Accuracy assessment of MODIS,NOAA and IRS data in snow cover mapping under Himalayan conditions

Snow cover information is an essential parameter for a wide variety of scientific studies and management applications, especially in snowmelt runoff modelling. Until now NOAA and IRS data were widely and effectively used for snow‐covered area (SCA) estimation in several Himalayan basins. The suit of snow cover products produced from MODIS data had not previously been used in SCA estimation and snowmelt runoff modelling in any Himalayan basin. The present study was conducted with the aim of assessing the accuracy of MODIS, NOAA and IRS data in snow cover mapping under Himalayan conditions. The total SCA was estimated using these three datasets for 15 dates spread over 4 years. The results were compared with ground‐based estimation of snow cover. A good agreement was observed between satellite‐based estimation and ground‐based estimation. The influence of aspect in SCA estimation was analysed for the three satellite datasets and it was observed that MODIS produced better results. Snow mapping accuracy with respect to elevation was tested and it was observed that at higher elevation MODIS sensed more snow and proved better at mapping snow under mountain shadow conditions. At lower elevation, IRS proved better in mapping patchy snow cover due to higher spatial resolution. The temporal resolution of MODIS and NOAA data is better than IRS data, which means that the chances of getting cloud‐free scenes is higher. In addition, MODIS has an automated snow‐mapping algorithm, which reduces the time and errors incorporated during processing satellite data manually. Considering all these factors, it was concluded that MODIS data could be effectively used for SCA estimation under Himalayan conditions, which is a vital parameter for snowmelt runoff estimation.     

Estimating fractional snow cover from MODIS using the normalized difference snow index

Snow-cover information is important for a wide variety of scientific studies, water supply and management applications. The NASA Earth Observing System (EOS) Moderate Resolution Imaging Spectroradiometer (MODIS) provides improved capabilities to observe snow cover from space and has been successfully using a normalized difference snow index (NDSI), along with threshold tests, to provide global, automated binary maps of snow cover. The NDSI is a spectral band ratio that takes advantage of the spectral differences of snow in short-wave infrared and visible MODIS spectral bands to identify snow versus other features in a scene. This study has evaluated whether there is a “signal” in the NDSI that could be used to estimate the fraction of snow within a 500 m MODIS pixel and thereby enhance the use of the NDSI approach in monitoring snow cover. Using Landsat 30-m observations as “ground truth,” the percentage of snow cover was calculated for 500-m cells. Then a regression relationship between 500-m NDSI observations and fractional snow cover was developed over three different snow-covered regions and tested over other areas. The overall results indicate that the relationship between fractional snow cover and NDSI is reasonably robust when applied locally and over large areas like North America. The relationship offers advantages relative to other published fractional snow cover algorithms developed for global-scale use with MODIS. This study indicates that the fraction of snow cover within a MODIS pixel using this approach can be provided with a mean absolute error less than 0.1 over the range from 0.0 to 1.0 in fractional snow cover.     

Mapping dry/wet snow cover in the Indian Himalayas using IRS multispectral imagery

Application of remote sensing data has been made to differentiate between dry/wet snows in a glacierized basin. The present study has been carried out in the Gangotri glacier, Himalayas, using IRS-LISS-III multispectral data for the period March-November 2000 and the digital elevation model. The methodology involves conversion of satellite sensor data into reflectance values, computation of NDSI, determination of the boundary between dry/wet snows from spectral response data, and threshold slicing of the image data. The areas of dry snow cover and wet snow cover for different dates of satellite overpasses have been computed. The dry snow area has been compared with non-melting area obtained from the temperature lapse rate method, and the two are found to be in close mutual correspondence (< 15%). It is observed that there occur four water-bearing zones in the glacierized basin: dry snow zone, wet snow zone, exposed glacial ice and moraine-covered glacial ice, each of which possesses unique hydrological characteristics and can be distinguished and mapped from satellite sensor data. It is suggested that input of data on the position and extent of specifically wet snow and exposed glacial ice, which can be directly derived from remote sensing, should improve hydrological simulation of such basins.     

