面向无人机适飞场景的多气象要素格点预报方法

A Grid-Based Forecasting Method of Multiple Meteorological Elements for UAV Flyable Scenarios

摘要: 无人机低空飞行对气象条件具有高度敏感性，精准短临格点预报是保障飞行安全与任务规划的关键。现有预报多聚焦单站点或单要素，难以直接支撑面向适飞判断的多变量空间化分析。针对该问题，本文提出一种面向无人机适飞分析的多要素气象格点预报方法。首先，构建多阶段数据预处理流程，将多站点不规则观测统一转换为20 × 20规则格点序列；其次，根据气象要素物理属性设计分组建模策略，在统一时空预测框架下分别构建热力、风场与降雨分支，其中风场采用水平风分量建模以规避风向角周期边界问题，降雨采用对数变换与非负约束处理稀疏分布。基于研究区域多站点连续一年地面观测数据的实验表明，模型对未来4 h逐15 min的多要素场具有稳定的预测能力，温度、相对湿度与风速的MAE分别约为1.45℃、5.09%和0.61 m/s，能够有效保持各要素空间结构一致性。本文方法可为无人机适飞状态的空间化分析提供多要素格点预报基础。

Abstract: Low-altitude UAV flights are highly sensitive to meteorological conditions, and accurate short-term gridded forecasting is essential for ensuring flight safety and mission planning. Existing forecasts mainly focus on single-station or single-element predictions, making it difficult to directly support multivariable spatial analysis for flight-suitability assessment. To address this issue, this paper proposes a multi-element meteorological gridded forecasting method for UAV flight-suitability analysis. First, a multi-stage data preprocessing pipeline is constructed to convert irregular multi-station observations into uniform 20 × 20 regular gridded sequences. Second, a grouped modeling strategy is designed according to the physical properties of meteorological elements. Within a unified spatiotemporal prediction framework, separate thermal, wind, and rainfall branches are built: the wind branch adopts horizontal wind components to avoid the periodic boundary problem of wind direction angles, while the rainfall branch employs logarithmic transformation and non-negative constraints to handle sparse distributions. Experiments based on one year of continuous ground observation data from multiple stations in the study area show that the model achieves stable predictive capability for multi-element fields over the next 4 h at 15-min intervals, with MAEs of approximately 1.45˚C for temperature, 5.09% for relative humidity, and 0.61 m/s for wind speed, while effectively preserving the spatial structural consistency of each element. The proposed method provides a multi-element gridded forecasting foundation for spatialized UAV flight-suitability analysis.