Earth scientists describe a wide range of observational measurements as “proxy measurements”. By referring to such a vast body of measurements simply as “proxy”, researchers dilute significant differences in the various ways that measurements relate to the phenomena they intend to describe. The limited language around these measurements makes it difficult for the nonspecialist to assess the reliability and uncertainty of data generated from proxy measurements. Producers and reviewers of proxy data need a common framework for conveying proxy measurement methodology, uncertainty, and applicability for a given study. We develop a functional distinction between different forms of measurement based on the different ways that their outputs (values, interpretations) relate to the phenomena they intend to describe (e.g., temperature). Paleotemperature measurements, which are used to estimate temperatures of systems in Earth's past, serve as a case study to examine and apply this new functional proxy definition. We explore the historical development and application of two widely used paleotemperature proxies, calcite δ¹⁸O and TEX₈₆, to illustrate how different measurements relate to the phenomena they intend to describe. Both proxies are vulnerable to causal factors that interfere with their relationship with temperature but address those “confounding causal factors” in different ways. While the goal of proxy development is to fully identify, quantify, and calibrate to all confounding causal factors, the reality of proxy applications, especially for past systems, engenders unavoidable and potentially significant uncertainties. We propose a framework that allows researchers to be explicit about the limitations of their proxies and identify steps for further development. This paper underscores the ongoing effort and continued need for critical examination of proxies throughout their development and application, particularly in Earth's history, for reliable proxy interpretation.

中文翻译：

---

代理可以衡量吗？评估和传达代理可靠性的框架

地球科学家将广泛的观测测量描述为“代理测量”。通过将如此大量的测量简单地称为“代理”，研究人员以各种方式将测量与他们打算描述的现象相关联的重要差异进行了稀释。这些测量的语言有限，使非专业人员很难评估从代理测量生成的数据的可靠性和不确定性。代理数据的产生者和审阅者需要一个通用的框架来传达给定研究的代理度量方法，不确定性和适用性。我们基于输出的输出（值，解释）与打算描述的现象（例如温度）的不同方式，在不同形式的测量之间建立了功能上的区别。用于估计地球过去系统温度的古温度测量值用作案例研究，以研究和应用这一新的功能代理定义。我们探讨了两种广泛使用的古温度代理方解石的历史发展和应用δ ¹⁸ O和TEX ₈₆，以说明不同的度量值与它们打算描述的现象之间的关系。两种代理都容易受到因果因素的干扰，这些因果因素会干扰它们与温度的关系，但会以不同的方式解决那些“混杂的因果因素”。虽然代理开发的目标是完全识别，量化和校准所有混杂的因果因素，但代理应用程序的现实（尤其是对于过去的系统）会带来不可避免的潜在不确定性。我们提出了一个框架，使研究人员可以清楚地了解代理的局限性，并确定进一步开发的步骤。本文强调了在代理的整个开发和应用过程中（尤其是在地球历史上）的不断努力和对代理的严格检查的持续需求，