局域β多样性的纬度梯度格局可以同时反映局域过程（如生态位分化、扩散限制和生态漂变等）和大尺度生态过程（如气候和进化历史等），是生态学研究中最根本的格局之一。然而，局域生态过程的影响往往被低估，很大程度上是由于多数β多样性指数在统计上依赖于γ多样性，导致与γ多样性相关的大尺度生态因子的影响被高估。因此，到目前为止，局域和大尺度生态过程对β多样性纬度格局的相对贡献还在激烈的争论中。一般来说，基于多度信息的β多样性指数比仅包含物种出现与否信息的指数能反映更多的群落结构信息，而且对取样不足更加稳健，因而基于多度的指数所反映的生态过程也更加准确，但是，其与γ多样性是否在统计上相互独立，还缺乏系统性的研究。同时，生态学家有关于零模型能否有效消除β多样性指数与γ多样性统计依赖性的争论一直没有达成一致。另一方面，群落生态学中，不同群落构建过程对β多样性影响的相对重要性，通常由环境和空间变量结合变差分解的方法来进行量化。然而，当分析中某些重要的环境或空间变量缺失时，未被解释的空间相关性可能导致残差的空间自相关，进而得出有偏差甚至错误的结论。如何更加准确地衡量不同局域过程对β多样性纬度梯度格局的相对贡献，是未来亟待解决的问题。 针对以上问题，我们首先选择五类广泛应用的基于多度的β多样性指数，通过模拟和观测数据相互印证的方法，检验并比较其与γ多样性的统计相关性，从而筛选出适合进行区域间比较的指数。同时，我们还评估了零模型方法消除β多样性与γ多样性统计相关性的有效性。接着，我们沿纬度梯度收集了东亚地区由21个大型森林动态监测样地组成的数据集，结合局域地形因子和生态位分析方法以及区域气候变量等大尺度因子，在消除γ多样性的影响后，进一步分析局域和区域过程的相对贡献。最后，采用多尺度空间排序（MSO）方法，通过地统计学和回归分析的相关假设检验，来诊断是否存在显著的残差空间自相关。因此，我们首先利用模拟分析来检验MSO方法的有效性，然后，进一步利用中国东部的古田山24公顷大样地数据，在不同的空间尺度上，检验β多样性分析中残差空间自相关的影响。 模拟结果显示，基于多度的β多样性指数不同程度地在统计上依赖γ多样性。只有多样地香农熵指数相对独立于γ多样性，该指数采用了针对取样不足的偏差校正方法。同时，Jaccard - Chao指数也在一定程度上表现出γ多样性独立性，该指数通过估计群落间共同缺失的物种来校正取样偏差，但只有当群落取样量较大时有效。零模型方法可以完全消除Jaccard - Chao指数的γ多样性依赖性，但只能部分降低，而不能完全消除其它指数对γ多样性的依赖。基于东亚森林大样地数据集，我们发现，以往被广泛检验到的β多样性与纬度负相关的格局主要是由β多样性指数与γ多样性的相关性导致的。而局域过程相对于大尺度过程有更强的效应值，也进一步明确了局域过程在β多样性纬度格局形成中的重要作用。进一步地，局域群落的模拟分析发现，当重要的环境和空间因子缺失时，会导致模型出现残差空间自相关。另外，在古田山大样地的多尺度排序分析中，在较大尺度上（20 m × 20 m, 40 m × 40 m 和 50 m × 50 m）并未检测到显著的残差空间自相关。然而，即使解释变量包含多达24个高分辨率的地形和土壤因子以及足以表示所有尺度空间结构的MEM空间变量，残差空间自相关依然存在于10 m × 10 m 尺度上。 综上所述，我们得出以下结论：基于多度的β多样性指数中，多样地香农熵指数和结合零模型的Jaccard - Chao 在统计上相对独立于γ多样性，可用于β多样性区域间比较或生物地理梯度分析。而其它指数有待在数学独立性和取样偏差校正方面做出进一步改进，以更准确地进行区域间多样性比较。基于此，我们发现了与以往不同的非线性β多样性纬度梯度格局，并进一步明确了局域地形异质性和生态位分化过程对此格局形成的重要贡献。最后，残差空间自相关问题对β多样性纬度梯度格局分析存在潜在的影响，尤其是在较小的空间尺度上；MSO方法可有效检验残差空间自相关，为未来进一步分析不同局域过程对β多样性纬度梯度格局的相对贡献提供了方法上的补充。

The latitudinal gradient in local β diversity, which reflects both local (e.g., niche differentiation and ecological drift) and large scales (e.g., climate and evolutionary history) processes, is an integral component of the most fundamental patterns in ecology. However, effects of local processes have often been underestimated, in large part due to strong dependence of β diversity metrics on γ diversity. Consequently, the relative importance of local- and large-scale processes in shaping β diversity remain poorly understood. Inferences from abundance-based β metrics should be more powerful due to being more informative and robust to undersampling than incidence-based metrics, but the problem of γ diversity dependence remains largely unexamined. Meanwhile, an intense debate is going on regarding whether null model approaches are adequate to fix this problem. On the other hand, the relative roles of community assembly processes in structuring β diversity are usually quantified by variation partitioning with respect to environmental and spatial variables. However, if important environmental or spatial factors are omitted in the analysis, unaccounted spatial correlation may lead to residual spatial correlation and problematic inferences. Therefore, how to measure the relative contribution of local processes for latitudinal gradient of β diversity more accuratly becomes a problem demanding prompt solution. Here, we first chose five widely-used β metrics to examine their magnitude