An Alternative Machine Learning-Based Methodology for H-SAF H35 Fractional Snow Cover Product

2021 ◽  
Author(s):  
Semih Kuter ◽  
Cansu Aksu ◽  
Kenan Bolat ◽  
Zuhal Akyurek
Keyword(s):  
Machine Learning ◽  
Snow Cover ◽  
Computational Cost ◽  
Multivariate Adaptive Regression Splines ◽  
Pixel Resolution ◽  
European Organization ◽  
Finnish Meteorological Institute ◽  
Fractional Snow Cover ◽  
Turkish State ◽  
Sentinel 2

<p>The fractional snow cover (FSC) product H35 is a daily operational product based on multi-channel analysis of AVHRR onboard to NOAA and MetOp satellites. H35 is supplied by the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) Satellite Application Facility on Support to Operational Hydrology and Water Management (HSAF). The “traditional” H35 FSC product is generated at pixel resolution by exploiting the brightness intensity, which is the convolution of the snow signal and the fraction of snow within the pixel and the sampling is carried out at 1-km intervals. The product for flat/forested regions is generated by Finnish Meteorological Institute (FMI) and the product for mountainous areas is generated by Turkish State Meteorological Service (TSMS). Both products, thereafter, are merged at FMI. This presentation aims to represent the latest findings of our efforts in developing an “alternative” H35 FSC product for the mountainous part by using two data-driven machine learning methodologies, namely, multivariate adaptive regression splines (MARS) and random forests (RFs). In total, 332 Sentinel 2 images over Alps, Tatra Mountains and Turkey acquired between November 2018 and April 2019 are used in order to generate the necessary reference FSC maps for the training of the MARS and RF models. AVHRR bands 1-5, NDSI and NDVI are used as predictor variables. Binary classified Sentinel 2 snow maps, ERA5 snow depth and MODIS MOD10A1 NDSI data are employed in the validation of the models. The results show that both MARS- and RF-based H35 product are i) in good agreement with reference FSC maps (as indicated by low RMSE and relatively high R values) and ii) able to capture the spatial variability of the snow extend. However, MARS-based H35 is preferred for an operational FSC product generation due to the high computational cost required in RF model.</p>

Latest Advances in Fractional Snow Cover Mapping on MODIS Data by Machine Learning Algorithms

2020 ◽  
Author(s):  
Semih Kuter ◽  
Zuhal Akyurek
Keyword(s):  
Machine Learning ◽  
Snow Cover ◽  
General Circulation ◽  
Snow Water Equivalent ◽  
Machine Learning Algorithms ◽  
Multivariate Adaptive Regression Splines ◽  
Support Vector ◽  
Landsat 8 ◽  
European Alps ◽  
Fractional Snow Cover

<p>Spatial extent of snow has been declared as an essential climate variable. Accurate modeling of snow cover is crucial for the better prediction of snow water equivalent and, consequently, for the success of general circulation and weather forecasting models as well as climate change and hydrological studies. This presentation mainly focuses on the representation of the latest findings of our efforts in fractional snow cover mapping on MODIS images by data-driven machine learning methodologies. For this purpose, a dataset composed of 20 MODIS - Landsat 8 image pairs acquired between Apr 2013 and Dec 2016 over European Alps were employed. Artificial neural networks (ANN), multivariate adaptive regression splines (MARS), support vector regression (SVR) and random forest (RF) models were trained and tested by using reference FSC maps generated from higher spatial resolution Landsat 8 binary snow maps. ANN, MARS, SVR and RF models exhibited quite good performance with average R ≈ 0.93, whereas the agreement between the reference FSC maps and the MODIS’ own product MOD10A1 (C5) was slightly poorer with R ≈ 0.88.</p>

scholarly journals Gap-Filling of MODIS Fractional Snow Cover Products via Non-Local Spatio-Temporal Filtering Based on Machine Learning Techniques

