Integration of optical and synthetic aperture radar (SAR) images to differentiate grassland and alfalfa in Prairie area

doi:10.1016/j.jag.2013.10.003

International Journal of Applied Earth Observation and Geoinformation

Volume 28, May 2014, Pages 12-19

https://doi.org/10.1016/j.jag.2013.10.003 Get rights and content

Highlights

•
Proposed low cost earth observation sensor imagery: MODIS and ScanSAR, to identify spatial distribution of alfalfa to estimate its biofuel potential at a regional level.
•
Investigated an innovative wavelet and IHS integration image fusion technique in combing MODIS and ScanSAR.
•
Demonstrated the image fusion technique in improving grassland and alfalfa differentiation significantly compared with commonly used multi-source data combination methods.

Abstract

Alfalfa presents a huge potential biofuel source in the Prairie Provinces of Canada. However, it remains a challenge to find an ideal single satellite sensor to monitor the regional spatial distribution of alfalfa on an annual basis. The primary interest of this study is to identify alfalfa spatial distribution through effectively differentiating alfalfa from grasslands, given their spectral similarity and same growth calendars. MODIS and RADARSAT-2 ScanSAR narrow mode were selected for regional-level grassland and alfalfa differentiation in the Prairie Provinces, due to the high frequency revisit of MODIS, the weather independence of ScanSAR as well as the large area coverage and the complementary characteristics SAR and optical images. Combining MODIS and ScanSAR in differentiating alfalfa and grassland is very challenging, since there is a large spatial resolution difference between MODIS (250 m) and ScanSAR narrow (50 m). This study investigated an innovative image fusion technique for combining MODIS and ScanSAR and obtaining a synthetic image which has the high spatial details derived from ScanSAR and the colour information from MODIS. The field trip was arranged to collect ground truth to label and validate the classification results. The fusion classification result shows significant accuracy improvement when compared with either ScanSAR or MODIS alone or with other commonly-used data combination methods, such as multiple files composites. This study has shown that the image fusion technique used in this study can combine the structural information from high resolution ScanSAR and colour information from MODIS to significantly improve the classification accuracy between alfalfa and grassland.

Introduction

Biofuels, being advocated as a renewable, cost-effective alternative to petroleum-based liquid fuels, require development of a cellulosic-based biofuels industry (Campbell, 2012). Perennials rich in fibre are generally a suitable feedstock for bioenergy, while those abundant in foliage are efficient as a feed for livestock. Alfalfa (Medicago sativa L.) is proposed as a biofuel feedstock, since its stems could be processed to produce energy or fuel and the leaves used as a livestock feed (McCaslin and Miller, 2007). In order to understand and develop alfalfa biofuel potential the spatial distribution needs to be determined more accurately than currently available spatial information extraction methods.

SAR is used often in vegetation mapping, due to its independence from solar illumination, and its different imaging principles when compared with the optical image (Buckley, 2004, Smith and Buckley, 2011). The brightness of a SAR image depends on the roughness, geometry, and material contents of the targeted surface and the wavelength of SAR. The grey information from optical image represents the reflectance of solar energy from a target area (Jensen, 2005). Combing microwave and optical sensors can help in discriminating the different classes since they are complementary to each other (Pohl and Van Genderen, 1998). Many studies have combined optical image and microwave image to improve mapping accuracy in agricultural scenarios (Brisco et al., 1989, Schistad-Solberg et al., 1994, Brisco and Brown, 1995, Le Hegarat-Mascle et al., 2000, Ban, 2003, Blaes et al., 2005, Michael et al., 2005, McNairn et al., 2009). SAR and optical imagery can be integrated in different ways to improve the data and information content during image processing for information extraction. Image fusion is a technique which can combine the optical and SAR sensor data, prior to information extraction. The purpose of radar and optical image fusion is mainly for feature enhancement and confusion reduction (van der Sanden and Thomas, 2004, Schistad-Solberg et al., 1994). The roles of image fusion can be reflected in three ways:

(1)
Take maximum advantage of the merits of each single sensor. As every sensor has disadvantages and advantages there is a potential synergy in integrating data from different sensors to take advantage of those sensors’ advantages without significantly distorting the desirable characteristics of every sensor (Lewis et al., 1998, Amarsaikhan and Douglas, 2004).
(2)
Reduce information redundancy caused by multisource data (Schistad-Solberg et al., 1994). Images acquired over the same geographic area by different sensors have two possibilities: partially redundancy, since they cover the same geographic area; and partially complementarity since sensors cover different spectral ranges, which is very apparent when comparing optical and microwave images. The aim of image fusion is not only to use their complementarities to reduce confusion by getting a more complete description of land cover type features, but also to use multisource data redundancy to reduce imprecision and classification errors (Le Hegarat-Mascle et al., 2000), thus improving classification results (Tso and Mather, 2001).
(3)
Un-mix mixed pixels. Generally, the spatial resolution of the selected SAR image is higher than the optical multispectral image. The fused image can un-mix the mixed pixels in the lower spatial resolution multispectral image to a certain level (Van der Meer, 1997, Robinson et al., 2000, Bachmann and Habermeyer, 2003, Hong et al., 2011).

This study collected early-season remote sensing images to differentiate perennials from annuals. An early-season MODIS and ScanSAR narrow mode image were selected in this study for regional level grassland and alfalfa differentiation in the Prairie area, due to grassland and alfalfa growing at the same time. MODIS can be used in operational mapping since it includes 7 spectral channels, primarily for use in land mapping applications, large coverage, high revisit frequency, small data volume, and no cost since February 2000. These characteristics are good for regional-level mapping applications. However, the spatial resolution is relatively coarse (250 m for the first two channels and 500 m for other 5 channels). The coarse resolution causes mixed pixel problems and lower classification accuracy. ScanSAR data from the Radarsat-2 observation, is a good data source for obtaining high-resolution spatial information (50 m) for regional mapping at a frequent repeat rate due to its all-weather and day–night collection capability, its low cost for Canadian Government-related projects, and the large geographic coverage (300 km × 300 km). However, a single SAR image produces low separability between different land use activities including grassland versus alfalfa. The objective of this study is to investigate an image fusion technique to improve grassland and alfalfa differentiation by combining MODIS and ScanSAR imagery. Specifically, we aimed to answer the following questions:

(1)
Does the incorporation of radar information in the classification process between alfalfa and grassland improve accuracy?
(2)
What kind of radar/optical data combination(s) is/are more suitable to provide this information?

Section snippets

Study area

A pilot study area was selected in Southern Saskatchewan (Fig. 1). The geographic coverage of this area is about 211 km × 236 km. The study area is primarily semiarid and land use is dominantly cereal production, with pasture, forage, oilseeds, pulse production and some conservation parks.

Data sets

Two early-season MODIS and ScanSAR data sets were acquired; the MODIS on June 2, 2009 and the ScanSAR narrow mode on June 20, 2009. The early season data selection was mainly to avoid other spectral confusion

ScanSAR ortho-rectification process

ScanSAR images were ortho-rectified with the DEM data (1:250,000 scale) downloaded from GeoBase (http://www.geobase.ca/). The radar-specific model in PCI OrthoEngine was used in the ortho-rectification process. As the terrain relief in the study area is not too high, the final image residual errors in X and Y direction are both less than 1 pixel. The ortho-rectified image was resampled to 50 m by using the nearest neighbour resampling method.

Speckle removal

The original ScanSAR image is contaminated with

Fusion

The original HV (Fig. 3(a)) was selected to fuse with MODIS (bands 6, 2, and 1, Fig. 3(b)) since HV is more sensitive to vertical structures compared with HH mode, with the final fusion result listed in Fig. 3(c). Fig. 3(c) appears similar to Fig. 3(b) from the point of colour, as no serious colour distortion was identified in the fusion result. To clarify the details of those images, the close view of a subset area highlighted in Fig. 3(a) are listed in Fig. 4(a)–(c). Fig. 4(c) shows that the