Subpixel mapping of snow cover in forests by optical remote sensing

Forest represents a challenging problem for snow-cover mapping by optical satellite remote sensing. To investigate reflectance variability and to improve the mapping of snow in forested areas, a method for subpixel mapping of snow cover in forests (SnowFrac) has been developed. The SnowFrac method is based on linear spectral mixing modelling of snow, trees and snow-free ground. The focus has been on developing a physically based reflectance model which uses a forest-cover map as prior information. The method was tested in flat terrain covered by spruce, pine and birch forests, close to the Jotunheimen region of South Norway. Experiments were carried out using a completely snow-covered Landsat Thematic Mapper (TM) scene, aerial photos and in situ reflectance measurements. A detailed forest model was photogrammetrically derived from the aerial photos. Modelled and observed TM reflectances were compared. In the given situation, the results demonstrate that snow and individual tree species, in addition to cast shadows on the snow surface from single trees, are the most influencing factors on visible and near-infrared reflectance. Modelling of diffuse radiation reduced by surrounding trees slightly improve the results, indicating that this effect is less important. The best results are obtained for pine forest and mixed pine and birch forest. Future work will focus on deriving a simplified reflectance model suitable for operational snow-cover mapping in forests.     

Potential operational multi-satellite sensor mapping of snow cover in maritime sub-polar regions

Scottish snow cover as an example of a maritime sub-polar region has two principal problems for an operational monitoring programme: the often ephemeral nature of the snow cover, and the loss of direct access to snow imagery due to clouds. At present only the NOAA AVHRR series provides images with the required temporal and spatial resolutions. Based on the availability of data from the Dundee satellite data receiving station a range of NOAA-12, -14 and -15 day and night passes were collected and processed. Three snow cover products were produced from the NOAA AVHRR/2 data: snow area based on channel 134 ISODATA classifications, percentage snow cover based on multi-temporal Normalized Difference Vegetation Index (NDVI), and daily maximum snow surface temperature maps using split-window combinations of thermal channels. Noted improvements were evident in the accuracy and resolution of snow cover classifications based on provisional testing of AVHRR/3 data. Maximum snow surface temperature maps indicated a potential for mapping areas of snow melt. The principal limitation in the operational snow cover mapping with AVHRR, however, remains the loss of temporal resolution due to cloud cover.     

A new water-resistant snow index for the detection and mapping of snow cover on a global scale

An up-to-date spatio-temporal change analysis of global snow cover is essential for better understanding of climate–hydrological interactions. The normalized difference snow index (NDSI) is a widely used algorithm for the detection and estimation of snow cover. However, NDSI cannot discriminate between snow cover and water bodies without use of an external water mask. A stand-alone methodology for robust detection and mapping of global snow cover is presented by avoiding external dependency on the water mask. A new spectral index called water-resistant snow index (WSI) with the capability of exhibiting significant contrast between snow cover and other cover types, including water bodies, was developed. WSI uses the normalized difference between the value and hue obtained by transforming red, green, and blue, (RGB) colour composite images comprising red, green, and near-infrared bands into a hue, saturation, and value (HSV) colour model. The superiority of WSI over NDSI is confirmed by case studies conducted in major snow regions globally. Snow cover was mapped by considering monthly variation in snow cover and availability of satellite data at the global scale. A snow cover map for the year 2013 was produced at the global scale by applying the random walker algorithm in the WSI image supported by the reference data collected from permanent snow-covered and non-snow-covered areas. The resultant snow-cover map was compared to snow cover estimated by existing maps: MODIS Land Cover Type Product (MCD12Q1 v5.1, 2012), Global Land Cover by National Mapping Organizations (GLCNMO v2.0, 2008), and European Space Agency’s GlobCover 2009. A significant variation in snow cover as estimated by different maps was noted, and was was attributed to methodological differences rather than annual variation in snow cover. The resultant map was also validated with reference data, with 89.46% overall accuracy obtained. The WSI proposed in the research is expected to be suitable for seasonal and annual change analysis of global snow cover.     

