
# Research Plan: Birds Migrate Longitudinally in Response to the Resultant Asian Monsoons of the Qinghai-Tibet Plateau Uplift

## Problem

The uplift of the Qinghai-Tibet Plateau (QTP) represents one of Earth's most significant geological events, fundamentally shaping biogeographic patterns across continents. While the plateau's high elevation is widely recognized as an orographic barrier influencing various taxonomic groups, the role of the resultant Asian monsoon system in shaping species movement patterns remains poorly understood. Most existing studies have conceptualized the QTP merely as a topographical obstacle, overlooking the complex climatic consequences of its uplift.

We hypothesize that the QTP uplift has systematically altered avian migration patterns, potentially shifting migratory directions from latitudinal to longitudinal orientations. We propose that the Asian monsoon system generated by the uplift, rather than high elevation per se, is the primary driver of these changes. Our research addresses the fundamental question: how has the QTP uplift influenced migratory strategies of birds that must traverse this massive geological feature to complete their annual cycles?

This investigation is crucial because it will provide insights into how major geological events influence biogeographic patterns of migratory species and yield testable hypotheses for understanding observed avian distributions in a changing world.

## Method

We will employ a multi-faceted approach combining species distribution modeling with environmental reconstruction to compare avian migration patterns before and after the QTP uplift. Our methodology centers on creating counterfactual scenarios to isolate the effects of plateau uplift on bird migration strategies.

For environmental reconstruction, we will use the Community Earth System Model (CESM) version 1.0.4 to simulate pre-uplift climatic conditions. We will model two scenarios: actual current elevation versus a maximum elevation of 300m to represent pre-uplift conditions. The model will incorporate dynamic atmosphere, land, ocean, and sea-ice components with preindustrial simulation parameters including 280 ppmv atmospheric CO2 concentration and modern orbital parameters.

We will focus on five key environmental variables: monthly wind patterns, annual temperature, annual precipitation, elevation, and annual vegetation cover. These variables will enable us to distinguish between the effects of topographic barriers versus monsoon-driven climatic changes.

For species distribution modeling, we will employ Maximum Entropy (MaxEnt) models to correlate current species distributions with environmental conditions, then project these relationships onto pre-uplift environmental scenarios. We will complement this with an Adaptive Spatio-Temporal Model (AdaSTEM) to capture dynamic seasonal distributions and migration patterns.

## Experiment Design

We will analyze migration patterns of 50 avian species that migrate across the QTP, selected from comprehensive species lists and filtered to include only complete migrants. Our experimental design involves two primary analytical approaches:

**Species Distribution Analysis**: We will use eBird Basic Dataset from 2019 to model current seasonal distributions of target species. We will apply rigorous data filtering criteria including complete checklists only, specific protocol types (Traveling or Stationary), observation durations between 5-300 minutes, and removal of observers with expertise below the 2.5% percentile. To address spatiotemporal sampling biases, we will conduct spatiotemporal subsampling using a global hexagonal hierarchical geospatial indexing system.

We will extract 106 predictor variables across six categories: sampling effort variables, temporal variables, topographic variables, land cover data, bioclimate variables, and vegetation indices. The AdaSTEM framework will model abundance patterns using XGBoost as the base classifier, with 10 ensemble folds and specific spatial-temporal parameters.

**Migration Direction Analysis**: We will quantify migratory directions using azimuth angles between breeding and wintering areas, as well as between adjacent stopover sites. For species with available tracking data from Movebank and literature sources, we will digitalize migration routes to obtain precise movement patterns.

We will calculate wind connectivity using modified rWind package algorithms, replacing default wind data with our CESM monthly wind data. This will enable assessment of wind costs and their influence on migration directions across different geographic regions (West QTP, Central QTP, East QTP).

**Statistical Modeling**: We will employ both random forest models for prediction and Bayesian multivariate regression models to measure environmental influences on migration directions. We will analyze different season-stage combinations (spring/autumn × overall/regional) and include species as random variables through hierarchical modeling.

All environmental variables will be standardized for comparison, and Bayesian models will be implemented using PyMC with NUTS sampling. We will assess model convergence using potential scale reduction factors and effective sample sizes, ensuring all parameters meet criteria of Rhat < 1.03 and ESS > 400.

This experimental design will enable us to test our hypothesis that QTP uplift has shifted avian migration from latitudinal to longitudinal patterns, and determine whether monsoon systems or elevation changes are the primary drivers of these shifts.