基于卷积神经网络（Convolutional Neural Network，CNN）的降水降尺度技术能够利用深度学习模型从低分辨率全球气候模式生成高分辨率降水数据，是评估气候变化在区域和局地尺度上对环境造成影响的关键技术，具有重要的现实意义。本文提出了一种基于 CNN 的降水降尺度模型-地形引导注意力网络（Terrain-guided Attentian Network，TGAN），能够实现 2°大气变量的降尺度操作，并得到 0. 1°的高分辨率降水场。该模型采用拉普拉斯金字塔作为渐进式降尺度框架，逐级提升分辨率并重建降水场；同时引入地形引导注意力模块，通过注意力机制以多尺度方式聚合大气数据和相应尺度的高程信息，加强模型的特征学习能力，从而实现模型模拟精度的提高。本研究以黄河中游地区为研究区域，使用 2001-2010 年的 ERA 5 日大气变量、GPM IMERG 日降水数据和高程数据进行模型训练，并利用 2011-2020 年的数据验证 TGAN 的有效性。结果表明，相较于传统 CNN 模型，TGAN 在日、月和年尺度的时空降水模拟中均表现出更高精度，尤其是在日尺度上优势更为显著：其平均均方根误差低至 5. 10 mm·d⁻¹，平均相关系数高达 0. 42，预测的第 95 百分位数和第 99 百分位数也更接近于 GPM IMERG 观测数据。同时，TGAN 模拟的日降水概率密度分布与 GPM IMERG 也更加吻合，尤其是在大降水区间中，表现出了更高的一致性。此外，本文进一步分析了不同损失函数对 TGAN 模型降尺度性能的影响。采用 RMSE 损失函数可以提升模型整体的预测精度，但在极端降水模拟上存在明显低估；而 Bernoulli-Gamma 损失函数虽整体精度略低，但却能更准确地重现极端降水事件，其概率密度分布与 GPM IMERG 及站点观测数据基本保持一致，表现出较好的极端降水捕捉能力。总体而言，TGAN 通过结合地形引导注意力机制以及Bernoulli-Gamma 损失函数，在黄河中游地区的降水降尺度任务中展现出显著优势，不仅提升了整体模拟精度，同时能够更好地捕捉极端降水事件，为复杂地形区域的高分辨率降水模拟提供了有力支撑。

As a powerful deep-learning downscaling technique，convolutional neural networks （CNNs） are widely utilized to generate high-resolution precipitation data from low-resolution global climate models through data-driven approaches，playing a critical role in assessing the impacts of climate change at both regional and lo‐ cal scales. This study proposes a CNN-based precipitation downscaling model，the Topography-Guided Attention Network（TGAN），which downscales coarse-resolution（2°）atmospheric variables to produce high-resolution precipitation fields at 0. 1° . The model adopts a Laplacian pyramid as a progressive，multi-level downscaling framework，in which the spatial resolution of precipitation fields is incrementally enhanced and precipitation structures are reconstructed through successive stages. In addition，a topography-guided attention module is in‐ corporated，which leverages an attention mechanism to integrate atmospheric variables with elevation data at cor‐ responding spatial scales. By combining these multi-scale features，the module strengthens the network’s capaci‐ ty for feature representation and learning，thereby improving the accuracy and reliability of simulated precipita‐ tion. Focusing on the middle reaches of the Yellow River，TGAN is trained on daily ERA5 atmospheric vari‐ ables，GPM IMERG daily precipitation data from 2001 to 2010，together with static elevation data，and validat‐ ed using data from 2011 to 2020. The results indicate that，compared with a conventional CNN model，TGAN achieves higher accuracy in spatiotemporal precipitation simulations at daily，monthly，and annual scales. At the daily scale，TGAN achieves a lower average root mean square error（5. 10 mm/day）and a higher average corre‐ lation coefficient（0. 42）compared with the conventional CNN model. Additionally，TGAN more accurately captures extreme precipitation events（95th and 99th percentiles）and better aligns with GPM IMERG observa‐ tions in probability density distribution，particularly for heavy precipitation range. This study further investigates the impact of different loss functions on the downscaling performance of TGAN. Using the RMSE loss function improves overall predictive accuracy but leads to underestimation of extreme precipitation events，whereas the Bernoulli-Gamma loss function，although slightly less accurate overall，more faithfully reproduces extreme pre‐ cipitation events. Its probability density distributions are highly consistent with both GPM IMERG data and sta‐ tion observations，indicating that the model has an enhanced capability to capture extreme precipitation events， thereby better reproducing the distribution characteristics of precipitation frequency. Overall，by combining the topography-guided attention mechanism with the Bernoulli-Gamma loss function，TGAN demonstrates clear ad‐ vantages in downscaling precipitation over the middle reaches of the Yellow River，not only improving overall simulation accuracy but also better representing extreme precipitation events，providing a robust and reliable tool for high-resolution precipitation modeling in complex terrain regions.