An improved method of soil moisture meter calibration for satellite data validation at watershed scale

Elijah Cheruiyot,Collins Mito,Massimo Menenti

DOI: https://doi.org/10.1007/s12145-023-01170-w

2023-11-25

Earth Science Informatics

Abstract:This work presents an improved gravimetric algorithm to derive reference soil moisture, with removal of some of the hypothesis on which its original expression was based, and addition of a new corrective term that takes into account the interdependence between temperature and non-unitary water density. The temperature correction term improves reference measurements by up to 0.55% of their values in the temperature range 10–35°C. The temperature-corrected reference measurements were applied to the calibration of a hand-held soil moisture meter (Lutron PMS-714) for three soil texture types: medium, fine, and very fine. Linear regression models were used to calibrate the meter for each soil type, and the resulting calibration equations were validated with field data sampled from Sondu-Miriu watershed in Western Kenya. The validation produced errors (RMSE = 0.022, 0.010, 0.010 m 3 /m 3 ) that are significantly better than the meter's reported factory calibration errors of ± 0.05 m 3 /m 3 . While calibrations did not improve correlation statistics (R 2 and RMSE), they did significantly reduce biases (+ 0.009, + 0.004, -0.001 m 3 /m 3 ) compared to uncalibrated ones (-0.216, -0.181, -0.184 m 3 /m 3 ). Additionally, the calibrated meter values compared well with Soil Moisture Active Passive (SMAP) surface moisture data, with errors (RMSE = 0.010, 0.007, 0.008 m 3 /m 3 ) well within SMAP recommended value of ± 0.04 m 3 /m 3 . A spatial scalability test showed that the calibrations are adequately robust (with R 2 = 0.81, RMSE = 0.016 m 3 /m 3 , and Bias = + 0.005 m 3 /m 3 ), permitting calibration equations derived from one site to be scaled out to other sites of similar soil texture regime.

geosciences, multidisciplinary,computer science, interdisciplinary applications