Detection of Hazard Areas of Frost Damage Using Numerical Terrestrial Data and NOAA/AVHRR Data

Abstract

Detection of the hazard areas for frost damage was attempted by using numerical terrestrial data and NOAA/AVHRR data. The areas in this analysis were the regions around the foot of Mt. Youtei (referred to here as Youtei), Ishikari plain (referred to here as Sorachi), the Tokachi plain (referred to here as Tokachi) and the Nayoro basin (referred to here as Nayoro). The study areas were 60km×50km.
The following three methods were employed for detecting valleys and basins using numerical terrestrial data; 1) Counting the number of negative gradient values of the altitude between a center cell and its eight adjacent cells (the 8 adjacent cells method); 2) Counting the number of negative gradient values between a center cell and its sixteen surrounding cells that are one cell width apart from it (the 16 adjacent cells method); 3)-A Counting the number of valleys that drain into a single cell (the contiguous maximum gradient method). This method involves tracing the maximum negative gradient values from cell to cell until a low point is reached from which only positive gradient values radiate. This technique was repeated for each cell in the entire grid. That is, each cell was used as a starting point, and the valley emanating from it was traced until a basin was reached; 3)-B Determining the number of adjacent cells for which the negative gradient value is greatest between the adjacent cell itself and the center cell (the maximum gradient method).
Two methods were employed to detect areas having low temperatures using NOAA/AVHRR data; 1) Counting the number of negative difference in temperature between the center cell and the eight adjacent cells (the 8 adjacent temp. method); 2) Locating cells with temperatures lower than the threshold temperature (the threshold temp. method).
The topographic data and temperature data for Youtei were then combined on a cell by cell basis to find how well the two types of data matched. The 16 adjacent cells method gave the same results as the 8 adjacent cells method.
The valleys and basins detected by the 8 adjacent cells method matched well with the low temperature areas located by both low temperature detecting methods. However, the 8 adjacent temperature method revealed which cells were more likely to get cold due to the local topography. On the other hand, the threshold temperature method revealed where valleys and basins with low temperatures were located.
The valleys and basins detected using the maximum gradient method did not match well with the areas of low temperature detected by either method.
The analysis discussed above was performed on the data for Youtei only. In order to compare the data obtained in all four areas, the coincidence index was calculated for each of the four study areas. The coincidence index is defined as the ratio of cells which represent either valleys or basins and have temperatures below threshold values, to the total number of cells with temperatures below the threshold.
The results indicated that these methods are effective for detecting hazard areas for frost damage in Youtei and Nayoro. However, the effectiveness of these methods is reduced for Sorachi and Tokachi. The small coincidence index in Sorachi is caused by a smooth and flat topography. In Tokachi hilly areas are separated by plains at a distance greater than can be covered by a center cell and its 8 adjacent cells.
The standard deviations in altitude between each cell and its 8 adjacent cells for which low temperatures were detected were calculated to determine topographic differences in each region. The results are shown in Table 4. The regions of Youtei and Nayoro have large standard deviations. On the other hand, the values for Sorachi and Tokachi are small. This demonstrates that low temperatures occur in smooth and flat areas in Sorachi and Tokachi.
The incidence matrix of the ranges for altitude in these regions and the inertial momentum around the diagonal