This article describes the development of a machine learning (ML)-based algorithm for snowfall retrieval (Snow retrievaL ALgorithm fOr gpM–Cross Track…This article describes the development of a machine learning (ML)-based algorithm for snowfall retrieval (Snow retrievaL ALgorithm fOr gpM–Cross Track, SLALOM-CT), exploiting ATMS radiometer measurements and using the CloudSat CPR snowfall products as references. During a preliminary analysis, different ML techniques (tree-based algorithms, shallow and convolutional neural networks—NNs) were intercompared. A large dataset (three years) of coincident observations from CPR and ATMS was used for training and testing the different techniques. The SLALOM-CT algorithm is based on four independent modules for the detection of snowfall and supercooled droplets, and for the estimation of snow water path and snowfall rate. Each module was designed by choosing the best-performing ML approach through model selection and optimization. While a convolutional NN was the most accurate for the snowfall detection module, a shallow NN was selected for all other modules. SLALOM-CT showed a high degree of consistency with CPR. Moreover, the results were almost independent of the background surface categorization and the observation angle. The reliability of the SLALOM-CT estimates was also highlighted by the good results obtained from a direct comparison with a reference algorithm (GPROF).more
Averaging a set of individual measurements can reduce the stochastic error but can introduce a sampling error particularly for irregularly sampled dat…Averaging a set of individual measurements can reduce the stochastic error but can introduce a sampling error particularly for irregularly sampled data. We present a general method to estimate the total error of an averaged quantity as a combination of the measurement error and the sampling error without knowledge about the true average value of the distribution. Our approach requires covariance matrices connecting the retrieved measurement values to an independent reference data set. These covariance matrices can be obtained from a representative validation data set. We confirm the validity of the method by estimating the temporal sampling error of monthly mean cloud fractional cover (CFC) data derived from the Spinning-Enhanced Visible and Infrared Imager radiometer onboard the METEOSAT Second Generation (MSG) spacecraft, operated by the European Organization for the Exploitation of Meteorological Satellites. The estimated sampling errors are then compared with the true sampling errors calculated from an hourly sampled complete data set. For this purpose, we use ten sampling scenarios. Some of them address typical sampling problems like systematic over- and undersampling as well as hourly, daily, and random data gaps. Two additional sampling scenarios are directly related to the satellite application facility on climate monitoring monthly mean CFC data record. These are used to estimate the worst case sampling errors of this data record. The estimated total and sampling errors agree well with corresponding calculated values. We derive the needed covariance matrices by analyzing synoptic observations of the cloud fraction which are MSG diskwide available, the majority of them over European land surfaces. The method is not limited to temporal averaging cloud fraction data. Moreover, it is a general method that is also applicable to temporal and spatial averaging of other parameters as long as appropriate covariance matrices are available.more
The low accuracy of satellite cloud fraction (CF) data over the Arctic seriously restricts the accurate assessment of the regional and global radiativ…The low accuracy of satellite cloud fraction (CF) data over the Arctic seriously restricts the accurate assessment of the regional and global radiative energy balance under a changing climate. Previous studies have reported that no individual satellite CF product could satisfy the needs of accuracy and spatiotemporal coverage simultaneously for long-term applications over the Arctic. Merging multiple CF products with complementary properties can provide an effective way to produce a spatiotemporally complete CF data record with higher accuracy. This study proposed a spatiotemporal statistical data fusion framework based on cumulative distribution function (CDF) matching and the Bayesian maximum entropy (BME) method to produce a synthetic 1∘ × 1∘ CF dataset in the Arctic during 2000–2020. The CDF matching was employed to remove the systematic biases among multiple passive sensor datasets through the constraint of using CF from an active sensor. The BME method was employed to combine adjusted satellite CF products to produce a spatiotemporally complete and accurate CF product. The advantages of the presented fusing framework are that it not only uses the spatiotemporal autocorrelations but also explicitly incorporates the uncertainties of passive sensor products benchmarked with reference data, i.e., active sensor product and ground-based observations. The inconsistencies of Arctic CF between passive sensor products and the reference data were reduced by about 10 %–20 % after fusing, with particularly noticeable improvements in the vicinity of Greenland. Compared with ground-based observations, R2 increased by about 0.20–0.48, and the root mean square error (RMSE) and bias reductions averaged about 6.09 % and 4.04 % for land