Abstract Ground-based evaporative fraction (EF) observations have been used widely for validation purposes in previous remote sensing-based EF models. Few studies have investigated whether such measurements can be utilized for calibration use. In this paper, an observation-driven optimization method is proposed to quantify EF over a large heterogeneous area within the surface temperature-vegetation index framework. It is designed at both daily scale and seasonal scale with MODIS products and in-situ EF observations over the Southern Great Plains in the US. The goal is to search for the optimal dry edge within the allowable range that minimizes the difference between the estimated and observed EF of a given site. Results show that the accuracy produced using only one site for calibration has reached a level comparable to those produced by traditional triangle methods. Compared with the daily-scale optimization method, the seasonal-scale optimization method has not only demonstrated its superiority in accuracy but also held distinctive advantages over the traditional triangle methods. Specifically, the dry edge produced by our optimization method holds true under both clear sky and partially cloudy conditions. This has not only bypassed the repetitive work of previous triangle methods but also made it possible to conduct a continuous monitoring of EF. Besides, the optimization method is characterized by its simplicity in algorithm, stability in accuracy and extensibility in parameterization, which makes it a suitable tool for providing a quick and reasonable estimation of EF over large heterogeneous areas from a limited number of in-situ EF observations.

中文翻译：

---

一种用于连续估计大面积非均质区域蒸发率的观测驱动优化方法

摘要 在以前的基于遥感的 EF 模型中，地基蒸发率 (EF) 观测已被广泛用于验证目的。很少有研究调查此类测量是否可用于校准用途。在本文中，提出了一种观测驱动的优化方法来量化地表温度-植被指数框架内大面积异质区域的 EF。它是在日尺度和季节性尺度上设计的，使用 MODIS 产品和美国南部大平原的原位 EF 观测。目标是在允许范围内搜索最佳干边，以最大限度地减少给定站点的估计和观察到的 EF 之间的差异。结果表明，仅使用一个站点进行校准所产生的精度已达到与传统三角形方法产生的精度相当的水平。与日尺度优化方法相比，季节尺度优化方法不仅在精度上表现出优越性，而且与传统的三角形方法相比具有明显的优势。具体来说，我们的优化方法产生的干边在晴天和部分多云的条件下都适用。这不仅绕过了以前三角法的重复工作，而且使对 EF 进行连续监测成为可能。此外，该优化方法具有算法简单、精度稳定、参数化可扩展等特点，