Integration of Sentinel-derived NDVI to reduce uncertainties in the operational field monitoring of maize

Canopy cover (CC) is a key parameter in calibration and validation of crop growth models, especially those used in operational field monitoring. However, CC direct measurements require intense field campaigns, increasing the cost in time-series data acquisition for large agricultural areas. Normalized Difference Vegetation Index (NDVI) is a commonly used remote-sensing vegetation index, expressing crop water-status, being indirectly related to CC. In this paper, we explore the relationship between on-site CC and the high-resolution NDVI data acquired via Sentinel 2 products. This relationship was utilized to produce CC time series over the cultivation period in four maize fields in northern and central Greece. Subsequently, the expression linking CC and NDVI was used to operationally validate CC change in a crop model capable to simulate the maize growth cycle (AquaCrop). The proposed method involves the dynamic in-season re-adjustment to a number of key model input parameters, based on the remotely acquired CC time series, namely maximum CC, canopy growth and decline coefficient, growing degree days needed to the beginning of senescence stage. These re-adjusted parameters were imported to model's crop file to improve simulations in CC, soil water content, final biomass and yield. Results showed that the remotely acquired CC time series could be successfully used as an alternative mean to validate CC simulations. Moreover, the ingestion of re-estimated parameters to crop file, improved model's capability to simulate CC (R2 0.98; RMSE < 5.12%), biomass (Pe < 12%) and yield (Pe < 12%). No significant differences were observed in model's performance regarding soil water content simulation.

Automated summary

The study explored the relationship between high-resolution NDVI data and on-site CC, using it for operational validation and improvement of crop growth models. Results showed that remotely acquired CC time series can be successfully used as an alternative means for validating CC simulations, while re-estimated parameters enhanced model performance in simulating CC, biomass, and yield. No significant differences were observed in soil water content simulation.