Surface heat flow (SHF) serves as a vital parameter for assessing the heat transfer from deep Earth to the surface, which can provide crucial insights into internal geodynamic processes. As the “roof of the world,” the Tibetan Plateau and its tectonic evolution are highly important in terms of global climate change and geodynamic study. However, a comprehensive understanding of the SHF distribution across most regions of the Tibetan Plateau is limited due to sparse measurement data. To surmount this limitation, a spatially intelligent approach has been developed: The geographically neural network weighted regression with enhanced interpretability (EI-GNNWR). This method integrates spatial heterogeneity and nonlinear interactions between geophysical and geological factors to predict the SHF distribution across the Tibetan Plateau. In this study, the EI-GNNWR model is used to accurately predict SHF across the entire region. After evaluating the effectiveness and interpretability of the EI-GNNWR model, our results demonstrate that medium to high SHF values are predominantly concentrated in the southern, northeastern, and southeastern sectors of the Tibetan Plateau. These observations suggest that the formation of zones with high SHF values may be strongly influenced by the Moho depth, ridges, topography, and average curvature of satellite gravity gradients. Especially, higher SHF values may indicate more profound geodynamic activities such as collisional orogeny, shear deformation zones, or lithospheric extension. These findings offer novel insights into the spatial patterns of SHF and deepen our understanding of the geothermal formation mechanisms driven by underlying tectonic activities.

中文翻译：

---

数据驱动方法揭示青藏高原地表热流分布


地表热流 (SHF) 是评估从地球深处到地表的传热的重要参数，可以为内部地球动力学过程提供重要的见解。作为“世界屋脊”的青藏高原及其构造演化对于全球气候变化和地球动力学研究具有重要意义。然而，由于测量数据稀疏，对青藏高原大部分地区的SHF分布的全面了解受到限制。为了克服这一限制，开发了一种空间智能方法：具有增强可解释性的地理神经网络加权回归（EI-GNNWR）。该方法综合了空间异质性和地球物理和地质因素之间的非线性相互作用来预测青藏高原的SHF分布。在本研究中，EI-GNNWR 模型用于准确预测整个地区的 SHF。在评估 EI-GNNWR 模型的有效性和可解释性后，我们的结果表明，中高 SHF 值主要集中在青藏高原的南部、东北部和东南部。这些观测结果表明，高SHF值区域的形成可能受到莫霍面深度、山脊、地形和卫星重力梯度平均曲率的强烈影响。特别是，较高的 SHF 值可能表明更深刻的地球动力学活动，例如碰撞造山、剪切变形带或岩石圈伸展。这些发现为SHF的空间模式提供了新的见解，并加深了我们对底层构造活动驱动的地热形成机制的理解。