Modeling yield and backscatter using satellite derived biophysical variables of rice crop based on Artificial Neural Networks

Mahesh Palakuru MAHESH PALAKURU,Sirisha Adamala SIRISHA ADAMALA,Harish Babu Bachina HARISH BABU BACHINA

doi:10.54386/jam.v22i1.120

Abstract

In this study, ‘observed rice yield (ton acre-1)’ and ‘remotely sensed backscatter’are modelled using artificial neural network (ANN) and multiple linear regression (MLR) methods for East and West Godavari districts of Andhra Pradesh in India. The biophysical variables viz. backscatter (bs), normalized difference vegetation index (NDVI), Chlorophyll (chfl), fraction of absorbed photosynthetically active radiation (FAPAR), leaf area index (LAI), canopy water content (CWC), and fraction of vegetation cover (Fcover) were derived from Scatterometer Satellite-1 (SCATSAT-1), Moderate Imaging Spectrometer (MODIS) and Sentinel-2 satellite data.Inputs selected are bs, NDVI, chfl, FAPAR, LAI, CWC, and Fcover for rice yield model, whereas NDVI, chfl, FAPAR, LAI, CWC, and Fcover are inputs for backscatter models. The performance of ANN and MLR models was evaluated using three indices such as root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2). The results concluded that the ANN models achieved R2 of 0.908 and 0.884 which are 42.73% and 28.85% higher than that of the MLR method for rice yield and backscatter, respectively.

Highlights

Rice (Oryzasativa) is one of the world’s major staple and higher demand foods, especially in India.The rice production as a whole is sensitive to weather fluctuations (Mallick et al, 2007; Dari et al, 2017), under threat due to decrease in its acreage and increase in occurrence of extreme events like floods, droughts, hailstormsetc (Bal and Minhas, 2017)
The crop acreage for Godavari command area was estimated from the satellite imageries and this was based on the unsupervised classification using Erdas imagine software.The downstream area of study area is dominated with more rice crop area as compared to upstream area (Fig. 2)
The estimates of root mean squared error (RMSE), mean absolute error (MAE) and R2for artificial neural network (ANN) model were good with values 0.031, 0.025 and 0.969 as compared to multiple linear regression (MLR) model in yield prediction during training

Summary

Introduction

Rice (Oryzasativa) is one of the world’s major staple and higher demand foods, especially in India.The rice production as a whole is sensitive to weather fluctuations (Mallick et al, 2007; Dari et al, 2017), under threat due to decrease in its acreage and increase in occurrence of extreme events like floods, droughts, hailstormsetc (Bal and Minhas, 2017). Lopez-Sanchez et al (2011) investigated the potential of polarimetric synthetic aperture radar (SAR) imagery with X-band for identifying the different phenological stages of rice fields. This approach exploits the known sensitivity of polarimetry to the structure or morphology of the observed scene to differentiate the physical changes and conditions followed by rice crops during its growth cycle. Zhang et al (2018) mapped the rice phenology using ANN approach in China with Landsat 8 satellite imagery They have used land surface temperature (LST) and NDVI as auxiliary input data. The above all studies confirms the potentiality of ANN models to estimate rice crop yield using satellite derived biophysical parameters as input

Methods

Results

Conclusion