To project future urban climate, the simulations apply dynamical downscaling to three GCMs (MIROC3.2-medres, MRI-CGCM2.3.2a, CSIRO-Mk3.0) and use the ensemble average for results. The results provide estimates of the heat stress to future residents of Tokyo, Osaka, and Nagoya. The WRF-UCM model reproduces the observed spatial distribution of the surface air temperature in the 2000s’ August, giving an all-domain mean bias of -1.2°C and RSME of 2.7°C. For Tokyo, Nagoya, and Osaka, these biases are -0.6, -0.1, and -0.4°C. Moreover, the diurnal temperature variations at these urban stations are well reproduced. The projected monthly average August temperatures in the 2070s are about 2.3°C higher than those in the 2000s at the three urban areas and comparable to those in the record-breaking hot summer of 2010. (Predictions by individual ensemble members differ by 0.8-1.2°C.) As a result, urban areas will experience uncomfortable sleeping nights nearly every day in August, with roughly as many heat-induced sleeping-discomfort nights as those in 2010. Moreover, application of the wet-bulb globe temperature (WBGT) shows that people in Tokyo will be warned not to strenuously exercise outdoors for 62% of the daytime hours in the 2070s, a sharp increase from the 30% of the 2000s. (Predictions by individual ensemble members range from 54-67%). Osaka and Nagoya will have even more restrictions on outdoor exercise. Finally, the urban heat island intensity is 1.5°C in Tokyo of the 2070s, comparable to the background climate warming of 2.3°C.
Regional climate projections associated with global warming are of great importance for the development of mitigation and adaptation strategies but are subject to a variety of uncertainties. This study developed a probabilistic strategy to consider every conceivable uncertainty in a climate analogue with the use of a pattern-scaling methodology and bootstrap resampling. The uncertainty of the regional climate model (RCM) simulations, which is associated with the physics and dynamics of the RCMs, is comparable to the uncertainties due to emission scenarios of the greenhouse gases and the transient climate responses of the general circulation model. Comparison of the projections between the probabilistic and deterministic viewpoints demonstrated a benefit of the former method in applications to impact studies.
This study evaluated the downscaled downward shortwave radiation (Sd) and daily mean, maximum, and minimum surface air temperatures (Tm, Tx, and Tn) over Japan derived from four dynamical models and one statistical model for the period 1985-2004. These variables, Sd, Tm, Tx, and Tn, are often used as input data for impact models, such as crop growth models. Therefore, the evaluation of these variables is essential prior to the application of the downscaled climate data to impact model simulation. All models except for the statistical model remarkably overestimated the Sd throughout the year, whereas the area-averaged seasonal change of the temperature variables was accurately simulated. On the other hand, the statistical model accurately simulated the area-averaged seasonal change of Sd and temperatures. These tendencies were also observed for the area- and time-averaged values. The inconsistency found for dynamical models in the relationship between the Sd bias and each of the temperature variable biases is attributed to the large inter-model differences, such as the difference of the radiation scheme, the sub-grid scale cloud parameterization, and the land surface scheme.
This study assessed the sensitivity of the simulated future impact on forage yield over Japan to the precipitation change in growing season derived from the multiple downscaling models, taking the regional climate projection ensemble dataset for Japan as an example. Three regional climate models (RCMs: NHRCM, NRAMS, and TWRF), and one statistical model (CDFDM) provided the fine-resolution (20-km) climate data over Japan from the climate projection performed by a global climate model (GCM: MIROCHI) under A1B scenario. With the common boundaries for the RCMs (and predictor for the statistical model), there is a consistency for the increased summer precipitation. However, discrepancies were found for the degree of precipitation increase and change in the mean summer maximum number of consecutive dry days. These discrepancies caused the spread of the simulated future change in forage yield over Japan by 3.3-11.4% (2081-2100), relative to the present-day one (1981-2000). These results showed that the direction (increase or decrease) and amplitude of the simulated future impact differ between the climate scenarios from the downscaling models and those from the parent GCM, indicating that climate downscaling is a source of uncertainty in simulating future impact.
This study assesses ocean surface winds in regional climate models (RCMs) and evaluates the ability of RCMs to downscale the features of tropical cyclones (TCs). RCMs show a smaller bias in the mean ocean surface wind around Japan during summer than the reanalysis data that is used as boundary data because of the better representation of land/ocean contrast in RCMs. However, for extreme values of ocean surface winds, all RCMs show a large bias over the ocean south of Japan. The RCMs reasonably simulate the TC tracks for about 40% of TCs, whereas these models fail to simulate realistic TC tracks for the remaining TCs. The TC track errors in the RCMs spread over a wide range with peaks ranging from 100 to 200 km. Although two RCMs underestimate the surface wind speed associated with TCs, one RCM simulates it reasonably. Therefore, it is suggested that the bias in the extreme values of ocean surface winds can be caused not only by an insufficient representation of surface winds associated with a model TC but also by the model TC track errors. Moreover, these errors may affect the extreme values of precipitation produced by the interaction between TCs and topographies in Japan; therefore the extreme values should be used with caution. Multi-model ensemble approach contributes to reduce TC track errors. As a result, number of the TCs with the relatively small TC track errors increases up to about 60%.