2019 ◽  
Vol 11 (1) ◽  
pp. 90 ◽  
Author(s):  
Jinliang Hou ◽  
Chunlin Huang ◽  
Ying Zhang ◽  
Jifu Guo ◽  
Juan Gu
Keyword(s):  
Machine Learning ◽  
Snow Cover ◽  
Machine Learning Techniques ◽  
Commission Error ◽  
Gap Filling ◽  
Temporal Filtering ◽  
Spatial Efficiency ◽  
Fractional Snow Cover ◽  
Non Local ◽  
Spatio Temporal

Cloud obscuration leaves significant gaps in MODIS snow cover products. In this study, an innovative gap-filling method based on the concept of non-local spatio-temporal filtering (NSTF) is proposed to reconstruct the cloud gaps in MODIS fractional snow cover (SCF) products. The ground information of a gap pixel was estimated by using the appropriate similar pixels in the remaining known part of an image via an automatic machine learning technique. We take the MODIS SCF product cloud gap filling data from 2001 to 2016 in Northern Xinjiang, China as an example. The results demonstrate that the methodology can generate almost continuous spatio-temporal, daily MODIS SCF images, and it leaves only 0.52% of cloud gaps long-term, on average. The validation results based on “cloud assumption” exhibit high accuracy, with a higher R 2 exceeding 0.8, a lower RMSE of 0.1, an overestimated error of 1.13%, an underestimated error of 1.4%, and a spatial efficiency (SPAEF) of 0.78. The validation based on 50 in situ snow depth observations demonstrates the superiority of the methodology in terms of accuracy and consistency. The overall accuracy is 93.72%. The average omission and commission error have increased approximately 1.16 and 0.53% compared with the original MODIS SCF products under a clear sky term.

scholarly journals Evaluation of Methods for Mapping the Snow Cover Area at High Spatio-Temporal Resolution with VENμS

2020 ◽  
Vol 12 (18) ◽  
pp. 3058
Author(s):  
Mohamed Wassim Baba ◽  
Simon Gascoin ◽  
Olivier Hagolle ◽  
Elsa Bourgeois ◽  
Camille Desjardins ◽  
...  
Keyword(s):  
Machine Learning ◽  
Snow Cover ◽  
Temporal Resolution ◽  
Near Infrared ◽  
Snow Accumulation ◽  
Learning Method ◽  
Snow Cover Area ◽  
High Atlas ◽  
Detection Algorithms ◽  
Sentinel 2

The VENμS mission launched in 2017 provides multispectral optical images of the land surface with a 2-day revisit time at 5 m resolution for over 100 selected sites. A few sites are subject to seasonal snow accumulation, which gives the opportunity to monitor the variations of the snow cover area at unprecedented spatial and temporal resolution. However, the 12 spectral bands of VENμS only cover the visible and near-infrared region of the spectra while existing snow detection algorithms typically make use of a shortwave infrared band to determine the presence of snow. Here, we evaluate two alternative snow detection algorithms. The first one is based on a normalized difference index between the near-infrared and the visible bands, and the second one is based on a machine learning approach using the Theia Sentinel-2 snow products as training data. Both approaches are tested using Sentinel-2 data (as surrogate of VENμS data) as well as actual VENμS in the Pyrenees and the High Atlas. The results confirm the possibility of retrieving snow cover without SWIR with a slight loss in performance. As expected, the results confirm that the machine learning method provides better results than the index-based approach (e.g., an RMSE equal to the learning method 1.35% and for the index-based method 10.80% in the High Atlas.). The improvement is more evident in the Pyrenees probably due to the presence of vegetation which complicates the spectral signature of the snow cover area in VENμS images.

scholarly journals MODIS Fractional Snow Cover Mapping Using Machine Learning Technology in a Mountainous Area

