| 16 Mar 2026
Improving geolocation and footprint characterisation of microwave radiometer observations using coastal crossings
Abstract. Spaceborne microwave radiometers (MWR) provide essential information on atmospheric humidity to enable wet tropospheric correction (WTC) in satellite altimetry. While WTC accuracy is generally high over the open ice-free ocean, it degrades in coastal areas due to land contamination in the radiometer's field of view. Correcting for this requires precise knowledge of the instrument's geolocation and spatial response characteristics.
This study presents a data-driven in-orbit approach to improve MWR geolocation and footprint characterisation for the Fundamental Data Record for Radiometry (FDR4RAD_V1), which constitutes the European Space Agency's (ESA) reference dataset of MWR observations from the ERS-1, ERS-2 and Envisat missions. By correlating observed brightness temperatures with effective land fractions at coastal transitions, we determine deviations between nominal and actual geolocation values, expressed as errors in along-track (ELON) and across-track (ECRO) direction, as well as footprint size, expressed as full width at half power (FWHP). A two-round iterative approach based on random-effects meta-analysis mitigates the effects of parameter interdependence and yields well-defined estimates with quantified uncertainties.
The analysis comprises more than 26,000 half-orbits sampled from ten repeat cycles per mission, spanning their combined full operational lifetimes (1991–2012). Substantial deviations between nominal and actual geolocations are observed for all instruments, with combined geolocation errors (ELON and ECRO via RSS) decreasing from 2.40 ± 0.07 km (ERS-1, 23.8 GHz) to 1.04 ± 0.08 km (Envisat, 23.8 GHz), indicating progressive improvements in positioning accuracy across missions. The standard errors of approximately 0.1 km, equivalent to an instrument pointing uncertainty of ~0.01°, reflect the precision achievable with this large-sample approach. For footprint size, the retrieved FWHP values range between 18.59 ± 0.04 km (Envisat, 36.5 GHz) and 21.78 ± 0.04 km (ERS-2, 36.5 GHz). Analysis of footprint geometry yields aspect ratios between 0.94 and 1.06, indicating small deviations from circularity. Temporal stability is confirmed for all mission–channel–parameter combinations except ELON for ERS-2 at 36.5 GHz, where a statistically significant trend coincides with documented gyroscope failures, revealing the method's sensitivity to platform changes.
The improved geolocation and footprint characterisation, provided for each instrument and channel, enables more accurate land contamination correction, enhancing wet tropospheric correction and sea level determination in coastal areas. The methodology is transferable to other near-nadir microwave radiometers, including those onboard the Sentinel-3 and Sentinel-6 missions, thereby supporting cross-mission consistency for multi-decadal sea-level records.
Comments to Authors:
This paper presents a data-driven in-orbit approach to improve microwave radiometer geolocation and footprint characterization using coastal crossings. By maximizing the correlation between observed brightness temperatures and modeled effective land fractions (ELF), the authors estimate along-track geolocation error (ELON), across-track error (ECRO), and full width at half power (FWHP). The study uses a large ensemble of global coastal crossings and a two-round iterative random-effects meta-analysis framework to reduce parameter interdependence and quantify uncertainties. The results demonstrate systematic geolocation biases and mission-dependent footprint characteristics, with implications for improving coastal wet tropospheric correction in satellite altimetry. Overall, I find this to be a technically strong and valuable study, and I would recommend the manuscript for publication after satisfactory responses to the comments provided below.
Major Comments:
1. Although the paper later evaluates footprint geometry and aspect ratios, the actual TB–ELF correlation optimization is still performed using a circular/symmetric Gaussian footprint model. This may be biased in cases where the true antenna response is elliptical or asymmetric. The manuscript would benefit from either incorporating elliptical footprint parameters directly into the optimization framework or providing quantitative evidence that the circular approximation does not significantly affect ELON, ECRO, and FWHP retrievals.
2. The methodology assumes that TB variations across coastal crossings are primarily controlled by land–water emissivity contrast. However, atmospheric humidity gradients, surface temperature variability, precipitation, and complex coastal morphology may also influence TB–ELF correlation maxima. A more detailed sensitivity analysis quantifying the impact of environmental heterogeneity on retrieval accuracy would strengthen confidence in the robustness of the approach.
3. The Mann-Kendall test is applied with N = 10 repeat cycles per mission and a relaxed significance threshold of α = 0.10, explicitly chosen to compensate for low statistical power (lines 538–541). With N = 10, the MK test has very limited power to detect subtle monotonic trends: for a two-tailed test at α = 0.10, the minimum detectable effect size (in terms of Kendall's τ) is large, meaning that moderate but physically meaningful drifts could easily go undetected. The paper correctly identifies one significant trend (ERS-2, 36.5 GHz ELON, p = 0.03) and attributes it to gyroscope failures, which provides important physical validation. However, for ERS-2's 23.8 GHz channel, split-period ELON estimates are provided (Table 4) showing a shift from 1.46 ± 0.13 km to 1.71 ± 0.13 km — a ~0.25 km change that, while not meeting the α = 0.10 threshold (p = 0.37), is physically plausible given the same gyroscope event. The authors should report explicit power calculations or confidence intervals for the trend estimates and consider whether bootstrapped uncertainty bounds on the MK statistic would be more appropriate given the small sample size and potential serial correlation across adjacent orbital cycles.
Minor Comments:
1. The manuscript would benefit from a brief discussion of computational cost and runtime, particularly given the large number of coastal crossings and iterative optimization procedure.