THE USE OF SATELLITE IMAGES IN THE FIELD OF AGRICULTURE

Researches Review DGTH | 47-1, 11–22 | 2018 ABSTRACT: The use of remote sensing in the field of agriculture is very diverse. It includes detection of the amount of chlorophyll in crops, the amount of water, the detection of pests, the creation of maps for selective spraying and feeding, etc. In this work, different vegetation indices were applied on satellite imageries obtained from platforms: Sentinel2, Landsat8, RapidEye, and PlanetScope, in order to detect trends in the vegetation of different cultures. RapidEye and PlanetScope imageries were used in order to detect nitrogen variation for agricultural parcels in the area of the municipality of Becej. In this case indices NDVI, NDRE, and MSAVI2 are applied for cultures: corn, sunflower, wheat, sugar beet, and soybean. Sentinel2 and Landsat8 imageries were used for the analysis of the orchard near Celarevo in the municipality of BackaPalanka and analysis of agricultural parcel with a predominantly agricultural crop near Stapar in the municipality of Sombor. The detection of an area with health and non-health vegetation was made, as well as a comparison of the obtained vegetation trends between two platforms. The applied indices were: NDVI, TNDVI, GNDVI, EVI2, WDRVI, and GLI.

ABSTRACT: The use of remote sensing in the field of agriculture is very diverse.It includes detection of the amount of chlorophyll in crops, the amount of water, the detection of pests, the creation of maps for selective spraying and feeding, etc.In this work, different vegetation indices were applied on satellite imageries obtained from platforms: Sentinel2, Landsat8, RapidEye, and PlanetScope, in order to detect trends in the vegetation of different cultures.RapidEye and PlanetScope imageries were used in order to detect nitrogen variation for agricultural parcels in the area of the municipality of Becej.In this case indices NDVI, NDRE, and MSAVI2 are applied for cultures: corn, sunflower, wheat, sugar beet, and soybean.Sentinel2 and Landsat8 imageries were used for the analysis of the orchard near Celarevo in the municipality of BackaPalanka and analysis of agricultural parcel with a predominantly agricultural crop near Stapar in the municipality of Sombor.The detection of an area with health and non-health vegetation was made, as well as a comparison of the obtained vegetation trends between two platforms.The applied indices were: NDVI, TNDVI, GNDVI, EVI2, WDRVI, and GLI.

INTRODUCTION
Production of quality agricultural goods is a primary focus for each country.Nowadays, the development of science and technology effects agriculture to reach its peak through the application of geospatial technologies which involve the collecting of a large amount of quality data as well as its analysis and interpretation.In the most cases, collecting data is done using remote sensing, which contributes monitoring of large areas for the purpose of planning, forecasting and management strategies for agricultural resources.In order to emphasize the individual properties of plants, the mathematical combination of different values of spectral ranges is made.These combinations represent the vegetation indices (VI further) which are designed to point up the level of plant stress, nitrogen, chlorophyll, water, biomass estimation, yields, etc.
In this work NDVI, NDRE and MSAVI2 vegetation indices were applied on the satellite imageries from RapidEye and PlenetScope platforms, in order to detect nitrogen variability on the agricultural-cultivated crops: corn, sunflowers, wheat, sugar beet and soybean in the municipality of Becej.Moreover, NDVI, TNDVI, GNDVI, WDRVI, GLI and EVI2 vegetation indices were applied on the satellite images from Sentinel2 and Landsat8 platforms for the purpose of detecting trends of good and bad vegetation conditions, with a special review of the comparative analysis of the results obtained on both platforms.

Landsat8
Landsat8 is the eight satellites of Landsat series, developed and launched by NASA in 2013.It has the ability to record spectral energy in 11 ranges with a spatial resolution of 15m to 100m and temporal resolution of 16 days.

Sentinel2
Sentinel2 constellation has been developed by the ESA, and it consists of 2 satellites capable of capturing energy in 13 spectral ranges.Temporal resolution is 5 days.Comparative overview of the spectral bands of Sentinel2 and Landsat8 missions are represented on Figure 1.

RapidEye
RapidEye has a constellation of 5 identical satellites with the spatial resolution of 5m.It records energy in 5 spectral bands.The high temporal resolution of 24h is one of the main characteristics of this platform.