Conclusion

This study proposed an earth observation based method to identify alfalfa spatial distribution due to its huge biofuel potential in the Prairie Provinces. The challenging parts exist in two aspects: alfalfa has a similar growing season and spectral confusion to other crops and cloud-free remote sensing data is not easy to acquire during crop growing season. The early season remote sensing imagery was acquired to avoid spectral confusion with other annual crops. This study proposed to combine

Acknowledgements

The authors would like to thank Canadian Space Agency for providing the Radarsat-2 data through the Climate Change Geoscience Program of the Earth Sciences Sector, Natural Resources Canada. Two anonymous reviewers’ critical comments are greatly appreciated to improve this manuscript significantly. Financial support from Canadian Space Agency, through the Government Related Initiatives Program, and from York University contract faculty research grants fund (CUPE 3903) is acknowledged.

References (32)

X. Blaes et al.
Efficiency of crop identification based on optical and SAR image time series
Remote Sensing of Environment
(2005)
R.G. Congalton
A review of assessing the accuracy of classifications of remotely sensed data
Remote Sensing of Environment
(1991)
H. McNairn et al.
Integration of optical and Synthetic Aperture Radar (SAR) imagery for delivering operational annual crop inventories
ISPRS Journal of Photogrammetry and Remote Sensing
(2009)
G.D. Robinson et al.
Evaluation of two applications of spectral mixing models to image fusion
Remote Sensing Environment
(2000)
D. Amarsaikhan et al.
Data fusion and multisource image classification
International Journal of Remote Sensing
(2004)
Y. Ban
Synergy of multitemporal ERS-1 SAR and landsat TM data for classification of agricultural crops
Canadian Journal of Remote Sensing
(2003)
M. Bachmann et al.
Evaluation of image fusion techniques for large-scale mapping of non-green vegetation
B. Brisco et al.
Multi-date SAR/TM synergism for crop classification in Western Canada
Photogrammetric Engineering and Remote Sensing
(1995)
B. Brisco et al.
Early season crop discrimination with combined SAR and TM data
Canadian Journal of Remote Sensing
(1989)
R.J. Buckley
Enhanced classification of prairie landscapes using simulated RADARSAT-2 imagery
Canadian Journal of Remote Sensing
(2004)

J. Campbell

Mapping grasslands for biofuel potential

USGS Newsroom

(2012)

R.G. Congalton et al.

Assessing the Accuracy of Remotely Sensed Data: Principles and Practices

(2008)

G.M. Foody

Classification accuracy assessment

IEEE Geoscience and Remote Sensing Society Newsletter

(2011)

G. Hong et al.

Crop type identification potential of Radarsat-2 and MODIS images in prairie area

Canadian Journal of Remote Sensing

(2011)

G. Hong et al.

Fusion of MODIS and Radarsat data for crop type classification – an initial study

G. Hong et al.

A wavelet and IHS integration method to fuse high resolution SAR with moderate resolution multispectral images

Photogrammetric Engineering and Remote Sensing

(2009)

Cited by (55)

