Crop modelling and remote sensing for yield predictions of asparagus cultivated in Peru

Abstract

In Peru, asparagus is an important crop for the export market. It is grown on a large scale in the North and South of the Pacific coastal desert with irrigation systems. Forecasting the yields is important in planning ahead sales to exporters. Therefore, farmers currently apply an empirical method of yield prediction by counting the number of mature buds per crown per meter and making linear comparisons with the harvests of previous years. Prior to this investigation, there was no simulation model for a continuous cycling crop grown in Peru, where temperatures are much higher than in seasonal production zones of Europe. This research describes the development of an appropriate simulation model for asparagus crop performance and yield predictions in Peru. It integrates crop modelling techniques and remote sensed data from Sentinel-1 to develop "TURION", a mechanistic crop model coded in Python which includes 26 physiological parameters and crop variables based on Wilson and Cloughley (1999); Wilson et al. (2002, 2005); Wilson and & Drost (2008) and field data. Growth rate, thermal time per phenological stages, biomass partition, stem diameter variations, spear volume and leaf areaindex (LAI) across the crop cycle were determined for model calibration. An advantage of the TURION model is that it can be easily adapted by farmers in Peru and Latin-America. The nine inputs required for TURION are readily available for asparagus farmers: brix% in roots (a widely used indicator of carbohydrate CHOcontent) at the time canopy are initially cut before harvest, solar radiation, daily mean temperature, three key dates detected with Sentinel-1 and three crop features. TURION includes three sub-models: (1) spears production and its root CHO depletion, (2) stems establishment and its root CHO depletion and (3) replenish CHO storage in roots by photosynthesis, LAI and CHO translocation. This model predicted: yield, numbers of spears, biomass of spears /stems, and root CHO in three brix% values: (1) at the end of the harvest, (2) at stem establishment and (3) at the end of the campaign. Therefore, TURION model was validated on crops ranging from 3 to 12 years post-establishment, for 75 commercial harvests reported between July 2018 and May 2020 over 38 different plots. Predictions provide outputs at plant level. Thus, results showed a relative root mean square error (RRMSE) of 2.45% for final yield, 1.95% for CHO at the end of harvest and 1.44% CHO at the end of the crop cycle.

Date of Award	2022
Original language	English
Awarding Institution	Aberystwyth University
Supervisor	John Doonan (Supervisor) & Alan Gay (Supervisor)