Format

Send to

Choose Destination
Int J Biometeorol. 2015 Dec;59(12):1747-59. doi: 10.1007/s00484-015-0983-0. Epub 2015 Apr 1.

Rubber yield prediction by meteorological conditions using mixed models and multi-model inference techniques.

Author information

1
Institute of Plant Production and Agroecology in the Tropics and Subtropics, University of Hohenheim, Garbenstrasse 13, 70599, Stuttgart, Germany. golbon@uni-hohenheim.de.
2
Institute of Crop Science, University of Hohenheim, Fruwirthstrasse 23, 70599, Stuttgart, Germany.
3
Institute of Plant Production and Agroecology in the Tropics and Subtropics, University of Hohenheim, Garbenstrasse 13, 70599, Stuttgart, Germany.

Abstract

Linear mixed models were developed and used to predict rubber (Hevea brasiliensis) yield based on meteorological conditions to which rubber trees had been exposed for periods ranging from 1 day to 2 months prior to tapping events. Predictors included a range of moving averages of meteorological covariates spanning different windows of time before the date of the tapping events. Serial autocorrelation in the latex yield measurements was accounted for using random effects and a spatial generalization of the autoregressive error covariance structure suited to data sampled at irregular time intervals. Information theoretics, specifically the Akaike information criterion (AIC), AIC corrected for small sample size (AICc), and Akaike weights, was used to select models with the greatest strength of support in the data from a set of competing candidate models. The predictive performance of the selected best model was evaluated using both leave-one-out cross-validation (LOOCV) and an independent test set. Moving averages of precipitation, minimum and maximum temperature, and maximum relative humidity with a 30-day lead period were identified as the best yield predictors. Prediction accuracy expressed in terms of the percentage of predictions within a measurement error of 5 g for cross-validation and also for the test dataset was above 99 %.

KEYWORDS:

Hevea brasiliensis; Meteorological conditions; Mixed models; Multi-model inference; Prediction; Yield

PMID:
25824122
DOI:
10.1007/s00484-015-0983-0
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Springer
Loading ...
Support Center