of γ diversity dependence using simulated and empirical data, and we also assess the effectiveness of null model approaches. Subsequently, we collected a dataset comprising 21 large plots throughout East Asian forests to determine the relative contributions of local effects (topographic heterogeneity and niche differentiation) and regional effects on local β diversity along a latitudinal gradient, after accounting for dependence of γ diversity. Finally, Multi-scale ordination (MSO) can be used to diagnose residual spatial correlation by checking assumptions related to geostatistics or regression analysis. We illustrate the performance of MSO using a simulated data set with known properties, and then test for residual spatial correlation in β diversity analyses of the Gutianshan (GTS) 24 - ha subtropical forest plot in eastern China. Simulation analyses showed that most abundance-based metrics were dependent on γ diversity in varying degrees. One exception was multiple-site Shannon entropy, which was relatively independent of γ diversity when a reliable bias correction method of undersampling was applied. Jaccard-Chao was also relatively insensitive to γ diversity when community size was larger, because bias of undersampling was corrected by estimating the effect of undetected shared species. Null model approach adequately removed γ diversity dependence of Jaccard-Chao, but partly reduced γ diversity dependence of other metrics without bias correction of undersampling. Next, based on forest plots throughout East Asia, we showed that the well-known pattern of local β diversity along latitude was mainly caused by the dependence of β diversity metrics on γ diversity. Stronger effects of local relative to large scale processes imply that local processes are likely more important in shaping spatial pattern of local β diversity along latitudinal gradient. Finally, simulation analyses showed that residual spatial correlation would be generated when environmental and spatial variables were eliminated. In GTS plot, even though we used up to 24 topographic and edaphic variables mapped at high resolution and spatial variables representing spatial structures at all scales, significant residual spatial correlation still existed at the 10 m ×10 m quadrat scale. This invalidated the analysis and inferences at this scale. Multiple-site Shannon entropy and β deviation of Jaccard-Chao are relatively independent of γ diversity except when sample size is unreasonably small, providing robust comparison of β diversity among regions. For other metrics, more improvement on mathematical independence and undersampling bias correction are needed to facilitate comparison of β diversity among regions. Based on these foundings, we found disparate pattern of β diversity along latitude, compared with previous studies. And furtherly confirmed the important contribution of local topographic heterogeneity and community - level niche differentiation. We also show that MSO provides a complementary tool to test for residual spatial autocorrelation in β diversity analyses, supporting for furtherly examining the relative contribution of local processes for latitudinal gradient of β diversity.