PlanetScope
PlanetScope constellation has approximately 150 satellites with the spatial resolution of 3m.It records energy in 4 spectral bands.PROCESSING DATA

Atmospheric correction
An important source of radiometric errors arises from the atmospheric weakness caused by the dissipation and absorption of EM energy by the atmosphere.When it is necessary to extract biophysical parameters of water masses or vegetation (chlorophyll, biomass, etc.) from satellite imageries, the use of atmospheric correction is mandatory.
The atmospheric correction of Landsat8 imageries was performed in QGIS software, where is also performed sharpening of the other bands based on the pan band -pansharpening algorithm.[4]Atmospheric correction was performed in Sen2Core software for Sentinel2, while Erdas Imagine software was used for RapidEye and PlanetScope.

Applicability of vegetation idicesand classification
The used vegetation indices were applied within the Erdas Spatial Model Editor tool, which allows the graphical creation of the flow of events and its execution.After the creation of the VI, next step was coloring the obtained raster known as engineering classification.Simple engineering classification was made for easier visual observation of the state of vegetation -trends.It was done in Erdas Knowledge Engineer tool.It was performed on 20 equal ranges.The used color range was from red to green (Figure 2).In order to detect locations of the same anomalies, an engineering classification with several decision parameters was done within the mentioned tool.There were multiple rasters (in this case VI) in the input and their classification was done according to certain rules (values) corresponding to the observed anomaly.The result is a raster with identifies areas that have a given range of values on all inputs.

RESULTS OF THE PROCESSING OF PLANETSCOPE AND RAPIDEYE FOOTAGE -MUNICIPALITY OF BECEJ
The analysis of the variance of the nitrogen was performed in the period from April to July by RapidEyeimageries. Figure 3 shows theclassification results after applying the NDVI index for parcels under corn.The marked part of the parcel represents the potential place of the missing of nitrogen concentration or excess of it.
For the corresponding parcels of sunflower,the results of the MSAVI2 index application are shown, as well as the identified sites of potential changes are characterized by the low reflection value.(Figure4) The application of the NDRE index, which is highly correlated with nitrogen variation, is shown in Figure 5 for the wheat culture and Figure 6 for the soybean culture.Throughout all four time moments of the analysis of the images, the low reflection rate prevails with a very small percentage of healthy vegetation (green parts of the parcel).
The use of the MSAVI2 indexis shown in Figure 7 on sugar beet plots.The anomaly with potential excess or deficiency of nitrogen is marked.
RapidEye and PlanetScopeimageries were used for date 9.6.2017.inorder to make the comparative analysis, with the accent on a better spatial resolution of the PlanetScope imagery -3m (Figure 8).On each pair of imageries can be noted the same trends of health and non-health vegetation; though there are some parts with deviations.A trend with less health vegetation was observed on all used indices (the area is marked in Figure 12) ranging from 90 to 195 (0-255).The engineering classification was executed for the mentioned range of values and all indices were used for the Sentinel-2 imageries from the date 4.7.2017.The result is shown in Figure 11, where can be detected a few more areas corresponding to the defined range.Considering that this area is covered by low crop cultivation, WDRVI and GLI produce poorer results because they are sensitive to the presence of dense and high vegeta-

Figure 2 .
Figure 2. The used color range

Figure 8 .
Figure 8. Results of NDVI index for RapidEye and PlanetScope

Figure 11 .
Figure 11.results of simple engineering classifications of TNDVI and GNDVI index for date 24.6.2017.

Figure 14 .
Figure 14.Results of simple engineering classifications of GNDVI, WDRVI and GLI index for dates 14.7.2017.(Sentinel2), 07.10.2017 and 17.07.2017.(Landsat8) EVI2 (The enhanced vegetation index 2) is developed for optimization of vegetation signal in order to increase sensibility in high biomass regions and to improve monitoring of vegetation.Also, it reduces the atmospheric effect on vegetation signal.Unlike NDVI, which is sensitive to chlorophyll, EVI2 is more sensitive to structural variations in the canopy.So it is often used for trees canopy classifica-WDRVI (The Wide Dynamic Range Vegetation Index) is designed to overcome some limitations of NDVI by introducing a weight factor on NIR band in the formula of NDVI.If medium to high biomass is analyzed (LAI>2), it is recommended to use WDRVI, while for low vegetation (LAI<1), NDVI is a better solution.[1] 1. NDVI (Normalized Difference Vegetation Index) is used for global vegetation analysis, based on linear regression of NDVI values with direct examination of LAI, APAR and biomass factor.It shows problems in scaling with saturated signals that often occur under high biomass condition of the forest regions,which is more favorable in the areas of low biomass.[7] 6. GLI (Green Leaf Index) uses only visible part of the spectrum.It is sensitive to the presenceof chlorophyll and nitrogen in plants.It gives better results for moderate to dense biomass.[5]