2020 ◽  
Vol 12 (6) ◽  
pp. 962 ◽  
Author(s):  
Changyu Liu ◽  
Xiaodong Huang ◽  
Xubing Li ◽  
Tiangang Liang
Keyword(s):  
Machine Learning ◽  
Random Forest ◽  
Snow Cover ◽  
Back Propagation ◽  
Network Models ◽  
Support Vector ◽  
Learning Models ◽  
Neural Network Models ◽  
Fractional Snow Cover ◽  
Machine Learning Models

To improve the poor accuracy of the MODIS (Moderate Resolution Imaging Spectroradiometer) daily fractional snow cover product over the complex terrain of the Tibetan Plateau (RMSE = 0.30), unmanned aerial vehicle and machine learning technologies are employed to map the fractional snow cover based on MODIS over this terrain. Three machine learning models, including random forest, support vector machine, and back-propagation artificial neural network models, are trained and compared in this study. The results indicate that compared with the MODIS daily fractional snow cover product, the introduction of a highly accurate snow map acquired by unmanned aerial vehicles as a reference into machine learning models can significantly improve the MODIS fractional snow cover mapping accuracy. The random forest model shows the best accuracy among the three machine learning models, with an RMSE (root-mean-square error) of 0.23, especially over forestland and shrubland, with RMSEs of 0.13 and 0.18, respectively. Although the accuracy of the support vector machine and back-propagation artificial neural network models are worse over forestland and shrubland, their average errors are still better than that of MOD10A1. Different fractional snow cover gradients also affect the accuracy of the machine learning algorithms. Nevertheless, the random forest model remains stable in different fractional snow cover gradients and is, therefore, the best machine learning algorithm for MODIS fractional snow cover mapping in Tibetan Plateau areas with complex terrain and severely fragmented snow cover.

scholarly journals Estimating Fractional Snow Cover in Open Terrain from Sentinel-2 Using the Normalized Difference Snow Index

2020 ◽  
Author(s):  
Simon Gascoin ◽  
Zacharie Barrou Dumont ◽  
César Deschamps-Berger ◽  
Florence Marti ◽  
Germain Salgues ◽  
...  
Keyword(s):  
Snow Cover ◽  
Time Lapse ◽  
Global Scale ◽  
Terrestrial Lidar ◽  
Fractional Snow Cover ◽  
High Resolution Images ◽  
Normalized Difference Snow Index ◽  
Topographic Variability ◽  
Sentinel 2

Sentinel-2 provides the opportunity to map the snow cover at unprecedented spatial and temporal resolution at global scale. Here we calibrate and evaluate a simple empirical function to estimate the fractional snow cover (FSC) in open terrain using the normalized difference snow index (NDSI) from 20 m resolution Sentinel-2 images. The NDSI is computed from flat surface reflectances after masking cloud and snow-free areas. The NDSI-FSC function is calibrated using Pléiades very high resolution images and evaluated using independent datasets including SPOT 6/7 satellite images, time lapse camera photographs, terrestrial lidar scans and crowd-sourced in situ measurements. The calibration results show that the FSC can be represented with a sigmoid-shaped function 0.5×tanh(a×NDSI+b)+0.5 where a = 2.65 and b = -1.42 yielding a root mean square error of 25%. Similar RMSE are obtained with different evaluation datasets with a high topographic variability. With this function, we estimate that the confidence interval on the FSC retrievals is 38% at the 95% confidence level.

scholarly journals Estimating Fractional Snow Cover in Open Terrain from Sentinel-2 Using the Normalized Difference Snow Index

2020 ◽  
Vol 12 (18) ◽  
pp. 2904
Author(s):  
Simon Gascoin ◽  
Zacharie Barrou Dumont ◽  
César Deschamps-Berger ◽  
Florence Marti ◽  
Germain Salgues ◽  
...  
Keyword(s):  
Snow Cover ◽  
Time Lapse ◽  
Global Scale ◽  
Terrestrial Lidar ◽  
Fractional Snow Cover ◽  
High Resolution Images ◽  
Normalized Difference Snow Index ◽  
Topographic Variability ◽  
Sentinel 2

