Modeling vineyard leaf area index using remote sensing data at different spatial resolutions

Abstract

The estimation of Leaf area index (LAI) at large spatial scales (vineyard level) is a crucial factor for several Precision Viticulture (PV) applications, associated with the plant growth modeling and decision support for irrigation and canopy management. The present study is focused on the assessment of LAI estimations using remote sensing data (Normalized Difference Vegetation Index – NDVI) at different spatial resolutions. NDVI values were calculated at three levels of spatial resolution:  i) Landsat 8 (30 m spatial resolution); ii) Sentinel-2A (10 m spatial resolution) and iii) unmanned aerial vehicle – UAV (0.06 m spatial resolution). The NDVI-LAI relationship for measures registered with the UAV was developed using in-situ LAI allometric measurements taken in 24 ground calibration sites inside of the experimental vineyard plot (1.8 ha) during the grapevine growing season 2016-2017. The vineyard plot cv. Malbec was trained on VSP with a distance between rows of 2.5 m and within rows of 1.5 m, and drip irrigated. In this study, the NDVI-LAI relationship calculated using images captured by the UAV presented a high agreement with an r² equal to 0.86 and was used as a comparison method for the satellite approaches. Results of low spatial resolution NDVI-LAI relationships, obtained using Landsat 8 and Sentinel-2A, had coefficients of determination (r²) equal to 0.23 and 0.45, respectively. These results give a straightforward evidence of the spatial resolution impact in the estimation of LAI in vineyards when the intra-block spatial variability is considered in the analysis.