Fractional snow cover mapping through artificial neural network analysis of MODIS surface reflectance

Accurate areal measurements of snow cover extent are important for hydrological and climate modeling. The traditional method of mapping snow cover is binary where a pixel is considered either snow-covered or snow-free. Fractional snow cover (FSC) mapping can achieve a more precise estimate of areal snow cover extent by estimating the fraction of a pixel that is snow-covered. The most common snow fraction methods applied to Moderate Resolution Imaging Spectroradiometer (MODIS) images have been spectral unmixing and an empirical Normalized Difference Snow Index (NDSI). Machine learning is an alternative for estimating FSC as artificial neural networks (ANNs) have been successfully used for estimating the subpixel abundances of other surfaces. The advantages of ANNs are that they can easily incorporate auxiliary information such as land cover type and are capable of learning nonlinear relationships between surface reflectance and snow fraction. ANNs are especially applicable to mapping snow cover extent in forested areas where spatial mixing of surface components is nonlinear. This study developed a multilayer feed-forward ANN trained through backpropagation to estimate FSC using MODIS surface reflectance, NDSI, Normalized Difference Vegetation Index (NDVI) and land cover as inputs. The ANN was trained and validated with higher spatial-resolution FSC maps derived from Landsat Enhanced Thematic Mapper Plus (ETM+) binary snow cover maps. Testing of the network was accomplished over training and independent test areas. The developed network performed adequately with RMSE of 12% over training areas and slightly less accurately over the independent test scenes with RMSE of 14%. The developed ANN also compared favorably to the standard MODIS FSC product. The study also presents a comprehensive validation of the standard MODIS snow fraction product whose performance was found to be similar to that of the ANN.     

A hybrid of deep learning and hand-crafted features based approach for snow cover mapping

Monitoring the extent of snow cover plays a vital role for a better understanding of current and future climatic, ecological, and water cycle conditions. Previously, several traditional machine learning models have been applied for accomplishing this while exploring a variety of feature extraction techniques on various information sources. However, the laborious process of any amount of hand-crafted feature extraction has not helped to obtain high accuracies. Recently, deep learning models have shown that feature extraction can be made automatic and that they can achieve the required high accuracies but at the cost of requiring a large amount of labelled data. Fortunately, despite the absence of such large amounts of labelled data for this task, we can rely on pre-trained models, which accept red-green-blue (RGB) information (or dimensions-reduced spectral data). However, it is always better to include a variety of information sources to solve any problem, especially with the availability of other important information sources like synthetic aperture radar (SAR) imagery and elevation. We propose a hybrid model where the deep learning is assisted by these information sources which have until now been left out. Particularly, our model learns from both the deep learning features (derived from spectral data) and the hand-crafted features (derived from SAR and elevation). Such an approach shows interesting performance-improvement from 96.02% (through deep learning alone) to 98.10% when experiments were conducted for Khiroi village of the Himalayan region in India.     

Toward improved daily snow cover mapping with advanced combination of MODIS and AMSR-E measurements

Taking three snow seasons from November 1 to March 31 of year 2002 to 2005 in northern Xinjiang, China as an example, this study develops a new daily snow cover product (500 m) through combining MODIS daily snow cover data and AMSR-E daily snow water equivalent (SWE) data. By taking advantage of both high spatial resolution of optical data and cloud transparency of passive microwave data, the new daily snow cover product greatly complements the deficiency of MODIS product when cloud cover is present especially for snow cover product on a daily basis and effectively improves daily snow detection accuracy. In our example, the daily snow agreement of the new product with the in situ measurements at 20 stations is 75.4%, which is much higher than the 33.7% of the MODIS daily product in all weather conditions, even a little higher than the 71% of the MODIS 8-day product (cloud cover of ~ 5%). Our results also indicate that i) AMSR-E daily SWE imagery generally agrees with MOD10A1 data in detecting snow cover, with overall agreement of 93.4% and snow agreement of 96.6% in the study area; ii) AMSR-E daily SWE imagery underestimates the snow covered area (SCA) due to its coarse spatial resolution; iii) The new snow cover product can better and effectively capture daily SCA dynamics during the snow seasons, which plays a significant role in reduction, mitigation, and prevention of snow-caused disasters in pastoral areas.     

