Bernardino J. Buenaobra1*, Mark V. Manhuyod2 and Roland E. Otadoy1,3
1 USC Phil-LiDAR Research Center, University of San Carlos, Cebu City, Philippines 6000
2 Plantation Systems and Analytics, Del Monte Philippines Inc., Bukidnon, Philippines 8705
3 Department of Physics, University of San Carlos, Cebu City, Philippine 6000




Launched in 1999 and 2013, the former being Landsat 7 and the latter being Landsat 8 [1] missions; these remote sensing satellite even at the time of this writing has persisted and continued to provide researchers, educators and remote sensing practitioners a set of user selectable downloadable and free 30 meter resolution imagery. With the facility of GLOVIS and Earth Explorer at USGS website where it is hosted, a careful selection from a range of images from acquisition time window, considering for example cloud cover as a filter, these images can be a quick and practical source of data of Earth’s land surfaces for the analyst. Furthermore, in Landsat 7 sensor also known as Enhanced Thematic Mapper (ETM+) there is eight discrete bands with Band 8 as having a 15m resolution called Panchromatic Band. With Landsat 8 sensors also known as Operational Land Imager (OLI) it has 11 discrete bands with 8th band being at 15m also a Panchromatic Band. For an analyst which favors localized or smaller area change detection at greater resolution, the process of subsetting followed by a pansharpening between B-G-R 30m bands in the usual sequence and a 15m panchromatic 8th band results to a bringing these visible bands to a halved better from their original resolution. In this paper we present a test case of the Landsat 8 data that was acquired on 2015-12-15 at WRS_PATH of 115 and a WRS_ROW of 55 with practically zero cloud cover. The geographic location is M.Fortich, Bukidnon, Philippines in central Mindanao. This area is known where the 2nd largest grower and supplier of pineapple fruit and cannery products in the world are located. In this paper we demonstrate the carrying out of the procedures of data processing data set imagery using open source GIS and RS remote sensing software. Specifically the QGIS and its QGIS Plug-in tools [2] and GRASS GIS [3], which all can provide an improved visualization of various vegetation indices. For this paper we select Normalized Difference Vegetation Index (NDVI), Green Chlorophyll Index (CIgreen), Ratio Vegetation Index (RGVI) also called Simple Ratio and Green Ratio Vegetation Index (GRVI).The emphasis is of the use Band 4 or the NIR band in Landsat 8 sensors. We then move the data from prior processing results to ImageJ [4] image processing environment where the operating variables made up of vegetation indices will be calibrated in terms of distance in meter/pixel and dimensionless vegetation indices value in pixels. At the end the analyst can have the facility of using line profiling and region of interest (ROI) calculations in image, to see the histogram distribution of the VI values.




Data Flow Diagram


To proceed, we refer to the data flow diagram in Figure 1.0 in the following page. This shows the key steps to arrive at the results. The first stages in this diagram are essentially pre-processing steps these comprise of the analysts action beginning with downloading the study area Landsat dataset. It will be the judgment of the analyst to provide a reasonable criterion for quality and quantity of images as needed. The percentage cloud cover is primary importance. For imagery that comes from cities or locations on earth with air particulates and aerosols atmospheric corrections are needed, a commercial software package ENVI may be used for this purpose. However, free and open source remote sensing tool called Monteverdi Orfeo Toolbox [5] also an algorithm in QGIS plugins developed by a French university can also be used. For color channel operations the tool Multispec [6] a multispectral data analysis software from Purdue University can also be an alternative.



Figure 1. Data flow for processing and visualization of vegetation indices from Landsat imagery

    Atmospheric correction may not be necessary if the field or study area itself is practically pristine from polluting aerosols and particulates like in cities. The two last most important procedures called subsetting followed by pansharpening (panchromatic band sharpening), are the key operations for image preprocessing the data set. We have used a USD $99.95 utility software called Pancroma [5] for this purpose. Monteverdi Orfeo Toolbox open or Multispec free RS software will do as well. The results will improve overall resolution from 30m to 15m. For an objective where vegetation vigor is what is sought – then the visible bands R, G, B i.e. absorption by the plant leaf and the invisible NIR band i.e. reflection from the plant leaf will be primarily considered.


