Weather forecasters had predicted heavy rainfall ahead of the devastating floods that hit central China this week, but issued a warning for the wrong place and wrong time. Su Aifang, deputy director of Henan’s meteorological service, said on Wednesday that the provincial government had been warned of the risk from extreme weather last Thursday. But the forecasters had predicted that the heaviest rain would hit Jiaozuo, a city at the foot of Taihang Mountain a day before the worst downpours happened. Past experience suggested that the wet air from the Pacific Ocean was more likely to rise and form clouds when it hit the 2,000-metre (6,560ft) peak. On Saturday the local authorities issued a notice warning that Jiaozuo could see up to 500mm (19.7 inches) of rain on Monday that risked “once-in-a-century” floods and relocated some residents from low-lying areas. China floods caused by typhoon airflow hitting area of high pressure Other areas, including the provincial capital Zhengzhou less than 100 kilometres (60 miles) to the south, were told to expect a smaller impact. But in the event, the worst of the rainfall hit Zhengzhou, a city of 12 million, a day later than forecast. On Tuesday morning, the city issued a red alert, the country’s highest warning level, but most residents were already on their way to work by then. The city saw more than 200mm of rain fall in just one hour, the highest rainfall ever recorded in China. By the afternoon, Zhengzhou saw more rain than it typically had in half a year, flooding the streets and trapping hundreds of passengers in subway trains. At least 25 people were killed in the floods, while power and drinking water supplies were knocked out in district after district. Some weather stations had no data because their equipment was damaged. Chen Tao, chief forecaster at the National Meteorological Centre, said in Beijing on Wednesday that China had made efforts to improve its extreme weather forecasting “but it remains a worldwide challenge”. Chen said modern forecasting models worked well for normal conditions, but the dynamics of extreme weather were poorly understood. “There are many uncertainties in extreme weather systems that can affect the accuracy of a forecast,” he said. Su said her team had been working 24-hour shifts over the past week, constantly updating their estimates as new data came in and issuing more than 1,000 warning messages by Wednesday morning. Previous studies have found that rainfall is one of the hardest things to forecast and most meteorologists say it is still not possible to predict how much rain will fall in a specific hour the following day. The National Meteorological Centre in Beijing is one of nine recognised by the World Meteorological Organization, a United Nations body. But despite the satellites and super computers used by the centre, its rain estimates are wrong more often than they are right. According to a study carried out last year by the State Key Laboratory of Severe Weather, China’s weather authorities got their 24-hour rain forecasts right about 15 per cent of the time in 2008, a figure that rose to 20 per cent by 2019. Even in the United States, Japan and Europe, where there is a much longer history of meteorology and generally smaller land areas to cover, the accuracy level – also known as the threat score – was about 30 per cent. One senior meteorologist said that although it would have been hard to forecast the record rainfall in Zhengzhou this week, the disaster had highlighted some areas that had been neglected, including a lack of monitoring stations. The scientist from the Nanjing University of Information Science and Technology, who asked not to be named because he was also working for the military, said hourly rainfall predictions were still overreliant on ground radar. These can only obtain a limited amount of information from the clouds and the greater use of satellite images and other sources of information would help give a more accurate picture. Although the number of radar and other monitoring facilities have been gradually increasing, the scientist said areas such as Beijing and Shanghai were better served than poorer cities such as Zhengzhou. China floods: rail passengers tell of rising water and last goodbyes In recent years, Chinese forecasters have started using artificial intelligence to improve accuracy, but the meteorologist warned that its performance “depends on the quality of data you feed into it”. “We not only need to increase the number of radar stations in cities but rural areas as well. There are too many blind spots across the country,” he said.