It is of great significance to study the deviation characteristics in intelligent grid forecasting of precipitation during the flood season in Hedong area of Gansu for improving the accuracy level of regional precipitation forecasting and warning, and enhancing the ability of disaster prevention and reduction services. By using precipitation data in flood season from 1 766 automatic meteorological observation stations in Hedong area, 264 precipitation cases from 2018 to 2020 were selected. Based on the 3-hour interval intelligent grid precipitation forecasts issued by the Central Meteorological Observatory, the Contiguous Rain Area (CRA) of forecast field and the actual field are identified and matched, and classified according to hit, miss and false alarm, to further study the CRA deviation characteristics. The results show that, for the precipitation cases hit by the forecast, the falling area deviation of the intelligent grid forecast of the warm forcing precipitation is the largest, followed by precipitations of the oblique frontal generation and cold forcing categories. The intensity deviation of the cold forcing precipitation is the largest, followed by precipitations of the warm forcing and baroclinic frontogenetic categories. The maximum morphological deviation is found in baroclinic frontogenetic precipitation, followed by the cold forcing and warm forcing categories. The forecast area of the warm forcing and oblique frontal precipitations is biased towards the north and east, while the cold forcing precipitation is biased towards the south and east. The forecasted precipitation area of three types precipitation is larger for the beta scale and below, and smaller for the the alpha scale. Forecasters can establish localized model correction schemes based on the results of CRA spatial testing to improve the service capability of intelligent grid precipitation forecast.

Based on the hourly precipitation data from 302 regional automatic weather stations in the Loess Plateau arid region of eastern Gansu from 2013 to 2020, digital elevation model data from shuttle radar topography mission and ERA5 reanalysis data of European Centre for Medium-Range Weather Forecasts, etc., the spatio-temporal distribution characteristics of short-term heavy rainfall are analyzed, and on this basis that the relationship between it and topography, geography factors is discussed. Then combined with an extremely short-term heavy rainfall event in 2021, the influence mechanism of terrain is summarized. The results are as follows: (1) The short-term heavy rainfall mainly occurs in summer in the Loess Plateau arid region of eastern Gansu, and in July the proportion of short-term heavy rainfall days （35.9%） is the most and the extremity is the strongest, while in August the proportion of short-term heavy rainfall times (46.9%) is the most and the rainfall intensity is the strongest. The rainfall intensity mainly ranges from 22.0 to 31.0 mm·h^-1, and it shows a multi-modal diurnal distribution with the most active, strongest and most extreme from 17:00 BST to next 00:00 BST, whose proportion of times is 56.8%. (2) The spatial distribution of the occurring times and hourly precipitation extremum of short-term heavy rainfall events is extremely uneven. The occurring times is less in the northwest and more in the southeast, it decreases sharply with the increase of rainfall intensity, and the short-term heavy rainfall occurs frequently in the area with valley bell-mouth topography, moreover, the palm landform is also the high incidence area of heavy precipitation above 30.0 mm·h^-1. The extremum is small in the middle and large in the northeast and southwest, the large value mostly distributes in eastern Qingcheng and western Heshui. (3) The influence of geographical and topographic factors on occurring times of short-term heavy rainfall is significant, and its contribution comes from geographical location, while their influence on precipitation extremum isn’t obvious. In general, the topographic forced uplift isn’t main influence mechanism on short-term heavy rainfall in the Loess Plateau arid region of eastern Gansu. (4) The mountain-valley wind circulation and surface mesoscale convergence line induced by it are important inducement for the formation of short-term heavy rainfall in valley bell-mouth topography area in the Loess Plateau arid region of eastern Gansu.

Based on the hourly precipitation from automatic weather stations, radar data at Xifeng station and radiosonde data at Pingliang station from 2008 to 2018, the temporal and spatial distribution, physical quantities and radar echo characteristics of short-term heavy rainfall in the Loess Plateau of Qingyang were analyzed statistically. The results are as follows: (1) The short-term heavy rainfall was easily to occur in the north and southeast of the Loess Plateau of Qingyang, and the business forecast should focus in Huan county and Zhengning county. (2) The earliest and latest short-term heavy rainfall occurred in mid-April and mid-October in the Loess Plateau of Qingyang from 2008 to 2018, respectively, and it occurred mostly in July and August, the peaks were in mid-July and mid-August. The diurnal variation of short-term heavy rainfall appeared typical double peaks pattern, the major and minor peaks occurred at about 17:00 BST and 09:00 BST, respectively, and the frequency of short-term heavy rainfall with different levels was higher from 16:00 BST to 18:00 BST. (3) The early warning indicators of water vapor condition of short-term heavy rainfall were 700 hPa specific humidity equal to or more than 7 g·kg-1 and temperature dew-point difference equal to or less than 4 ℃ in the Loess Plateau of Qingyang, as to energy condition, CAPE was equal to or more than 100 J·kg-1 and CIN was equal to or less than 120 J·kg-1, and for instability condition, △T7-5 was equal to or more than 15 ℃ and SI was equal to or less than 1 ℃. In addition, the early warning indicators of -20 ℃ and 0 ℃ layers height averaged 8.6 km and 5.1 km, and the thickness of frozen layer averaged 3.4 km. (4) The radar echo intensity of short-term heavy rainfall exceeded 38 dBZ, the height of echo top and the strongest echo center were more than 7.0 km and 3.6 km, respectively, and the VIL was more than 20 kg·m-2, which could be used for early warning indicators of short-term heavy rainfall. Upon inspection, these early warning indicators could provide some useful references for potential forecast and nowcasting of short-term heavy rainfall in the Loess Plateau of Qingyang.