On the contribution of volume scattering to the microwave backscattered signal from wet snow and wet soil

Abstract

The contribution of volume scattering from the subsurface stratum of soil or snow to the backscattered radar signal decreases rapidly with increasing moisture content as compared to the contribution by rough-surface scattering. In order to obtain an evaluation of the relative importance of volume scattering of moist strata, a heterogeneous two-component mixture of water inclusions in a dry background medium is assumed and the formulation of the distorted Born approximation according to Tsang et al. is applied. Computed permittivities and backscatter cross-sections are compared with experimental data of situations with modest moisture content. From there, conclusions on the relative portion of volume scattering and on shape and size of the effectively-scattering water inclusions can be drawn.     

Spatiotemporal variability of snow cover on Tibet,China using MODIS remote-sensing data

ABSTRACT

Snow cover is an important component of the cryosphere, and the study on spatial and temporal variations of snow cover is essential for understanding the consequences and impacts of climate change and water resources management. In this study, the temporal variation of snow-covered area (SCA) and spatial variability of snow-cover frequency (SCF) on Tibet is analysed based on the Moderate Resolution Imaging Spectroradiometer (MODIS)/Terra snow cover product (MOD10A2) from 2000 to 2015, and relationships with main climate variables are investigated. Results are as follows: (1) there is a very weak decreasing trend in annual mean SCA, and a slight increasing trend in autumn and winter and a slight decreasing trend in spring and more robust decreasing trend in summer for SCA are found. (2) The temporal variation of SCA is negatively correlated with temperature, whereas it is little correlated with corresponding precipitation. (3) The general trend of spatial SCF variation on Tibet, predominated by snow-cover variations in spring and autumn, tends to decrease in spring while it tends to increase in autumn. (4) The spatial variability of SCF is attributed to snow-cover variations in autumn and spring, which is more obvious in higher latitudes in autumn while it is more noticeable in lower-latitude southeastern plateau in spring. (5) The regions with higher variability of snow cover are main pastoral land and more prone to snow-related disaster in Tibet, becoming key zone of snow-cover monitoring and disaster prevention and mitigation.     

Accuracy of land cover area estimated from coarse spatial resolution images using an unmixing method

Many techniques intended to estimate land coverage of multiple categories occupied within each pixel from such coarse resolution data have been proposed. However, in traditional unmixing studies with coarse resolution imagery such as Advanced Very High Resolution Radiometer (AVHRR) data, it is assumed that only a few endmembers exist throughout an entire image. Therefore, it is essential to evaluate how well an unmixing method would work for various categories within pixels of coarse resolution images. In this study, the land coverage of eight classes in National Oceanic and Atmospheric Administration (NOAA) AVHRR imagery by using finer resolution Landsat Thematic Mapper (TM) imagery was estimated, and the accuracy of these estimated classes was evaluated. The results suggest that this method may be generally useful for comparing multi-spectral images in space and time.     

Hierarchical mapping of Northern Eurasian land cover using MODIS data

The Northern Eurasian land mass encompasses a diverse array of land cover types including tundra, boreal forest, wetlands, semi-arid steppe, and agricultural land use. Despite the well-established importance of Northern Eurasia in the global carbon and climate system, the distribution and properties of land cover in this region are not well characterized. To address this knowledge and data gap, a hierarchical mapping approach was developed that encompasses the study area for the Northern Eurasia Earth System Partnership Initiative (NEESPI). The Northern Eurasia Land Cover (NELC) database developed in this study follows the FAO-Land Cover Classification System and provides nested groupings of land cover characteristics, with separate layers for land use, wetlands, and tundra. The database implementation is substantially different from other large-scale land cover datasets that provide maps based on a single set of discrete classes. By providing a database consisting of nested maps and complementary layers, the NELC database provides a flexible framework that allows users to tailor maps to suit their needs. The methods used to create the database combine empirically derived climate–vegetation relationships with results from supervised classifications based on Moderate Resolution Imaging Spectroradiometer (MODIS) data. The hierarchical approach provides an effective framework for integrating climate–vegetation relationships with remote sensing-based classifications, and also allows sources of error to be characterized and attributed to specific levels in the hierarchy. The cross-validated accuracy was 73% for the land cover map and 73% and 91% for the agriculture and wetland classifications, respectively. These results support the use of hierarchical classification and climate–vegetation relationships for mapping land cover at continental scales.     

