Sugarcane production estimates using Landsat-2 MSS Data and agrometeorological models

Date

1983

Abstract

The capability of Landsat MSS Data and weather-based yield models in predicting the two parameters of sugarcane production, namely, cropped area and cane/sugar yield, were tested against ground truth data of Hda. Luisita, San Miguel, Tarlac. Cropped area estimated were determined from the Image-100 multrispectral analysis of Landsat-2 MSS data taken on three acquisition dates in 1976. Cane/sugar yield estimates were obtained from numerical models developed using the step-wise regression procedure wherein (1) 13-year record of yields were regressed against six weather variables, and (2) the 1976 cane/sugar yields against nine vegetation indices calculated from the brightness values extracted from the Landset-2 imageries. The cane-cropped area was best estimated, delieneated and mapped in the August 26, 1976 imagery (relative difference, R.D. = 1.2%) because of this date, the actively growing canes presented sharp contrast in brightness with the residentials, fallows and the surrounding ricelands. Both the weather-based and Landsat yield models gave significant coefficients of determination (R^2) but the former proved to be superior in terms of stability, reproducibility and the ability to give estimates at least two months in advance of the start of milling. Of the six climatic variables tested, namely, monthly maximum and minimum temperature, rainfall, percent relative humidity, rainy days and sunshine duration, the monthly sunshine duration showed the highest potential in predicting cane yield (R^2 = 99.20**) and sugar yield (R^2 = 96.24**). When the combined effect of the six weather variables were regressed against yield, comparable results were obtained: R^2 = 98.89** for cane yield and R^2 = 96.45** for sugar yield. Further computer analysis of Landsat data showed that the study area can be accurately classified into age groups of three-month invervals. Visual comparisons of computer print-out maps with ground truth were presented and quantitatively supported by age-predicting equations using vegetation indices from Landsat data. Integrating the area and yield estimates, the sugarcane production computed from the two sets of data both passed the set acceptable limit of 10 per cent relative difference. The true production value was more closely approximated by the area and yield determined entirely from Landsat data. However, the combined approach using Landsat data for area and against agromet model for yield estimates is recommended again in view of its greater stability and reproducibility. The use of Landsat data for cropped area estimates is commendable but for cane yield prediction, it needs further testing in different parts of the country to conclusively establish its applicability in the Philippines. | **Significant at P = 0.01 level

Document Type

Master Thesis

Degree

Master of Agriculture

Major Course

Major in Sugar Technology

Language

English

Location

UPLB Main Library Special Collections Section (USCS)

Call Number

LG 995 1983 A38 A73

This document is currently not available here.

Share

COinS