Assessment of machine learning classifiers for global lake ice cover mapping from MODIS TOA reflectance data

Wu, Yuhao; Duguay, Claude R.; Xu, Linlin

doi:10.1016/j.rse.2020.112206

Cited by 50 publications

(32 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, a manual procedure is a heavy task, especially if SIMBA operation covers a long period or one would need real-time SIMBA results. Several studies have been carried out aimed at development of an algorithm to obtain snow depth and ice thickness automatically (Liao et al, 2019;Zuo et al, 2018;Cheng et al, 2020). Below we present an example of the application of the Cheng et al (2020) algorithm to retrieve snow depth and ice thickness from SIMBA data observed in Lake Orajärvi.…”

Section: Simba Snow Depth and Ice Thicknessmentioning

confidence: 99%

Inter-annual variation in lake ice composition in the European Arctic: observations based on high-resolution thermistor strings

Cheng

Vihma

et al. 2021

Earth Syst. Sci. Data

View full text Add to dashboard Cite

Abstract. Climate change and global warming strongly impact the cryosphere. The rise of air temperature and change of precipitation patterns lead to dramatic responses of snow and ice heat and mass balance. Sustainable field observations on lake air–snow–ice–water temperature regime have been carried out in Lake Orajärvi in the vicinity of the Finnish Space Centre, a Flagship Supersite in Sodankylä in Finnish Lapland since 2009. A thermistor-string-based snow and ice mass balance buoy called “Snow and ice mass balance apparatus (SIMBA)” was deployed in the lake at the beginning of each ice season. In this paper, we describe snow and ice temperature regimes, snow depth, ice thickness, and ice compositions retrieved from SIMBA observations as well as meteorological variables based on high-quality observations at the Finnish Space Centre. Ice thickness in Lake Orajärvi showed an increasing trend. During the decade of data collection (1) the November–May mean air temperature had an increasing trend of 0.16 ∘C per year, and the interannual variations were highly correlated (r = 0.93) with the total seasonal accumulated precipitation; (2) the maximum granular ice thickness ranged from 15 % to 80 % of the maximum total ice thickness; and (3) the snow depth on lake ice was not correlated (r = 0.21) with the total precipitation. The data set can be applied to investigate the lake ice surface heat balance and the role of snow in lake ice mass balance and to improve the parameterization of snow to ice transformation in snow and ice models. The data are archived at https://doi.org/10.5281/zenodo.4559368 (Cheng et al., 2021).

show abstract

Section: Simba Snow Depth and Ice Thicknessmentioning

confidence: 99%

Inter-annual variation in lake ice composition in the European Arctic: observations based on high-resolution thermistor strings

Cheng

Vihma

et al. 2021

Earth Syst. Sci. Data

View full text Add to dashboard Cite

show abstract

“…Interestingly, even for Sentinel-1 imagery it proved beneficial to employ a network pre-trained on close-range RGB data. Very recently Wu et al (2021) compared the capabilities of four different ML methodologies: Multinomial Logistic Regression (MLR), SVM, RF, and Gradient Boosting Trees (GBT) for lake ice observation using MODIS Top of Atmosphere (TOA) product. They modelled lake ice monitoring as a 3-class (ice, water, cloud) supervised classification problem.…”

Section: Lake Ice Observation With Machine and Deep Learningmentioning

confidence: 99%

Recent Ice Trends in Swiss Mountain Lakes: 20-year Analysis of MODIS Imagery

Tom¹,

Wu²,

Baltsavias³

et al. 2021

Preprint

View full text Add to dashboard Cite

Depleting lake ice can serve as an indicator for climate change, just like sea level rise or glacial retreat. Several Lake Ice Phenological (LIP) events serve as sentinels to understand the regional and global climate change. Hence, monitoring the long-term lake freezing and thawing patterns can prove very useful. In this paper, we focus on observing the LIP events such as freeze-up, break-up and temporal freeze extent in the Oberengadin region of Switzerland, where there are several small-and medium-sized mountain lakes, across two decades (2000-2020) from optical satellite images. We analyse time-series of MODIS imagery (and additionally cross-check with VIIRS data when available), by estimating spatially resolved maps of lake ice for these Alpine lakes with supervised machine learning. To train the classifier we rely on reference data annotated manually based on publicly available webcam images. From the ice maps we derive long-term LIP trends. Since the webcam data is only available for two winters, we also validate our results against the operational MODIS and VIIRS snow products. We find a change in Complete Freeze Duration (CFD) of -0.76 and -0.89 days per annum (d/a) for lakes Sils and Silvaplana respectively. Furthermore, we correlate the lake freezing and thawing trends with climate data such as temperature, sunshine, precipitation and wind measured at nearby meteorological stations.

show abstract

“…The lake ice phenology was extracted using the convolutional neural network method [12]. The lake ice phenology was mapped using MODIS/Terra L1B TOA (MOD02) product based on four machine learning classifiers (multinomial logistic regression, MLR; support vector machine, SVM; random forest, RF; gradient boosting trees, GBT) [13]. The sentinel-2 and auxiliary TanDEM-X topographic data were trained for automated mapping of Antarctic supraglacial lakes using RF [14].…”

Section: Introductionmentioning

confidence: 99%

“…In recent years, machine learning has been adopted for remote sensing applications, including neural networks, SVM, and RF [19,20]. RF was relatively insensitive to the choice of the hyperparameters compared to the other classifiers [13,21]. The RF algorithm is being increasingly applied to satellite and aerial image classification and continuous field datasets.…”

Section: Introductionmentioning

confidence: 99%

Spatial-Temporal Distribution of the Freeze–Thaw Cycle of the Largest Lake (Qinghai Lake) in China Based on Machine Learning and MODIS from 2000 to 2020

Han

Huang

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

The lake ice phenology variations are vital for the land–surface–water cycle. Qinghai Lake is experiencing amplified warming under climate change. Based on the MODIS imagery, the spatio-temporal dynamics of the ice phenology of Qinghai Lake were analyzed using machine learning during the 2000/2001 to 2019/2020 ice season, and cloud gap-filling procedures were applied to reconstruct the result. The results showed that the overall accuracy of the water–ice classification by random forest and cloud gap-filling procedures was 98.36% and 92.56%, respectively. The annual spatial distribution of the freeze-up and break-up dates ranged primarily from DOY 330 to 397 and from DOY 70 to 116. Meanwhile, the decrease rates of freeze-up duration (DFU), full ice cover duration (DFI), and ice cover duration (DI) were 0.37, 0.34, and 0.13 days/yr., respectively, and the duration was shortened by 7.4, 6.8, and 2.6 days over the past 20 years. The increased rate of break-up duration (DBU) was 0.58 days/yr. and the duration was lengthened by 11.6 days. Furthermore, the increase in temperature resulted in an increase in precipitation after two years; the increase in precipitation resulted in the increase in DBU and decrease in DFU in corresponding years, and decreased DI and DFI after one year.

show abstract

Assessment of machine learning classifiers for global lake ice cover mapping from MODIS TOA reflectance data

Cited by 50 publications

References 60 publications

Inter-annual variation in lake ice composition in the European Arctic: observations based on high-resolution thermistor strings

Inter-annual variation in lake ice composition in the European Arctic: observations based on high-resolution thermistor strings

Recent Ice Trends in Swiss Mountain Lakes: 20-year Analysis of MODIS Imagery

Spatial-Temporal Distribution of the Freeze–Thaw Cycle of the Largest Lake (Qinghai Lake) in China Based on Machine Learning and MODIS from 2000 to 2020

Contact Info

Product

Resources

About