Snow water equivalent (SWE) is a key parameter in the Earth’s energy budget and water cycle. It has been demonstrated that SWE can be retrieved using active microwave remote sensing from space. This necessitates the development of forward models that are capable of simulating the interactions of microwaves and the snow medium. Several proposed models have described snow as a collection of sphere- or ellipsoid-shaped ice particles embedded in air, while the microstructure of snow is, in reality, more complex. Natural snow usually forms a sintered structure following mechanical and thermal metamorphism processes. In this research, the bi-continuous vector radiative transfer (bi-continuous-VRT) model, which firstly constructs snow microstructure more similar to real snow and then simulates the snow backscattering signal, is used as the forward model for SWE estimation. Based on this forward model, a parameterization scheme of snow volume backscattering is proposed. A relationship between snow optical thickness and single scattering albedo at X and Ku bands is established by analyzing the database generated from the bi-continuous-VRT model. A cost function with constraints is used to solve effective albedo and optical thickness, while the absorption part of optical thickness is obtained from these two parameters. SWE is estimated after a correction for physical temperature. The estimated SWE is correlated with the measured SWE with an acceptable accuracy. Validation against two-year measurements, using the SnowScat instrument from the Nordic Snow Radar Experiment (NoSREx), shows that the estimated SWE using the presented algorithm has a root mean square error (RMSE) of 16.59 mm for the winter of 2009–2010 and 19.70 mm for the winter of 2010–2011.

Validation of satellite albedo products is an essential step because their quantitative application lie in their ability to record the real state of the earth surface. Upscaling in situ measurements to the corresponding pixel scale is necessary due to the spatial scale mismatch between in situ and satellite measurements. Machine learning-based models have been increasingly used for upscaling because they can yield more reliable results than traditional methods. Nevertheless, the main controlling factors on upscaled results have rarely been discussed. This article explores the control factors that bring uncertainties to the upscaled results based on machine learning models. Three machine learning models, including random forest (RF), $k$ -nearest neighbor (KNN), and Cubist models, were selected to upscale single site in situ -based albedo to the coarse pixel scale. The upscaled results were carefully assessed through comparison with pixel scale albedo reference. The results indicate that the accuracy of upscaled results depends on the machine learning models, the inclusion of key variables related to albedo, the dataset selection of these variables, the amount of training data, and the sensitivity of machine learning models to these factors. Despite the dependence on control factors, the machine learning-based upscaling methods generally have excellent applicability across different spatial scales and over other untrained areas. Therefore, they open the door to generating a time series of globally, spatially continuous distributed reference datasets with sufficient length, consistency, and continuity to adequately fulfill the requirement of a comprehensive validation.

Monitoring near-surface soil freeze/thaw patterns is becoming essential under the context of global changes as it is more sensitive to climatic fluctuation compared with subsurface thermal characteristics and its evolution could be an early warning of changes in near-surface permafrost. It requires continuous long term and stable data record for understanding hydrological, ecological and biogeochemical responses of permafrost to global climate change. AMSR2 (Advanced Microwave Scanning Radiometer 2) is designed as a successor of AMSR-E (Advanced Microwave Scanning Radiometer – Earth Observing System) to ensure continuity of such observation. In this study, a linear regression is used to inter-calibrate the AMSR-E and AMSR2 brightness temperatures. Then discriminant function algorithm is adopted to produce a long term freeze/thaw data record. It is compared with in situ air temperature measurements from both the temporal and spatial aspects. The results show that the accuracy is consistent between AMSR-E and AMSR2 with a value above 85 %, according to the result of spatial distribution accuracy. Analysis is conducted with this data record to explore the spatial distribution of frost days, its changing trend and the frost probability of each pixel on a specific date. The mean annual frost days of high northern latitude (HNL, > 45° N) zone is 214.2 ± 69.5 days and the trend of frost days indicates that the frost period is decreasing at a rate of −0.0065 day/month in 27 % of the domain which is defined by significance level of the F-test, and most of which are concentrated in the high latitude area specifically over the Northeast of Canada, Central and Eastern Russia and most part of Eastern Europe. The significant changes in frost days mostly occur in regions of discontinuous permafrost and transient permafrost. The spatial distribution of the frost days and its trend variations are found to be consistent with the minimum temperature anomalies trend. It indicates that the global warming is not constant at different regions over the globe. Further analysis over the Qinghai-Tibetan Plateau where discontinuous permafrost, island permafrost, seasonally frozen ground exist demonstrated that the frost period is shortening slightly over the past decade, and the last frost date is advanced over more than half of the region. It is considered to be a remarkable indication for permafrost degradation in this area.