Search

The European Space Agency’s Climate Change Initiative (ESA CCI) Soil Moisture (SM) COMBINED product is a more than 40-year-long data record on global SM for climate studies and applications. It merges SM observations derived from multiple active and passive satellite remote sensing instruments in the microwave domain. Differences in sensor characteristics (such as frequency or polarization) can cause structural breaks in the product, which are not completely removed during the merging process. These artificially caused discontinuities can adversely affect studies using the long-term data set. In this article, we compare three adjustment methods in terms of reducing the number of detected breaks in the SM record. We investigate their impact on the data with multiple validation metrics. Their potential (negative) influence is examined by comparing trends in the data before and after homogenization. We find that all three presented methods can reduce the number of detected breaks in ESA CCI SM. Differences between the methods mainly concern their ability to handle inhomogeneities in variance. Evaluation of the corrected data shows the limited impact of homogenization in terms of quantitative validation metrics. Changes in SM trends due to removing breaks are found in some areas. We find that break correction overall improves the already rather homogeneous data set while preserving its climate describing characteristics. Quantile category matching is identified as the preferred method in terms of correcting breaks in ESA CCI SM. [ABSTRACT FROM AUTHOR]

The blending of soil moisture (SM) time series derived from different active and passiveremote sensing instruments with varying sensor characteristics (frequency, polarization,radiometric accuracy) in the global ESA CCI SM climate data record (CDR), canpotentially lead to artificial, structural breaks at transition times between sensorproducts, which may negatively affect long term studies on climate changes orrisk assessments, which make use of 40 years of SM observations provided in thedataset.We detect relative inhomogeneities in mean and variance between sensor periods usingnon-parametric statistical tests with homogeneous reference SM from MERRA2 land surfacemodel simulations and adjust breaks with respect to corrections from models derivedfrom observations within homogeneous sub-periods of ESA CCI and referenceSM.We compare different methods for adjusting inhomogeneities (relative adjustment based onmatching linear regression models, higher order moments adjustment based on predictionsfrom (non-linear) regression models and adjustment by matching quantile categoriesof empirical distributions) and rate them in terms of their usability on ESA CCISM.We use in-situ SM observations from the international soil moisture network (ISMN)and surface model SM for validation to investigate the impact of break correctionmethods on satellite SM and define a modular, extendable framework to test andadjust multiple breaks in a single observation series, taking into account seasonality,correlations between ESA CCI and reference SM, varying spatial and temporalcoverage of satellite SM as well as different lengths of homogeneous sub-periods. Weexamine the impact of adjustment on long term climate dynamics in ESA CCI SM bydetecting monotonic, significant trends using a non-parametric Mann-Kendall test andcompare our results to trends in other soil moisture data sets and recommend theuse of future versions of ESA CCI SM, for which the number of breaks will bereduced. [ABSTRACT FROM AUTHOR]