On Drivers of Subpixel Classification Accuracy—An Example From Glacier Facies

Yousuf, Bisma; Shukla, Aparna; Arora, Manoj K.; Bindal, Ankit; Jasrotia, Avtar Singh

doi:10.1109/jstars.2019.2955955

Cited by 13 publications

(4 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, machine-learning-based classification methods have been applied to mapping glacier facies (Racoviteanu and Williams, 2012;Shukla and Yousuf, 2016;Zhang et al, 2019;Yousuf et al, 2020). Studies have shown that machine learning has advantages in extracting land-surface information from remote sensing images, which can effectively improve the accuracy of object recognition (Lary et al, 2017;Maxwell et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

Upward Expansion of Supra-Glacial Debris Cover in the Hunza Valley, Karakoram, During 1990 ∼ 2019

Xie

Liu

et al. 2020

Front. Earth Sci.

View full text Add to dashboard Cite

Supra-glacial debris cover is key to glacier ablation through increasing (thin debris layer) or decreasing (thick debris layer) melt rates, thereby regulating the mass balance of a glacier and its meltwater runoff. The thickening or lateral expansion of supra-glacial debris cover correlates with a reduction of glacier ablation and, consequently, runoff generation, which is also considered to be an influential factor on the rheology and dynamics of a glacierized system. Studies on supra-glacial debris cover have recently attracted wide attention especially for glaciers in the Himalayas and Karakoram, where the glaciers have heterogeneously responded to climate change. In this study, we used 32 images from the Landsat Thematic Mapper, Enhanced Thematic Mapper Plus, and Operational Land Imager archive, going back to 1990, which are available on the Google Earth Engine cloud-computing platform, to map the supra-glacial debris cover in the Hunza Valley, Karakoram, Pakistan, based on a band ratio segmentation method (normalized difference snow index [NDSI] < 0.4), Otsu thresholding, and machine learning algorithms. Compared with manual digitization, the random forest (RF) model was found to have the greatest accuracy in identifying supra-glacial debris, with a Kappa coefficient of 0.94 ± 0.01 and an overall accuracy of 95.5 ± 0.9%. Overall, the supraglacial debris cover in the study area showed an increasing trend, and the total area expanded by 8.1-21.3% for various glaciers from 1990 to 2019. The other two methods (Otsu thresholding and NDSI < 0.4) generally overestimated the supra-glacial debris covered area, by 36.3 and 18.8%, respectively, compared to that of the RF model. The supra-glacial debris cover has migrated upward on the glaciers, with intensive variation near the equilibrium-line altitude zone (4,500-5,500 m a.s.l.). The increase in ice or snow avalanche activity at high altitudes may be responsible for this upward expansion of supra-glacial debris cover in the Hunza Valley, which is attributed to the combined effect of temperature decrease and precipitation increase in the study area.

show abstract

Section: Introductionmentioning

confidence: 99%

Upward Expansion of Supra-Glacial Debris Cover in the Hunza Valley, Karakoram, During 1990 ∼ 2019

Xie

Liu

et al. 2020

Front. Earth Sci.

View full text Add to dashboard Cite

show abstract

“…Machine learning-based classification methods have been applied to measure glacier changes (Huang et al, 2011;Xie et al, 2020;Yousuf et al, 2020), and these methods can effectively improve the accuracy of identifying features on the glaciers and surrounding surfaces (Maxwell et al, 2018). However, machine learning requires a large amount of remote sensing data with a high spatial and temporal resolution for mining adequate information, which places high demands on data storage, computing, and analysis.…”

Section: Introductionmentioning

confidence: 99%

Measuring glacier changes in the Tianshan Mountains over the past 20 years using Google Earth Engine and machine learning

