The Muon Space GNSS-R Surface Soil Moisture Product

Abstract

Muon Space (Muon) is building a constellation of small satellites, many of which will carry global navigation satellite system-reflectometry (GNSS-R) receivers. In preparation for the launch of this constellation, we have developed a generalized deep learning retrieval pipeline, which now produces operational GNSS-R near-surface soil moisture retrievals using data from NASA's Cyclone GNSS (CYGNSS) mission. In this article, we describe the input datasets, preprocessing methods, model architecture, development methods, and detail the soil moisture products generated from these retrievals. The performance of this product is quantified against in situ measurements and compared to both the target dataset (retrievals from the Soil Moisture Active-Passive (SMAP) satellite) and the v1.0 soil moisture product from the CYGNSS mission. The Muon Space product achieves improvements in spatial resolution over SMAP with comparable performance in many regions. An ubRMSE of 0.032 cm$^3$ cm$^{-3}$ for in situ soil moisture observations from SMAP core validation sites is shown, though performance is lower than SMAP's when comparing in forests and/or mountainous terrain. The Muon Space product outperforms the v1.0 CYGNSS soil moisture product in almost all aspects. This initial release serves as the foundation of our operational soil moisture product, which soon will additionally include data from Muon Space satellites.

Figures (10)

• Figure 1: An illustration of the application of the Muon generalized GNSS-R retrieval pipeline to SM, which transforms L1 GNSS-R land surface observations into operational L2 and L3 data products. The green boxes and text represent the data pipeline components, while the blue box and text represent the model development process. The dark purple box and text shows operational product generation using a finalized DL model. The CYGNSS and SMAP satellites depict the matching of target information (i.e., “truth”) to each GNSS reflection measurement.
• Figure 2: Observation coverage from one day (May 18, 2022) sampled by the CYGNSS constellation (grey tracks). Also shown is the SMAP Level 3 gridded SM product showing AM coverage (colored swaths), as well as the locations of the in situ validation sites (blue dots). CYGNSS data are used as the primary model input, SMAP AM data are used as the target for training, and the in situ data are used for validation of the model output.
• Figure 3: False-color image from Sentinel-2 over the Salton Sea region in southern California (a) and long-term averages of CYGNSS reflectivity gridded to 3 km (b), 9 km (c), and 36 km (d).
• Figure 4: Evaluation of the model performance against the target and comparison to the UCAR product. a) distributions for the target SM (green), predicted SM (pink), and UCAR SM (blue). b) the relationship between predictions and targets with color on a logarithmic scale. c) the global difference between prediction and target averaged over the validation window. d-g) rolling average windows of SM from SMAP (green), the Muon product (pink), and the UCAR product (blue), error from SMAP, correlation with SMAP,and RSME from SMAP, respectively. The colored lines under the time series plots indicate the different data windows.
• Figure 5: Results of input-ablation study (left/middle), and input sensitivity study (right). The bar chart lists the sensitivity of all the inputs used in the final model.