Crop-yield-prediction-using-weather-data-and-NDVI-time-series

The proposed system will be able to predict the crop yield production which will be useful to farmers for harvesting and storage. The system will use the weather forecasting which includes the parameters like temperature, rainfall, humidity, dew point and the normalized difference vegetation index time series from Sentinel-2 satellite for selected region. By using both the results obtained,accurate crop yield prediction can be calculated which will help farmers in planning efficiently, minimize costs and maximize yields—and profits—as a result.

What is NDVI ? The Normalized Difference Vegetation Index is a simple indicator that indicates the green areas in the specified region. The NDVI is analyzed using the remote sensing measurements which are captured from the space platform. The calculation of NDVI requires the multispectral image from the satellite. The multispectral images contain more than 3 bands which are red, green, blue. The NDVI can be calculated by the simple formula that requires the values of 2 bands red and near infrared. The NDVI value ranges from -1 to +1, where positive values indicate more denser green region while negative values indicate lesser green area.The NDVI for the Sentinel-2 satellite can be calculated after 3-5 days interval and hence the time series can be analysed for the specific region of interest. NDV I = (NIR − Red) (NIR+Red) where, N DV I : Normalize Difference Vegetation Index NIR : Near Infrared

Problem Statement

Previously, the research was based on weather parameters or remote sensing data for crop yield prediction, but combination of both was nominal. MODIS and LANDSAT VII/VIII satellite images were suitable for state/country but not for small scale/regional farm lands.This work propose a novel hybrid approach which combines weather parameters and NDVI values from Sentinel II having spatial resolution of 10m and revisit cycle of 5 days which can provide better accuracy for crop yield prediction in regional farming areas for sustainable development. This approach will predict the yield well ahead of harvesting time using historical crop production, weather and NDVI parameters by applying varied machine learning techniques. The proposed system will be able to predict the crop yield production which will be useful to farmers for harvesting and storage. The system will use the weather forecasting which includes the parameters like temperature, rainfall, humidity, dew point and the normalized difference vegetation index time series from Sentinel-2 satellite for selected region. By using both the results obtained,accurate crop yield prediction can be calculated which will help farmers in planning efficiently, minimize costs and maximize yields—and profits—as a result.

Software Requirements

• Python 3.7 : This is required for basic coding of the different machine learning algorithms and basically required for jupyter notebook. • Jupyter Notebook 6.0 : This is required for the actual preprocessing on raw data, processing on the obtained dataset with different algorithms. Google Earth Engine: This is required to access the satellite images and find the NDVI time series for the previous years of the region of interest. • QGIS 3.10.5 : This is required for GIS Mapping of Maharashtra ESRI shapefiles. • Apache Tomcat 9.0 : The tomcat is a web server used for the host the web gis portal with geoserver. • Geoserver 2.17 : The geoserver is used as GIS server for hosting the shapefiles that can be visualized using web server in portal. • PostgreSQL 9.1 : This is the sql databse used for the storage of the data and the required for the postgis database. • PostGIS 2.2 : This is the geospatial database required to store the co ordinates of the shapefiles with its different attributes for the visualization process.