Zhuang,

Ke,

Cai

et al. 2023

J. Geogr. Sci.

View full text Add to dashboard Cite

Glaciers in the Tianshan Mountains are an essential water resource in Central Asia, and it is necessary to identify their variations at large spatial scales with high resolution. We combined optical and SAR images, based on several machine learning algorithms and ERA-5 land data provided by Google Earth Engine, to map and explore the glacier distribution and changes in the Tianshan in 2001, 2011, and 2021. Random forest was the best performing classifier, and the overall glacier area retreat rate showed acceleration from 0.87%/a to 1.49%/a, while among the sub-regions, Dzhungarsky Alatau, Central and Northern/Western Tianshan, and Eastern Tianshan showed a slower, stable, and sharp increase rates after 2011, respectively. Glacier retreat was more severe in the mountain periphery, low plains and valleys, with more area lost near the glacier equilibrium line. The sustained increase in summer temperatures was the primary driver of accelerated glacier retreat. Our work demonstrates the advantage and reliability of fusing multisource images to map glacier distributions with high spatial and temporal resolutions using Google Earth Engine. Its high recognition accuracy helped to conduct more accurate and time-continuous glacier change studies for the study area.

show abstract

“…Novel approaches have emerged since 2010 to automate the mapping of supraglacial debris, namely, shallow architecture machine learning algorithms such as Support Vector Machine (Huang et al, 2014;Yousef et al, 2020;Shukla et al, 2022), Maximum Likelihood Classifier (Shukla et al, 2010), Artificial Neural Networks (Karimi et al, 2012), and Random Forest Classifier Alifu et al, 2020;Khan et al, 2020;Lu et al, 2020). The application of Convolutional Neural Networks (CNNs), a member of the deep learning classifier family within machine learning, to delineate supraglacial debris extents has been successfully experimented with in a few studies (Nijhawan et al, 2018;Xie et al, 2020;Lu et al, 2021;Xie et al, 2021;Tian et al, 2022;Xie et al, 2022).…”

Section: Introductionmentioning

confidence: 99%

An integrated deep learning and object-based image analysis approach for mapping debris-covered glaciers

Thomas

Robson

Racoviteanu

2023

Front. Remote Sens.

View full text Add to dashboard Cite

Evaluating glacial change and the subsequent water stores in high mountains is becoming increasingly necessary, and in order to do this, models need reliable and consistent glacier data. These often come from global inventories, usually constructed from multi-temporal satellite imagery. However, there are limitations to these datasets. While clean ice can be mapped relatively easily using spectral band ratios, mapping debris-covered ice is more difficult due to the spectral similarity of supraglacial debris to the surrounding terrain. Therefore, analysts often employ manual delineation, a time-consuming and subjective approach to map debris-covered ice extents. Given the increasing prevalence of supraglacial debris in high mountain regions, such as High Mountain Asia, a systematic, objective approach is needed. The current study presents an approach for mapping debris-covered glaciers that integrates a convolutional neural network and object-based image analysis into one seamless classification workflow, applied to freely available and globally applicable Sentinel-2 multispectral, Landsat-8 thermal, Sentinel-1 interferometric coherence, and geomorphometric datasets. The approach is applied to three different domains in the Central Himalayan and the Karakoram ranges of High Mountain Asia that exhibit varying climatic regimes, topographies and debris-covered glacier characteristics. We evaluate the performance of the approach by comparison with a manually delineated glacier inventory, achieving F-score classification accuracies of 89.2%–93.7%. We also tested the performance of this approach on declassified panchromatic 1970 Corona KH-4B satellite imagery in the Manaslu region of Nepal, yielding accuracies of up to 88.4%. We find our approach to be robust, transferable to other regions, and accurate over regional (>4,000 km2) scales. Integrating object-based image analysis with deep-learning within a single workflow overcomes shortcomings associated with convolutional neural network classifications and permits a more flexible and robust approach for mapping debris-covered glaciers. The novel automated processing of panchromatic historical imagery, such as Corona KH-4B, opens the possibility of exploiting a wealth of multi-temporal data to understand past glacier changes.

show abstract

On Drivers of Subpixel Classification Accuracy—An Example From Glacier Facies

Cited by 13 publications

References 33 publications

Upward Expansion of Supra-Glacial Debris Cover in the Hunza Valley, Karakoram, During 1990 ∼ 2019

Upward Expansion of Supra-Glacial Debris Cover in the Hunza Valley, Karakoram, During 1990 ∼ 2019

Measuring glacier changes in the Tianshan Mountains over the past 20 years using Google Earth Engine and machine learning

An integrated deep learning and object-based image analysis approach for mapping debris-covered glaciers

Contact Info

Product

Resources

About