Journal of Agricultural Machinery (Sep 2017)

Remote Sensing of Residue Management in Farms using Landsat 8 Sensor Imagery

  • M. A Rostami,
  • H Afzali

DOI
https://doi.org/10.22067/jam.v7i2.50118
Journal volume & issue
Vol. 7, no. 2
pp. 388 – 400

Abstract

Read online

Introduction Preserving of crop residues in the field surface after harvesting crops, making difficult farm operations. The farmers for getting rid of crop residues always choose the easiest way, i.e. burning. Burning is one of the common disposal methods for wheat and corn straw in some region of the world. Present study was aimed to investigate the accurate methods for monitoring of residue management after wheat harvesting. With this vision, the potential of Landsat 8 sensor was evaluated for monitoring of residue burning, using satellite spectral indices and Linear Spectral Unmixing Analysis. For this purpose, correlation of ground data with satellite spectral indices and LSUA data were tested by linear regression. Materials and Methods In this study we considered 12 farms where remained plants were burned, 12 green farm, 12 bare farms and 12 farms with full crop residue cover were considered. Spatial coordinates of experimental fields recorded with a GPS and fields map were drawn using ArcGissoftware, version of 10.1. In this study,t wo methods were used to separate burned fields from other farms including Satellite Spectral Indices and Linear Spectral unmixing analysis. In this study, multispectral landsat 8 image was acquired over 2015 year. Landsat 8 products are delivered to the customer as radiometric, sensor, and geometric corrections. Image pixels are unique to Landsat 8 data, and should not be directly compared to imagery from other sensors. Therefore, DN value must be converted to radiance value in order to change the radiance to the reflectance, which is useful when performing spectral analysis techniques, such as transformations, band ratios and the Normalized Difference Vegetation Index (NDVI), etc. In this study, a number of spectral indices and Linear Spectral Unmixing Analysis data were imported/extracted from Landsat 8 image. All satellite image data were analyzed by ENVI software package. The spectral indices used in this study were NDVI, BAI, NBR and NBRT. Classification accuracy was evaluated and expressed by confusion matrix and Kappa coefficient. Natural surfaces are rarely composed of a single uniform material. Spectral mixing occurs when materials with different spectral properties are represented by a single image pixel. The condition where scale of the mixing is large (macroscopic), mixing would occur in a linear fashion. However for microscopic situations, the mixing is generally nonlinear. The linear model ahich wasadopted in this study, assumes that there is no interaction between materials. Assumption of LSUA is that each pixel on the surface is a physical mixture of several constituents weighted by surface abundance, and the spectrum of the mixture is a linear combination of the endmember reflectance spectra. Within the context of this study, LSUA is a classification method that can determine contribution of each material (or endmember) such as soil or residue for each image pixel. Results and Discussion The spectral response curve extracted from Landsat 8 image used as input into the LSUA model in ENVI software. As expected, crop burned residue (Ash) spectra had lower reflectance when compared to the soil, residue and green plant spectra. The contrast between residue, green plant, soil and residue ash spectra was particularly evident in the NIR and SWIR bands. It is suggested that these bands are essential for residue discrimination. Differences of reflectance in the visible bands were minimal, providing little discrimination between residue, green plant, soil and residue ash. Burned area estimated by LSUA method from Landsat 8 image was correlated against the ground data (measured coincident to the ground data). The overall accuracy of classification with BAI index and LSUA method was 91.7 and 88.3 and Kappa coefficient was 0.89 and 0.84 respectively. Results indicated that burned field area can be located and its area can be estimated using Landsat 8 images. The Index BAI was selected as discernment index for burned/unburned fields in Landsat 8 images. Conclusions Present study was aimed to evaluate the accurate methods for monitoring residue management after wheat harvesting. With this vision, the potential of Landsat 8 sensor local data for monitoring residue burning was evaluated using satellite spectral indices and Linear Spectral Unmixing Analysis. Results indicate that residue ash spectra had lower reflectance when compared to the residue, soil and green plant except NIR band spectra. The contrast between residue, soil, green plant and residue ash spectra was particularly evident in the NIR bands. Results indicated that burned field area can be located and its area can be estimated using Landsat 8 images. The Index BAI was selected as discernment index for burned/unburned fields in Landsat 8 images.

Keywords