Roads and infrastructure are fundamental preconditions for the large-scale conversion of natural forests into agricultural landscapes. Therefore, geographic differences in accessibility are critical for understanding land use and land cover change (LULCC) dynamics. As one of the most dynamic agricultural frontiers globally, the Amazon has much attention in LULCC research. While most studies account for roads and infrastructure, LULCC research often relies on one-dimensional measures of accessibility (e.g., Euclidean distances), which may not optimally represent the underlying latent variable concept. In this study, we demonstrate how alternative concepts and measures of accessibility (specifically travel time maps) can have considerably different explanatory value in the prediction of LULCC. We adopt a panel-data model to explain the geographical distribution of pasture and crop expansion in the Brazilian Amazon using land cover data and travel time maps generated with high-quality representations of existing official and non-official road infrastructure. Our approach's novelty consists of comparing travel time to different markets during the wet- and the dry season and their effect on the allocation of LULCC within a macro scale modeling approach. Our results suggest that (1) pronounced differences between wet- and rainy season accessibility (due to road quality) increase the likelihood of pasture expansion and reduce the likelihood for crops, and (2) that alternative measures of infrastructure access (e.g., to markets versus towns or processing facilities) can explain different socio-economic aspects of LULCC. Our findings suggest that bad infrastructure quality might severely limit the possibility of establishing a less land-intensive agricultural model in the Amazon and that LULCC research can significantly benefit from improved and context-specific measures of accessibility.

道路和基础设施是将天然林大规模转变为农业景观的基本先决条件。因此，可达性的地理差异对于理解土地利用和土地覆盖变化 (LULCC) 动态至关重要。作为全球最具活力的农业前沿之一，亚马逊在 LULCC 研究中备受关注。虽然大多数研究都考虑了道路和基础设施，但 LULCC 研究通常依赖于可达性的一维度量（例如，欧几里德距离），这可能无法最佳地代表潜在的潜在变量概念。在这项研究中，我们展示了可访问性的替代概念和度量（特别是旅行时间图）如何在 LULCC 的预测中具有相当不同的解释价值。我们采用面板数据模型来解释巴西亚马逊地区牧场和作物扩张的地理分布，使用土地覆盖数据和旅行时间地图，这些地图由现有官方和非官方道路基础设施的高质量表示生成。我们的方法的新颖之处在于比较在雨季和旱季到不同市场的旅行时间及其对宏观尺度建模方法中 LULCC 分配的影响。我们的结果表明 (1) 雨季和雨季可达性之间的显着差异（由于道路质量）增加了牧场扩张的可能性并降低了农作物的可能性，以及（2）基础设施准入的替代措施（例如，进入市场）与城镇或加工设施）可以解释 LULCC 的不同社会经济方面。