Accuracy assessment using sub-pixel fractional error matrices of global land cover products derived from satellite data

Information on land cover distribution at regional and global scales has become fundamental for studying global changes affecting ecological and climatic systems. The remote sensing community has responded to this increased interest by improving data quality and methodologies for extracting land cover information. However, in addition to the advantages provided by satellite products, certain limitations exist that need to be objectively quantified and clearly communicated to users so that they can make informed decisions on whether and how land cover products should be used. Accuracy assessment is the procedure used to quantify product quality. Some aspects of accuracy assessment for evaluating four global land cover maps over Canada are discussed in this paper. Attempts are made to quantify limiting factors resulting from the coarse spatial resolution of data used for generating land cover information at regional and global levels. Sub-pixel fractional error matrices are introduced as a more appropriate way for assessing the accuracy of mixed pixels. For classification with coarse spatial resolution data, limitations of the classification method produce a maximum achievable accuracy defined as the average percent fraction of dominant land cover of all pixels in the mapped area. Relationships among spatial resolution, landscape heterogeneity and thematic resolution were studied and reported. Other factors that can affect accuracy, such as misregistration and legend conversion, are also discussed.     

Comparison of methods of snow cover mapping by analysing the solar spectrum of satellite remote sensing data in China

Three methods, supervised classification (SC), digital number (DN) statistics and Normalized Difference Snow Index (NDSI), are used to map snow cover and then calculate snow cover area. Data sets from Landsat TM, Moderate Resolution Imaging Spectroradiometer (MODIS) and NOAA/AVHRR are selected because these sensors of different spatial resolution provide the most up to date remote sensing data for China. The results show that the best method for obtaining the snow index is different for each of these sensor products because of their different spatial and temporal resolutions and objectives of application. Reflectivity threshold statistics (RTs) should be used if the data series is incomplete; whereas SC needs a relatively accurate signature file for classification. A valid and rational method has been certified which selects NDSI for extracting snow pixels. Meanwhile, we introduce the brightness compensation method for decreasing the impact of topographic shading on distinguishing of snow pixels.     

Large-scale monitoring of snow cover and runoff simulation in Himalayan river basins using remote sensing

Various remote sensing products are used to identify spatial-temporal trends in snow cover in river basins originating in the Himalayas and adjacent Tibetan-Qinghai plateau. It is shown that remote sensing allows detection of spatial-temporal patterns of snow cover across large areas in inaccessible terrain, providing useful information on a critical component of the hydrological cycle. Results show large variation in snow cover between years while an increasing trend from west to east is observed. Of all river basins the Indus basin is, for its water resources, most dependent on snow and ice melt and large parts are snow covered for prolonged periods of the year. A significant negative winter snow cover trend was identified for the upper Indus basin. For this basin a hydrological model is used and forced with remotely sensed derived precipitation and snow cover. The model is calibrated using daily discharges from 2000 to 2005 and stream flow in the upper Indus basin can be predicted with a high degree of accuracy. From the analysis it is concluded that there are indications that regional warming is affecting the hydrology of the upper Indus basin due to accelerated glacial melting during the simulation period. This warming may be associated with global changes in air temperature resulting from anthropogenic forcings. This conclusion is primarily based on the observation that the average annual precipitation over a five year period is less than the observed stream flow and supported by positive temperature trends in all seasons.     

Mapping snow cover in the pan-Arctic zone,using multi-year (1998-2001) images from optical VEGETATION sensor

Timely information on spatial distribution and temporal dynamics of snow cover in the pan-Arctic zone is needed, as snow cover plays an important role in climate, hydrology and ecological processes. Here we report estimates of snow cover in the pan-Arctic zone (north of 45° N) at 1-km spatial resolution and at a 10-day temporal interval over the period of April 1998 to December 2001, using 10-day composite images of VEGETATION sensor onboard Système Pour l'Observation de la Terre (SPOT)-4 satellite. The results show that snow covered area (SCA) in North America (north of 45° N) increased from 1998 to 2001, while SCA in Eurasia (north of 45° N) decreased from 1998 to 2000 but increased in 2001. There were large spatial and temporal variations of snow cover in the pan-Arctic zone during 1998-2001.     

