Abstract: This paper describes a concept of forecast difficulty and its definition based on analysis of traditional evaluation methods of river runoff forecast levels, and develops two methods for its calculation using forecast error distributions and considering various forecast situations such as different rainfalls, forecast lead times and inflows. These methods can achieve effective calculations of forecast difficulty in various forecast situations under different standards of the forecast levels; and by combining them with evaluation of the forecasters' levels of runoff forecasting, we work out a comprehensive method for evaluating the forecasters' levels that takes into account of the forecast difficulty factor. A case study shows that compared with traditional methods, our comprehensive evaluation method is effective in considering various forecast difficulties in different forecast situations. And it can fully reflect the contribution of difficult forecast situations (such as long lead time) to the comprehensive forecast level; and provides more accurate, scientific and reasonable evaluation of runoff forecast levels, thus effectively promoting the forecasters' self-improvement of forecast levels.

Key words: Runoff forecasting, error, forecast difficulty, forecast level, comprehensive evaluation