Updated cadastral land values are a matter of critical importance for local governments: higher revenue of property taxes, more equitable treatment to taxpayers, a fundamental input in the design of public policies related to access to land and housing for the most vulnerable and a key feature in land value capture strategies to finance public infrastructure, to name just a few public policies that require correct valuations of land. However, in Latin America, outdated cadastral values are common to most cities. The reasons for this can be found in the complexity of the mass appraisal process, lack of institutional and fiscal capacity to undertake it and bureaucratic resistance to its implementation.

The objective of this paper is to present a mass appraisal methodology that uses only free and open data to achieve robust urban land valuations. Information from the OpenStreetMap Project is used to generate several land variables. In addition, the Global Human Settlement Layer of the European Commission is used to determine the level of consolidation of urban sprawl. Land value data were obtained from the Mapa de Valores de América Latina, a collaborative initiative that systemizes more than 68,000 data from more than 900 cities.

This information is used to train three tree-based machine learning models: Random Forest, Quantile Random Forest and Gradient Boosting Model. The results support the viability of the proposed strategy, simplifying the mass appraisal process in terms of costs, time and complexity of the information used.

更新的地籍土地价值对地方政府至关重要：更高的财产税收入，更公平地对待纳税人，在设计与最弱势群体获得土地和住房相关的公共政策方面的基本投入和一个关键特征在为公共基础设施融资的土地价值获取战略中，仅举几例需要正确评估土地的公共政策。然而，在拉丁美洲，大多数城市普遍存在过时的地籍价值。造成这种情况的原因在于大规模评估过程的复杂性、缺乏执行它的体制和财政能力以及对其实施的官僚主义阻力。

本文的目的是提出一种大规模评估方法，该方法仅使用免费和开放的数据来实现稳健的城市土地估值。来自 OpenStreetMap 项目的信息用于生成多个土地变量。此外，欧盟委员会的全球人类住区层用于确定城市扩张的巩固程度。土地价值数据来自 Mapa de Valores de América Latina，这是一项协作计划，系统化了来自 900 多个城市的 68,000 多个数据。

此信息用于训练三种基于树的机器学习模型：随机森林、分位数随机森林和梯度提升模型。结果支持了拟议战略的可行性，在成本、时间和所用信息的复杂性方面简化了大规模评估过程。