Automating land cover mapping of Scotland using expert system and knowledge integration methods

Effective land cover mapping often requires the use of multiple data sources and data interpretation methods, particularly when no one data source or interpretation method provides sufficiently good results. Method-oriented approaches are often only effective for specific land cover class/data source combinations, and cannot be applied when different classification systems or data sources are required or available. Here we present a method, based on Endorsement Theory, of pooling evidence from multiple expert systems and spatial datasets to produce land cover maps. Individual ‘experts’ are trained to produce evidence for or against a class, with this evidence being categorised according to strength. An evidence integration rule set is applied to evidence lists to produce conclusions of different strength regarding individual classes, and the most likely class identified. The only expert system design implemented currently within the methodology is a neural network model, although the system has been designed to accept information from decision trees, fuzzy k-means and Bayesian statistics as well. We have used the technique to produce land cover maps of Scotland using three classification systems of varying complexity. Mapping accuracy varied between 52.6% for a map with 96 classes to 88.8% for a map with eight classes. The accuracy of the maps generated is higher than when individual datasets are used, showing that the evidence integration method applied is suitable for improving land cover mapping accuracy. We showed that imagery was not necessarily the most important data source for mapping where a large number of classes are used, and also showed that even data sources that produce low accuracy scores when used for mapping by themselves do improve the accuracy of maps produced using this integrative approach. Future work in developing the method is identified, including the inclusion of additional expert systems and improvement of the evidence integration, and evaluation carried out of the overall effectiveness of the approach.     

Assessing the potential of VEGETATION sensor data for mapping snow and ice cover: A Normalized Difference Snow and Ice Index

The VEGETATION (VGT) sensor in SPOT 4 has four spectral bands that are equivalent to Landsat Thematic Mapper (TM) bands (blue, red, near-infrared and mid-infrared spectral bands) and provides daily images of the global land surface at a 1-km spatial resolution. We propose a new index for identifying and mapping of snow/ice cover, namely the Normalized Difference Snow/Ice Index (NDSII), which uses reflectance values of red and mid-infrared spectral bands of Landsat TM and VGT. For Landsat TM data, NDSII is calculated as NDSII_TM=(TM3-TM5)/(TM3+TM5); for VGT data, NDSII is calculated as NDSII_VGT=(B2-MIR)/(B2+MIR). As a case study we used a Landsat TM image that covers the eastern part of the Qilian mountain range in the Qinghai-Xizang (Tibetan) plateau of China. NDSIITM gave similar estimates of the area and spatial distribution of snow/ice cover to the Normalized Difference Snow Index (NDSI=(TM2-TM5)/(TM2+TM5)) which has been proposed by Hall et al. The results indicated that the VGT sensor might have the potential for operational monitoring and mapping of snow/ice cover from regional to global scales, when using NDSII_VGT.     

Land cover mapping of large areas using chain classification of neighboring Landsat satellite images

Satellite imagery is the major data source for regional to global land cover maps. However, land cover mapping of large areas with medium-resolution imagery is costly and often constrained by the lack of good training and validation data. Our goal was to overcome these limitations, and to test chain classifications, i.e., the classification of Landsat images based on the information in the overlapping areas of neighboring scenes. The basic idea was to classify one Landsat scene first where good ground truth data is available, and then to classify the neighboring Landsat scene using the land cover classification of the first scene in the overlap area as training data. We tested chain classification for a forest/non-forest classification in the Carpathian Mountains on one horizontal chain of six Landsat scenes, and two vertical chains of two Landsat scenes each. We collected extensive training data from Quickbird imagery for classifying radiometrically uncorrected data with Support Vector Machines (SVMs). The SVMs classified 8 scenes with overall accuracies between 92.1% and 98.9% (average of 96.3%). Accuracy loss when automatically classifying neighboring scenes with chain classification was 1.9% on average. Even a chain of six images resulted only in an accuracy loss of 5.1% for the last image compared to a reference classification from independent training data for the last image. Chain classification thus performed well, but we note that chain classification can only be applied when land cover classes are well represented in the overlap area of neighboring Landsat scenes. As long as this constraint is met though, chain classification is a powerful approach for large area land cover classifications, especially in areas of varying training data availability.