Search Filter
- Publications
- Abstract of Theses and Dissertations
- Database
- Rice yield prediction using leaf area index and normalized difference...
Rice yield prediction using leaf area index and normalized difference vegetation index under diverse cultural management practices in Myanmar
Dissertation Abstract:
Robust and reliable tools for crop monitoring and early forecasting of rice production are critical in heavy rice consuming regions including Myanmar. This study was conducted with the overall objective of the development of rice yield prediction models using remotely-sensed leaf area index (LAI) and normalized difference vegetation index (NDVI) under different cultural practices in Kyaukse Township of Myanmar during rice monsoon season, from July to December 2018. In this study, remotely-sensed 16-day composite NDVI and 8-day composite LAI values at different growth stages of rice, rice grain yields, field data and cultural practices of 60 selected farmers' fields were collected. Based on the results, farmers' cultural practices affected vegetation indices (LAI and NDVI) and rice grain yields. With favorable water conditions and high applied nitrogen (N) fertilizer rate, transplanted rice with 31-45 days old seedling showed higher values of LAI, NDVI and rice grain yields. The optimum planting time for monsoon rice in the study area was during 4th week of July to 1st week of August. Correlation analysis showed that LAI and NDVI were positively correlated with rice yield. From regression analyses, 8 yield prediction models from NDVI images, 13 models from LAI images and 9 models from combination of NDVI and LAI images were generated. All of the models were capable of predicting rice yield at different growth stages. Based on the statistical results of model validation, only 4 models (N1, L11, NL1, and NL2) had excellent accuracy for yield estimation. Among these 4 models, the model N1 generated using only one NDVI image of 56 DAT as an input was the best model. This model required only one NDV1 image while other models needed more inputs and it can predict rice yield before panicle initiation in which farmers have enough time to apply additional fertilizers and modify some of their cultural practices especially water management following the advisories if the predicted yield is low. The regression models using remotely-sensed LAI and NDVI data are potentially useful for yield forecasting and for crop monitoring in Myanmar.