On the automatic quality assessment of annotated sample data for object extraction from remote sensing imagery
2023, ISPRS Journal of Photogrammetry and Remote Sensing
Nowadays, the tasks of object extraction from remotely sensed imagery, such as built-up area, water-body, and other Earth surface information extraction, heavily rely on annotated sample data due to the wildly used machine learning algorithms, especially Deep Learning (DL) methods. Most algorithms, however, treat image-based annotations as ground truth and ignore the impact of unreliable annotations, leading to low-reliability solutions. Additionally, manual Annotation Quality aSsessment (AQS) of remote sensing data usually entails a significant workload in actual dataset production, particularly for large datasets. To alleviate these issues, we propose a novel Annotation Quality aSsessment Network (AQSNet) for the automatic quality assessment of remote sensing annotated samples. It uses a multi-scale channel-spatial attention module to simulate how an inspector examines annotated samples at local and global scales, enhancing the ability to identify poor-quality annotations. Besides, we present a simple method for generating effective training samples by analyzing the types and characteristics of low-quality annotated sample data. Furthermore, we introduce and make publicly available HBD4¹, a massive land cover dataset, with 1,945,034 samples of size 512 × 512 pixels. To validate our methodology, two AQS datasets are constructed with HBD4 and public building datasets. On the building AQS dataset, the proposed method outperforms the baseline model (HRNet) by 7.115% in IoU and 5.712% on the water-body AQS dataset, respectively. Extensive ablation and comparison experiments also demonstrate that the proposed approach can effectively assess the quality of remote sensing annotated samples. The high-quality sample data acquired by our AQSNet serves as a pre-trained dataset in the extended applications, from which robust pre-trained weights for downstream tasks are obtained, achieving accurate results with only 5% training samples in the LandCover.ai dataset for water-body extraction. To summarize, our method enables the automatic quality assessment of annotated sample data and the construction of high-quality sample data, both of which have the potential to improve the performance of machine learning models in the remote sensing community. The source code and the two constructed AQS datasets will be publicly available at https://github.com/zhilyzhang/AQSNet.
Evaluating NISAR's cropland mapping algorithm over the conterminous United States using Sentinel-1 data
2021, Remote Sensing of Environment
Accurate knowledge of the distribution, breadth and change in agricultural activity is important to food security and the related trade and policy mechanisms. Routine observations afforded by spaceborne Synthetic Aperture Radar (SAR) allows for high-fidelity monitoring of agricultural parameters at the field scale. Here we evaluate the approach to be used for generating NASA's upcoming NASA ISRO SAR (NISAR) mission's L-band cropland product using Sentinel-1C-band data. This study uses all ascending Sentinel-1A/B data collected over the conterminous United States in 2017 to compute the coefficient of variation (CV) at 150 m × 150 m resolution and evaluates the overall accuracy (OA) of CV-based crop/non-crop classifications at 100 one-by-one degree tiles. We calculate accuracies using two approaches: (a) using a literature-recommended constant CV threshold of 0.5 (CV_{thr_0.5}) and (b) determining optimal CV thresholds for every tile using Youden's J statistic (YJS), CV_{thr_YJS}. These accuracy comparisons are important for determining (1) the viability of using a computationally inexpensive and straightforward approach for cropland classification over large spatial scales/diverse land covers (i.e., can accuracies ≥80% be routinely achieved?), (2) how closely OA_0.5 compares to the performance ceiling (OA_YJS). This information will help determine whether approach (a) is appropriate and how much potential room of improvement there could be in modifying it. Results for OA_0.5 and OA_YJS are 81.5% and 86.8%, respectively. A breakdown by census geographic region, showed that OA_0.5 (OA_YJS) exceeded 80% (90%) in the South and Midwest, but was only 76.1% (73.5%) in the West. The improvement in OA_YJS mainly stems from tiles with >40% crop prevalence having about 10% greater OA values. To better examine the potential of the approach for land cover classification, results of approach (b) were also stratified by crop. Approach (b) accurately detected most non-crop classes as non-crop (>80%), but with low OA_YJS values for grasslands/pasture, especially in the West. CV values for crop were distinct from non-crop indicating that the approach is suitable for crop/non-crop classifications. Because results CV values have substantial overlap within crop/non-crop classes, indicating the approach is poorly suited for land cover classifications. We also detected a strong geographic dependence of CV_{thr_YJS}: values ranged from about 0.2 at the coasts and gradually increase to about 0.6 in the Central United States, most often falling close to 0.3 and 0.5.
Contribution of multispectral (optical and radar) satellite images to the classification of agricultural surfaces
2020, International Journal of Applied Earth Observation and Geoinformation
Citation Excerpt :
Numerous studies have shown the ability of optical imagery to detect the type and state of a crop (Joshi et al., 2016) and the ability of radar images to follow surface states and stages of development (Hadria et al., 2009; McNairn et al., 2014). Given the complementary nature of optical and radar signals (notably their different penetration capacities), they have been used in synergy to improve the ways agricultural surfaces are monitored, including the accuracy of mapping and of biophysical parameter estimations (Amarsaikhan and Douglas, 2004; Blaes et al., 2005; McNairn et al., 2009a; Fisette et al., 2013; Hong et al., 2014; Inglada et al., 2016). A variety of methods for detecting changes in land use by classifying multi-source, multi-temporal data have been proposed and evaluated in recent years (Lu and Weng, 2007; Mountrakis et al., 2011; Srivastava et al., 2012; Hussain et al., 2013; Tewkesbury et al., 2015).
The monitoring of different crops (cultivated plots) and types of surface (bare soils, etc.) is a crucial economic and environmental issue for the management of resources and human activity. In this context, the objective of this study is to evaluate the contribution of multispectral satellite imagery (optical and radar) to land use and land cover classification.
Object-oriented supervised classifications, based on a Random Forest algorithm, and majority zoning post-processing are used. This study emerges from the experiment on multi-sensor crop monitoring (MCM'10, Baup et al., 2012) conducted in 2010 on a mixed farming area in the southwest of France, near Toulouse. This experiment enabled the regular and quasi-synchronous collection of multi-sensor satellite data and in situ observations, which are used in this study. 211 plots with contrasting characteristics (different slopes, soil types, aspects, farming practices, shapes and surface areas) were monitored to represent the variability of the study area. They can be grouped into four classes of land cover: 39 grassland areas, 100 plots of wheat, 13 plots of barley, 20 plots of rapeseed, and 2 classes of bare soil: 23 plots of small roughness and 16 plots of medium roughness. Satellite radar images in the X-, C- and L-bands (HH polarization) were acquired between 14 and 18 April 2010. Optical images delivered by Formosat-2 and corresponding field data were acquired on 14 April 2010.
The results show that combining images acquired in the L-band (Alos) and the optical range (Formosat-2) improves the classification performance (overall accuracy = 0.85, kappa = 0.81) compared to the use of radar or optical data alone. The results obtained for the various types of land cover show performance levels and confusions related to the phenological stage of the species studied, with the geometry of the cover, the roughness states of the surfaces, etc. Performance is also related to the wavelength and penetration depth of the signal providing the images. Thus, the results show that the quality of the classification often increases with increasing wavelength of the images used.
Beyond Traditional Methods: Innovative Integration of LISS IV and Sentinel 2A Imagery for Unparalleled Insight into Himalayan Ibex Habitat Suitability
2023, Research Square
Remote sensing of grasslands in the South American Pampas (scientometrics analysis)
2023, Land Degradation and Development
Comparing Machine Learning Algorithms for Pixel/Object-Based Classifications of Semi-Arid Grassland in Northern China Using Multisource Medium Resolution Imageries
2023, Remote Sensing

