Estimation of tea yield using multiple linear regression modelling, A case Study in Kakonde Tea Estate.

Abstract

In Uganda, tea is the third most agricultural foreign exchange earner following coffee and fish although it is performing below its expected potential compared to other tea growing countries like Kenya. Various government and private and tea estate managers require advance information about the expected yield of tea plantation estates so as to develop early warning systems about fields which are experiencing difficulties and easily allocate scarce resources effectively so as to maximize yield in peak seasons. This clearly demonstrates the need to change from conventional methods that lie on data collection to better approach of planning mitigation measures before harvesting to maximize yield at peak seasons. Motivated by the operational use of remote sensing in various agricultural crop studies, this study evaluates the application and utility of remote sensing and GIS-based techniques in tea yield estimation. The potential of widely used vegetation indices like NDVI, elevation along with weather parameters has been evaluated for the prediction of green leaf tea yield before actual harvesting using a multiple linear regression model.

A regression analysis between tea yield and its growing parameters was done so as to check whether the model fits the data well and also find out which independent variables (NDVI, rainfall, temperature and elevation) contribute to dependent variable(tea yield). Rainfall and Normalized difference vegetation index were the most relevant variables in the prediction of tea yield, they had a strong positive correlation of 0.6741 and 0.6479 respectively with tea yield data. Temperature and elevation variables were not so significant, they had relatively moderate correlation coefficients of 0.2767 and 0.2599 respectively with tea yield. The estimated tea yield of Kakonde tea estate on 30th may 2022 was 3235.7864kg/ha, however the actual tea yield was 3237.18856kg/ha which brings about a difference of 1.4021kg/ha and a standard error of 0.1894. A multiple variate linear model was also found valid with a coefficient of determination (R²) of 0.8298. The results demonstrated that this model can be adopted to estimate tea yield of a particular tea estate.