On the possibility of minimizing the number of hydrometeorological stations

Abstract

Abstract. The task of minimizing the number of stationary hydrometeorological stations (HMS) without significant loss of weather forecast quality may be relevant both from the standpoint of cost indicators and for solving a special forecast task in an area where it is not desirable or impossible to install a stationary HMS, for example, on the airport runway. The article proposes a solution to this problem by temporarily installing a mobile HMS at a given point, impossible for installing a stationary HMS, point of the territory and determining (based on statistical analysis) a more or less reliable relationship between the indicators of hydrometeorological temporal HMS and stationary HMS located nearby. For successful solving of this complex problem, it is necessary to reconcile two contradictory criteria: the indicator of the weather forecast accuracy at a given point and the minimum complexity indicator and, as a result, the correct identification of the structure and parameters of the forecast model. The article discusses the top three tasks: the first is the construction of a simple and adequate structure of the model structure; the second is an adequate evaluation of its parameters in case of inaccurate noisy measurements of the state of the process; the third is filtering, without a mistake, the study of the useful component of the signals, which takes place due to the filter inertia that exists when filtering noise and the forecast of filtered variables. Thus, the article presents the formulation and solution of these three tasks: by building the structure of a predictive model that takes into account the physical essence of natural processes and has a minimum number of unknown parameters; by building a method of unbiased and efficient estimation of the parameters of this model from noisy measurements of object state variables; by constructing a filter predicate as a system with astatism of a given order and adaptation of filtering indicators to the nonstationarity (heteroscedasticity) of the noise component of signals.