Mapping Daily Maximum Temperature over Agricultural Areas with Complex Terrains.

Rises in the daytime temperature in summer cause high temperature stress, thereby adversely affecting the quality and yield of crops. In case the global warming continues, damages in the agricultural sector are expected to increase due to high temperature. Improving presumed reliability of the daily maximum temperature in an agricultural area where the observation density is sparse, compared with urban areas, is prerequisite to the development of agricultural technology that will cope with the situation and reduce damages caused by high temperature. Since the daily maximum temperature is closely related with insolation, in the case of a mountainous terrain, with solar irradiance varying depending on the slope, the mid-day temperature distribution also varies locally. This study proposes to develop practical methods of preparing a highly reliable precision map of the daily maximum temperature distribution targeted at catchment zones in a complex terrain. The 1500 LST temperature and the data covering the 4 hours immediately preceding 1500 (1100 - 1500) LST of solar irradiance integrated from 9 locations differing in the aspect from each other (3 locations at the altitude of a 50m level, 3 locations of a 100m level, 3 locations of a 300m level) sampled for a clear day with daily mean cloud amount standing at 0 during the whole year 2012 were collected.  The slope and aspect of each location were smoothed by means of grid cells numbering 5, 10, 15, 20, 25, 30 each in radius measuring 30m x 30m, and integrated irradiance of each cell and 1500 temperature were compared. As a result, in case it was smoothed with 25 grid cells (750m), a statistically significant regression model was derived where 1500 temperature increased or decreased by 0.83°C with each daily insolation integrated for 4 hours changing spatially by 1 MJ/m² (y = 0.83x + 0.04, r²=0.54). The regression model was applied to a rural catchment zone of about 50km² in an area with an altitude range of 5m – 1,100m where a high-density meteorological observatory network was in service and a 1500 temperature distribution map was prepared with a 30m × 30m grid cell as a unit for the year 2013, and compared with actual measurement data from the locations. An average root mean square error was calculated at 0.9°C, showing a substantial reduction from 1.2°C obtained by the existing method that had not applied the regression model. The mean error was calculated at –0.2°C, indicating that the bias was significantly improved, compared with –0.8°C obtained by the existing method.