Evaluation of water stress mapping methods in vineyards using airborne thermal imaging

Sammanfattning: With low cost and high efficiency, use of airborne infrared thermography imaging from an unmanned aerial vehicle (UAV) has become a widely used technique to measure plant water stress. With that, the use of UAV mounted thermal camera as a precision agriculture technique to map crop water stress is not trivial and its success is depended on variety of factors which are still needed to be studied and understood. The current research study sought to explore three crucial components of mapping water status based on airborne thermal imaging: a) the quality, quantity and distribution of the sampling points, b) the type of thermal index used, and c) the choice of interpolation method. The following three objectives were highlighted: a) Comparison of two commonly used thermal indices, the crop water stress index (CWSI) and the Jones stomatal conductance index (Ig), to assess index performance in regions with high climatic variability. b) Assessment of the impact of spatial and temporal changes on the performance of interpolation algorithms. c) Investigation of the possibility of mapping vineyard water stress through collection and analysis of the vineyard cover crop thermal data in comparison to the grapevine canopy. The study was conducted in two commercial vineyards in the Rheingau region of Germany. Airborne thermal imaging using UAV was collected at four different periods during the 2019 growing season (July-September) and was accompanied by ground measurements including mid-day stem water potential (Ψstem) and proximal thermal imaging. The airborne data was interpolated using the following interpolation algorithms: Inverse Distance Weighting (IDW), Kriging, Local Polynomial, and Spline. The resulting interpolated surfaces were evaluated through cross validation and, through comparison to the ground measurements. Cross-validation results showed definite preference for CWSI based interpolations. The result can be attributed to the range of potential values rather than the suitability of the index. In contrast, comparison to the ground measurement results show definite preference for Ig based interpolations. Both the cross validation and the comparison to the ground measurement results showed a high range of interpolation algorithms varying spatially between the two vineyards and temporarily between the different measurement dates. Cross-validation analyses and comparison with the proximal thermal imaging measurements revealed a preference for cover crop-based interpolation. While comparison with the Ψstem showed a preference for the cover crop during the July measurements and preference for the grapevine during the August-September measurements. It is inferred that the results are related to the small sample size of the Ψstem procedure and operator bias. Results of the study indicated that the Ig index exhibits much higher suitability than CWSI for mapping water stress index in regions with higher humidity and variable climate such as the Rheingau. Additionally, the study showed the affect of sampled data points on the resulting interpolated surface, the importance of evaluating different algorithms and choosing the most suitable one. Finally, the results demonstrated that cover crop-based data have the potential to producing better quality water stress maps in steep-slopped vineyards which are characterized by low soil depth. Further research is thus recommended to evaluate suitability of this method in vineyards with deeper soil profile to confirm that this phenomenon is not restricted to steeped sloped vineyards.