Uncovering hidden heat: Evaluating the plausibility of the Landsat Land Surface Temperature product over metal-roofed poultry farms in Northeast Oklahoma

By Mirte Riemens

Abstract
The Landsat Land Surface Temperature (LST) dataset is used across a wide range of disciplines, from monitoring plant stress or wildfires to assessing urban heat island effects. This dataset is derived by converting the satellite-measured thermal radiance into temperature values at the surface level. However, the accuracy of this conversion is highly dependent on the correct estimation of surface emissivity values. A particular problem in emissivity estimations is provided by metal surfaces, as these can have uncommonly low emissivity values.

Here, I investigate the plausibility of the Landsat LST over metal-roofed poultry barns, which are relatively isolated in the landscape. The Landsat LST product shows large differences in the temperatures over these poultry barns, reaching differences of 15 degrees Celsius between farms. By investigating the validity of these differences, I uncovered a systematic overestimation of emissivity over these metal roofs. As a result, there is a significant underestimation of LST. This error is most severe for newly constructed poultry farms, as the new galvanized steel of the roof has a significantly lower emissivity than old, galvanized steel, resulting in the large LST differences observed between differently aged poultry farms. I quantified the underestimation of LST by calculating a new LST for a poultry farm using a custom emissivity adjusted to the lower emissivity values of the metal barn roofs. This resulted in an underestimation of more than 17 degrees Celsius for a new poultry farm and approximately 3 degrees Celsius for an old poultry farm. Consequently, the Landsat LST product is likely to be significantly inaccurate for both sites.

Methods comparable to the emissivity estimation used in the Landsat LST product are found across multiple different LST products, making it likely that these products also significantly underestimate the LST over metal surfaces. As LST products are often used over urban areas, where the proportion of metal roofing can be substantial, these findings can have significant implications for the broader field of research. Errors of this magnitude have the potential to seriously mislead research conclusions, leading to subsequent misguided policy decisions. For example, such underestimations could result in a significant underestimation of urban heat island effects in areas with extensive metal roofing. Raising awareness of this issue in LST over metal is thus crucial to ensure the responsible and accurate use of the temperature products.