regions, respectively; these metrics for ocean regions were about 0.05–0.31, 2.85 %, and 3.15 %, respectively. Compared with active sensor data, R2 increased by nearly 0.16, and RMSE and bias declined by about 3.77 % and 4.31 %, respectively, in land; meanwhile, improvements in ocean regions were about 0.3 for R2, 4.46 % for RMSE, and 3.92 % for bias. The results of the comparison with ERA5 and the Meteorological Research Institute – Atmospheric General Circulation model version 3.2S (MRI-AGCM3-2-S) climate model suggest an obvious improvement in the consistency between the satellite-observed CF and the reanalysis and model data after fusion. This serves as a promising indication that the fused CF results hold the potential to deliver reliable satellite observations for modeling and reanalysis data. Moreover, the fused product effectively supplements the temporal gaps of Advanced Very High Resolution Radiometer (AVHRR)-based products caused by satellite faults and the data missing from MODIS-based products prior to the launch of Aqua, and it extends the temporal range better than the active product; it addresses the spatial insufficiency of the active sensor data and the AVHRR-based products acquired at latitudes greater than 82.5∘ N. A continuous monthly 1∘ CF product covering the entire Arctic during 2000–2020 was generated and is freely available to the public at https://doi.org/10.5281/zenodo.7624605 (Liu and He, 2022). This is of great importance for reducing the uncertainty in the estimation of surface radiation parameters and thus helps researchers to better understand the Earth's energy imbalance.more
Accurate solar surface irradiance data is a prerequisite for an efficient planning and operation of solar energy systems. Further, it is essential for…Accurate solar surface irradiance data is a prerequisite for an efficient planning and operation of solar energy systems. Further, it is essential for climate monitoring and analysis. Recently, the demand on information about spectrally resolved solar surface irradiance has grown. As surface measurements are rare, satellite derived information with high accuracy might fill this gap. This paper describes a new approach for the retrieval of spectrally resolved solar surface irradiance from satellite data. The method combines a eigenvector-hybrid look-up table approach for the clear sky case with satellite derived cloud transmission (Heliosat method). The eigenvector LUT approach is already used to retrieve the broadband solar surface irradiance of data sets provided by the Climate Monitoring Satellite Application Facility (CM-SAF). This paper describes the extension of this approach to wavelength bands and the combination with spectrally resolved cloud transmission values derived with radiative transfer corrections of the broadband cloud transmission. Thus, the new approach is based on radiative transfer modeling and enables the use of extended information about the atmospheric state, among others, to resolve the effect of water vapor and ozone absorption bands. The method is validated with spectrally resolved measurements from two sites in Europe and by comparison with radiative transfer calculations. The validation results demonstrate the ability of the method to retrieve accurate spectrally resolved irradiance from satellites. The accuracy is in the range of the uncertainty of surface measurements, with exception of the UV and NIR (≥ 1200 nm) part of the spectrum, where higher deviations occur.more
Knowledge of the wintertime sea-ice production in Arctic polynyas is an important requirement for estimations of the dense water formation, which driv…Knowledge of the wintertime sea-ice production in Arctic polynyas is an important requirement for estimations of the dense water formation, which drives vertical mixing in the upper ocean. Satellite-based techniques incorporating relatively high resolution thermal-infrared data from MODIS in combination with atmospheric reanalysis data have proven to be a strong tool to monitor large and regularly forming polynyas and to resolve narrow thin-ice areas (i.e., leads) along the shelf-breaks and across the entire Arctic Ocean. However, the selection of the atmospheric data sets has a large influence on derived polynya characteristics due to their impact on the calculation of the heat loss to the atmosphere, which is determined by the local thin-ice thickness. In order to overcome this methodical ambiguity, we present a MODIS-assisted temperature adjustment (MATA) algorithm that yields corrections of the 2 m air temperature and hence decreases differences between the atmospheric input data sets. The adjustment algorithm is based on atmospheric model simulations. We focus on the Laptev Sea region for detailed case studies on the developed algorithm and present time series of polynya characteristics in the winter season 2019/2020. It shows that the application of the empirically derived correction decreases the difference between different utilized atmospheric products significantly from 49% to 23%. Additional filter strategies are applied that aim at increasing the capability to include leads in the quasi-daily and persistence-filtered thin-ice thickness composites. More generally, the winter of 2019/2020 features high polynya activity in the eastern Arctic and less activity in the Canadian Arctic Archipelago, presumably as a result of the particularly strong polar vortex in early 2020.more