Estimating ice temperature from short records in thermally disturbed boreholes

• Published in: Journal of Glaciology
• Main Author: Humphrey, Neil
• Format: Article in Journal/Newspaper
• Language: English
• Published: Cambridge University Press (CUP) 1991
• Subjects: Earth-Surface Processes; Antarc*; Antarctica; Ice Stream B; Journal of Glaciology
• Online Access: http://dx.doi.org/10.1017/s0022143000005840; https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S0022143000005840

Abstract A technique to estimate undisturbed ice temperature is discussed for sensors placed in boreholes that have been heated to the melting point during drilling, and for which only a limited time span of temperature record is available. A short temperature record after the hole refreezes commonly results when using hot-water or steam drills, where measurements are constrained by logistics, ice deformation, sensor drift or other problems, or where the refreezing time is long because of near-freezing ice temperatures or large hole sizes. Short data records are also typical in ongoing drilling programs where temperature information may be necessary for the program itself. Building on analyses by Lachenbruch and Brewer (1959) and a numerical model by Jarvis and Clarke (1974), it is shown that estimates of undisturbed temperatures can be made from records of temperature that extend only marginally beyond the initial refreezing. Complex effects of hole size, heating history, and the thermodynamic and geometrical effects of a moving boundary (the freezing borehole walls) are important to temperature decay immediately after freeze-up, so that the standard technique of comparing temperature decay to an inverse of time model is not applicable, and comparsion has to be made to a numerical model of heat flow to a refreezing borehole. Data from Ice Stream B, Antarctica, are compared to the numerical model to illustrate the technique. Data are also compared to simpler (inverse time) thermal models, and a potential for error is pointed out, since a short data record can be spuriously matched with the simpler, one or two free-parameter, models.