View all citing articles on Scopus

View full text

Integration of optical and synthetic aperture radar (SAR) images to differentiate grassland and alfalfa in Prairie area

Highlights

Abstract

Introduction

Section snippets

Study area

Data sets

ScanSAR ortho-rectification process

Speckle removal

Fusion

Conclusion

Acknowledgements

Remote Sensing of Environment

Remote Sensing of Environment

ISPRS Journal of Photogrammetry and Remote Sensing

Remote Sensing Environment

Data fusion and multisource image classification

International Journal of Remote Sensing

Synergy of multitemporal ERS-1 SAR and landsat TM data for classification of agricultural crops

Canadian Journal of Remote Sensing

Evaluation of image fusion techniques for large-scale mapping of non-green vegetation

Multi-date SAR/TM synergism for crop classification in Western Canada

Photogrammetric Engineering and Remote Sensing

Early season crop discrimination with combined SAR and TM data

Canadian Journal of Remote Sensing

Enhanced classification of prairie landscapes using simulated RADARSAT-2 imagery

Canadian Journal of Remote Sensing

Mapping grasslands for biofuel potential

USGS Newsroom

Assessing the Accuracy of Remotely Sensed Data: Principles and Practices

Classification accuracy assessment

IEEE Geoscience and Remote Sensing Society Newsletter

Crop type identification potential of Radarsat-2 and MODIS images in prairie area

Canadian Journal of Remote Sensing

Fusion of MODIS and Radarsat data for crop type classification – an initial study

A wavelet and IHS integration method to fuse high resolution SAR with moderate resolution multispectral images

Photogrammetric Engineering and Remote Sensing