Sentinel-2 provides the opportunity to map the snow cover at unprecedented spatial and temporal resolutions on a global scale. Here we calibrate and evaluate a simple empirical function to estimate the fractional snow cover (FSC) in open terrains using the normalized difference snow index (NDSI) from 20 m resolution Sentinel-2 images. The NDSI is computed from flat surface reflectance after masking cloud and snow-free areas. The NDSI–FSC function is calibrated using Pléiades very high-resolution images and evaluated using independent datasets including SPOT 6/7 satellite images, time lapse camera photographs, terrestrial lidar scans and crowd-sourced in situ measurements. The calibration results show that the FSC can be represented with a sigmoid-shaped function 0.5 × tanh(a × NDSI + b) + 0.5, where a = 2.65 and b = −1.42, yielding a root mean square error (RMSE) of 25%. Similar RMSE are obtained with different evaluation datasets with a high topographic variability. With this function, we estimate that the confidence interval on the FSC retrievals is 38% at the 95% confidence level.

scholarly journals Estimating fractional snow cover from passive microwave brightness temperature data using MODIS snow cover product over North America

2021 ◽  
Vol 15 (2) ◽  
pp. 835-861
Author(s):  
Xiongxin Xiao ◽  
Shunlin Liang ◽  
Tao He ◽  
Daiqiang Wu ◽  
Congyuan Pei ◽  
...  
Keyword(s):  
North America ◽  
Land Cover ◽  
Brightness Temperature ◽  
Snow Cover ◽  
Temperature Data ◽  
Passive Microwave ◽  
Multivariate Adaptive Regression Splines ◽  
Spatiotemporal Resolution ◽  
Microwave Brightness Temperature ◽  
Fractional Snow Cover

Abstract. The dynamic characteristics of seasonal snow cover are critical for hydrology management, the climate system, and the ecosystem functions. Optical satellite remote sensing has proven to be an effective tool for monitoring global and regional variations in snow cover. However, accurately capturing the characteristics of snow dynamics at a finer spatiotemporal resolution continues to be problematic as observations from optical satellite sensors are greatly impacted by clouds and solar illumination. Traditional methods of mapping snow cover from passive microwave data only provide binary information at a spatial resolution of 25 km. This innovative study applies the random forest regression technique to enhanced-resolution passive microwave brightness temperature data (6.25 km) to estimate fractional snow cover over North America in winter months (January and February). Many influential factors, including land cover, topography, and location information, were incorporated into the retrieval models. Moderate Resolution Imaging Spectroradiometer (MODIS) snow cover products between 2008 and 2017 were used to create the reference fractional snow cover data as the “true” observations in this study. Although overestimating and underestimating around two extreme values of fractional snow cover, the proposed retrieval algorithm outperformed the other three approaches (linear regression, artificial neural networks, and multivariate adaptive regression splines) using independent test data for all land cover classes with higher accuracy and no out-of-range estimated values. The method enabled the evaluation of the estimated fractional snow cover using independent datasets, in which the root mean square error of evaluation results ranged from 0.189 to 0.221. The snow cover detection capability of the proposed algorithm was validated using meteorological station observations with more than 310 000 records. We found that binary snow cover obtained from the estimated fractional snow cover was in good agreement with ground measurements (kappa: 0.67). There was significant improvement in the accuracy of snow cover identification using our algorithm; the overall accuracy increased by 18 % (from 0.71 to 0.84), and the omission error was reduced by 71 % (from 0.48 to 0.14) when the threshold of fractional snow cover was 0.3. The experimental results show that passive microwave brightness temperature data may potentially be used to estimate fractional snow cover directly in that this retrieval strategy offers a competitive advantage in snow cover detection.

Sign in / Sign up

Export Citation Format

Share Document