In the age of big data, identifying causality has become the focus of different scientific disciplines. This importance is particularly acute in geography, where understanding complex systems is a tough task. This study provides an in-depth review of causal analysis techniques, assessing their strengths, assumptions, and limitations in geographic research. Using case studies of precipitation impacts on vegetation and runoff, we compare three key approaches: granger causality, the PC algorithm, and LiNGAM. Our findings reveal that (1) causal analysis is evolving from linear to nonlinear, and from bivariate to multivariate, despite challenges such as uncertainty testing. (2) Establishing causal direction is crucial, but distinguishing between direct and indirect causation is equally important. (3) While detailed assumptions enhance the refinement of causal approaches, they may limit their generalizability. Our results support the broader use of causal analysis in geographic research.

在大数据时代，识别因果关系已成为不同科学学科的焦点。这种重要性在地理学中尤其重要，因为理解复杂系统是一项艰巨的任务。本研究深入回顾了因果分析技术，评估了它们在地理研究中的优势、假设和局限性。通过降水对植被和径流影响的案例研究，我们比较了三种关键方法：格兰杰因果关系、PC 算法和 LiNGAM。我们的研究结果表明，（1）因果分析正在从线性演变为非线性，从双变量演变为多元，尽管存在不确定性测试等挑战。（2）确立因果方向固然重要，但区分直接因果和间接因果也同样重要。(3) 虽然详细的假设增强了因果方法的完善，但它们可能限制其普遍性。我们的结果支持因果分析在地理研究中更广泛的应用。