Meteorological calendar

After the Cultural Revolution started, I couldn’t do any research work in the countryside, because I had to work in the field day and night. But even the reign’s strongest power could only restrict people’s hands and feet, it can’t restrict the people’s spirit, and the head is still on my shoulder. I can dominate it by myself. So I began to think something about mathematics, I chose weather array as the virtual material for test and verify, I recorded the change of the weather everyday. With the weather record I recorded for 5 years, I began to probe the method of cycle analysis, and I found that the process of weather has its own periodicity. Though my major was not in this field. I created original theory <analysis of primary cycle> unexpected. At that time, I only had the weather array recorded by myself, and the calculating instruments were only abacus and slide ruler, but I had made up the weather forecast of the whole year day by day and the result was nearly as correct as the medium weather report in the location I lived.

Since1979, I had regained my profession and turned to work in a meteorological department, where the references were abundant and accurate, and I had electronic calculating device, what’s more, I could use medium-sized computer twice for the experiment. However, I failed on the contrary. Someone learning meteorology told me that weather couldn’t be calculated by mathematical method. But I asked in return: “Don’t you use mathematical method to your current forecasting?” In my opinion, weather forecasting could be done from two directions and used two methods: one for kinetics that obtained the future weather by using the most abundant data of the world weather circumstance; the most meticulous kinetics equation and integrate calculus to calculating. Another way was the method of movement, which analyzed the longest weather record array and then found its periodicity. It was considered by the astronomy that when the sun goes around by the earth, it could forecast solar eclipse and lunar eclipse, what it relied on was periodicity. Moreover, according to the philosophic theory, all regularity was in fact periodicity.

In the early 90s, I did the research in my forecast method with my own XT computer. Almost a year, my machines always ran whole day and night, but I couldn’t reach the conclusion. Because of absence effective scoring methods. The forecasting still failed when using hypothesis and simplified calculation procedure. I really felt puzzled: why couldn’t I reach the progeny that the datum was so exactitude and facilities is advanced several times, Did the weather really has not regularity period?

I began a new round research when the new century came. The material data I used wasn’t any change, it was still day’s precipitation and temperature in Xi’an during 1932to 1991. Though the computer I used is not the fastest (only about 1/10 of fastest), but it is 50-100 times faster than XT. Exceptionally it shows predict effective in form graphic. Of course, it didn’t completely succeed at present, but we have had delightful progenies to prove that <analysis of primary cycle >is correct. If parameter is selected good enough, the simulation can be much exact (all exact in 20000 data). Predict of extrapolate may reached 70 % of the simulation’s (temperature). Now precipitation forecasting is still bad, it only reached to 10% of the simulation. If precipitation glide average array is adopted, it can reach to 50% or so. As we can’t use square err directly, we have to use a set of temporary scoring method. Whether the method is good or not, directly master the process of experiment. So I publicize my scoring method and wish the experienced specialist forecaster to bring forward some ameliorate opinion.

I believe that the network can advance my research work, when I have a production with my method, which can calculate weather calendar to any area. If someone is interested in it, and can prepare good historical material for me, I will return him meteorological calendar with the calculating production. <Format of data > tells how to prepare the historical material, you can retain the last 400 of them as the checking up of forecast. Except the weather material (precipitation, temperature, sunlight and so on), other array will be ok, and the longer the better, and at least more than 10000, or to 50000,100000. I hope I can have practical content of <weather calendar> in the late this year.