Operational Implementation Plan

The test site covers an area of 25×25 km2 and is characterized by major annual crops viz., rice (Oriza sativa), cotton and sugarcane. The flat topography and the agricultural dominance of the study area make it useful for crop-related studies using SAR data sets. Crop inventory mapping is one of the essential tasks for our study, which is required to establish the estimates of crop production. As such, the accurate crop classification at pixel or parcel level is of significant interest to the agriculture community.

We utilize the SAR derived polarimetric parameters as feature set for classification. In recent years, polarimetric data have been widely used for crop mapping and monitoring, and several experiments have underscored the potential of land cover and land use classification using SAR data. Our objective of this research is to assess the classification accuracy result for day wise and time series data over the test sites, using supervised classifiers (SVM and random forest).

In addition to backscatter intensities (HH, HV and VV channels), the polarimetric features (e.g., Kennaugh matrix elements, scattering power decomposition parameters etc.) are used as feature dataset. In first phase, we majorly focus on C-band RADARSAT-2 quad-pol data sets for crop classification. Later, with the availability of multi-frequency (TerraSAR-X and ALOS-2 L-band) data sets, we incorporate a multi-frequency classification approach. Also, we compare several classification approach for multi-temporal classification. A schematic diagram of a multi-temporal classification is presented in Figure 1.

Figure 1. The process fowchart of crop classification based on KPCA+SVM and SVM method. [1]

References:

[1] D. Mandal, V. Kumar, Y. S. Rao (2020), “An assessment of temporal RADARSAT-2 SAR data for crop classification using KPCA based support vector machine”, Geocarto International, 2020, DOI: 10.1080/10106049.2020.1783577

Scale

Regional

Operational Implementation Plan

Rice growth monitoring using Synthetic Aperture Radar (SAR) is recognized as a promising approach for tracking the development of this important crop. Accurate spatio-temporal information of rice inventories is required for water resource management, production risk occurrence, and yield forecasting. This research investigates the potential of the radar derived vegetation indices for monitoring rice growth at different phenological stages. Polarimetric SAR can provide a rich source of data to track temporal dynamics of vegetation conditions. Similar to spectral indices that are well established in optical remote sensing, a vegetation index derived from SAR data could be an alternative, especially for crop growth monitoring during periods of cloud cover.

We focus our experiment to implement recently developed three radar vegetation indices (Figure 2):

• Generalized volume scattering model based Radar Vegetation Index (GRVI) [1]
• Compact-pol Radar Vegetation Index (CpRVI) [2]
• Dual-pol Radar Vegetation Index (DpRVI) [3]

Figure 2. Radar vegetation indices

References:

[1] D. Mandal, V. Kumar, D. Ratha, J. M. Lopez-Sanchez, A. Bhattacharya, H. McNairn, Y. S. Rao, K.V. Ramana,”Assessment of rice growth conditions in a semi-arid region of India using the Generalized Radar Vegetation Index derived from RADARSAT-2 polarimetric SAR data,” Remote Sensing of Environment, Vol. 237, 111561, 2020, DOI:10.1016/j.rse.2019.111561.

[2] D. Mandal, D. Ratha, A. Bhattacharya, V. Kumar, H. McNairn, Y. S. Rao, A. C. Frery ,” A Radar Vegetation Index for Crop Monitoring Using Compact Polarimetric SAR Data,” IEEE Transactions on Geoscience and Remote Sensing, 2020, DOI: 10.1109/TGRS.2020.2976661.

[3] D. Mandal, V. Kumar, D. Ratha, S. Dey, A. Bhattacharya, J. M. Lopez-Sanchez, H. McNairn, Y. S. Rao, “Dual Polarimetric Radar Vegetation Index for Crop Growth Monitoring Using Sentinel-1 SAR Data,” Remote Sensing of Environment, 2020. (In press)

Crop Conditions

• growth status

Operational Implementation Plan

Variations in crop phenological stages can be characterized by Synthetic Aperture Radar (SAR) data due to its high sensitivity to the dielectric and geometrical structure of the canopy. We utilize novel PolSAR scattering components and clustering algorithms to classify phenological stages of rice. The clustering of rice phenological stages via Cloude and Pottier’s H/alpha is first proposed by Lopez-Sanchez et al. [1]. However, we introduce a new H\theta plane (unified framework for quad and compact-pol SAR data [2]) to characterize rice phenological stages. Later these new parameters would be used for phenological stages classification.

In addition, for dual-pol SAR data, we utilize a temporal data set and clustering algorithm to identify transplantation stage of rice over the region using Sen4Rice processing chain [3].

References:

[1] Lopez-Sanchez, J.M., Vicente-Guijalba, F., Ballester-Berman, J.D. and Cloude, S.R., 2012. Polarimetric response of rice fields at C-band: Analysis and phenology retrieval. IEEE Transactions on Geoscience and Remote Sensing, 52(5), pp.2977-2993.

[2] S. Dey, D. mandal, D. Ratha, J.M. Lopez-Sanchez, A. Bhattacharya, H.McNairn, Y. S. Rao, “Novel Clustering Schemes for Full and Compact Polarimetric SAR Data: A Case Study for Rice Monitoring”, ISPRS Journal of Photogrammetry and Remote Sensing (Under review).

[3] Mandal, D., Kumar, V., Bhattacharya, A., Rao, Y.S., Siqueira, P. and Bera, S., 2018. Sen4Rice: A processing chain for differentiating early and late transplanted rice using time-series Sentinel-1 SAR data with Google Earth engine. IEEE Geoscience and Remote Sensing Letters, 15(12), pp.1947-1951.

Phenological Events

• Seedling
• Vegetative Growth
• Fruit Development
• Harvest

Operational Implementation Plan

Retrieval of crop biophysical parameter is one of the major goal for the current experiment. High spatio-temporal resolution and wide spatial coverage provided by EO satellite data are key inputs for operational crop monitoring. In Synthetic Aperture Radar (SAR) applications, a semi-empirical model (viz., Water Cloud Model (WCM) [1]) is often used to estimate vegetation descriptors individually from quad and dual-pol data. Plant Area Index (PAI) and wet biomass (W) are used as the vegetation descriptors in the WCM. The performance of the inversion approach is evaluated with in-situ measurements collected over the test site using RADARSAT-2 and Sentinel-1 data. We also explore compact-pol (simulated Radarsat Constellation Mission –RCM data) for biophysical parameter estimation. With the availability of temporal data (multi-year) we can perform the cross-validation of model estimation and its transferability.

References:

[1] Attema, E.P.W. and Ulaby, F.T., 1978. Vegetation modeled as a water cloud. Radio science, 13(2), pp.357-364.

Biophysical Variables

• Vegetation Indices (NDVI, EVI, SAVI, etc.)
• LAI (Leaf Area Index)
• Biomass
• LAIeff