Figure 2. Landsat 8 OLI selected dataset for case study


     The next succeeding functional blocks forms part of the visualization for the pre-processed data this operation is intended to arrive at converting the color channels into a single composite. By stacking of individual color planes carried out by use of band mathematics as shown in Figure 4 below e.g. floating point addition and division of band components in software to determine NDVI. Here we use two important open source and free QGIS and GRASS GIS. We use QGIS special miscellaneous function called Build Virtual Raster to calculate specially stacked combinations of color bands – this will also in effect provide the rendering to the resulting raster giving a false color to NIR making the vegetation as colored red with some gradient to convey health.




A select set of Vegetation Index for visualization


In order to arrive to a numerical indication of vegetation health and vigor various indices have been defined and developed by many remote sensing (RS) scientist and engineers as shown in Table 1. A notable description of which have been recently published by Andrés Viña et al [7] where the authors describe from experimental data the comparison of different vegetation indices to be used for the remote assessment of green leaf area index of crops here x is GLAI. For our purposes we select at least four indices that makes use of NIR Band 4 (Landsat 7), Band 5 (Landsat 8) and the visible color channels B,G,R that is Band 1, 2 and 3 (Landsat 7) and Band 2, 3 and 4 (Landsat 8).


Table 1. A select Vegetation Index (VI) definition




Calibration from raster values to pixel intensities


   Already available after applying the Band Math expression for the selected indices is a raster value per distance in GRASS GIS however, we are interested to arrive at a magnitude per pixel. This can be done by moving the profiling into an image processing environment here FIJI (ImageJ). In general given a resolution of an imaging sensor in number of horizontal pixel elements hpel and vertical pixel elements vpel and a scenery with known horizontal wd and vertical distance in meters vd we can arrive at a horizontal dh and vertical dv pixel/meter calibration by a simple unit analysis below:




With an image processing software facility of line profiling, histogram generation, region of interest only (ROI), point and click on a spatial calibrated image is possible at the analyst’s hand.


The procedure requires that a maximum and minimum value from the raster will be matched with the highest and lowest pixel available known in the image format e.g. 0-255 levels of intensity and creating the calibration curve or line with a known model of the index being considered (see Table 1). Figure 5.0 below shows an example output of this procedure in GRASS GIS and FIJI (ImageJ).



Figure 5. Calibration from raster value to pixel intensities for Simple Ratio Index




Results from Landsat 8 OLI dataset preprocessing


We performed the subsetting and panchromatic sharpening in Band 2, 3, 4 and Band 5 and used Band 8 to interpolate them to a higher resolution the results are shown on the Table 2 below:


Table 2 Results from pre-processing procedure for case study area



Results for visualization of NDVI, RVI, GRVI and Green Chlorophyll Index


We show below in Table 3 that with GRASS GIS and ImageJ the variation on VIs can be visualized with a line profile and ROI histogram distribution.


Table 3 Visualizations of selected Vegetation Indices multispectral, raster and pixel data types




Keywords: Landsat Missions, Visualizations, NDVI, Panchromatic, Pansharpeni




[1] Landsat link: (accessed on 3 April 2016)

[2] QGIS link: (accessed on 3 April 2016)

[3] GRASS GIS link: (accessed on 3 April 2016)

[4] ImageJ link: (accessed on 3 April 2016)

[5] Orfeo Monteverdi Toolbox link: (accessed on 3 April 2016)

[6] Multispec link: (accessed on 3 April 2016)

[7] Andrés Viña et al, 2011, Comparison of different vegetation indices for the remote assessment of green leaf area index of crops Remote Sensing of Environment 115 (2011) 3468–3478