Extreme values of central Japan snowfall attributable to global climate change

1. Objectives and data

To investigate trends of heavy snowfall attributable to climate change, we use maximum snow depth data calculated during 1951–2010 with a climate model (experimentation by present climate). The Japanese hydrological–meteorological community used a climate simulation model to construct a huge database of hydro-meteorological values (https://www.miroc-gcm.jp/d4PDF/index_en.html), designated as d4PDF. It has data calculated under the present climate and data calculated under a future climate, which are affected by global warming. The "T-year value" is a stochastically and statistically estimated hydrological value for which the return period is T years.

2. Extreme values of snowfall

We used hourly snow depth data of d4PDF around Japan. First, we calculated the annual maximum snow depth at each calculation grid. Because this database consists of ensemble calculations of 50 members, data for 3,000 years (60 years × 50 members) are included. We investigated trends of 50-year values for which the return period is 50 years. Specifically, 50-year snowfall and snow depth were estimated. The annual maximum of snow depth at each calculation grid was calculated. Our research theme evaluates the “T-year value” for which the return period is the T-year is itself, but we calculated it using a non-parametric method at this stage as described below.

• We assume that the population of this database is not stationary. Therefore, for example, for 1951, we use 500 data which consist of data of 10 years (1951–1960) of 50 members.
• A 50-year value is the tenth largest value among 500 descending data.
• For 1952–2010, 50-year values are estimated using the same procedure.

3. Result

Figure 1 presents the time variational 50-year-snow-depth at five grid points in central Japan. Figure 2 shows 50-year values of 24 h snowfall as the same grid points. These show no long clear increasing or decreasing trend. However, after 1985, it is apparent that the 50-year-annual maximum snowfall during 24 h decreases slightly.

4. Future work

Although we describe trends of 1951–2010 herein, we can describe some other results we are investigating now, as presented below.

• Optimal methods for estimating a T-year value require checking for optimal methods among non-parametric methods, methods using annual maxima, and methods using a partial duration series.
• Although we use only a database for present climate conditions, we will present trends of the near future which are affected by global changes.

 

Figure 1: Time variational 50-year snow depth at five grid points.

 

Figure 2: Time variational 50-year snowfall at five grid points.