An evaluation of smartphone applications for LAI estimation to facilitate canopy state assessment in vineyards

Crop development and foliar density as expressed with Leaf Area Index (LAI) is an important source of information for disease prevention. Canopy density in vineyards has been correlated with disease incidence, mainly concerning the impact of high density on intra-canopy ventilation and levels of humidity. LAI data can be used together with other data sources, such as temperature, humidity, rainfall etc., to enhance disease predictive models and continuous monitoring of crops. To improve the crowdsourcing aspect of data collection from farmers and agronomists capturing in-field observations, this study was implemented aiming to evaluate LAI smartphone applications. The applications selected for testing and evaluation were smart fLAIr (https://sys.cs.uos.de/smartflair) and VitiCanopy (https://viticanopy.com.au), selected based on their applicability, subscription pricing, user-friendliness and continued support from the developers among all available Android applications. The smartphone applications were evaluated against LiCOR 2200C plant canopy analyzer (https://www.licor.com/env/products/leaf_area/LAI-2200C) to demonstrate the measurement accuracy of each. Sampling for this experiment was carried out in four plots (25 points/plot, 100 total) applying gaiasense smart farming services (https://www.gaiasense.gr/en/gaiasense-smart-farming), located in two irrigated commercial vineyards in Stimagka, southern Greece. The collected samples were representing various canopy states considering foliar density. Sampling took place during early morning hours (after sunrise) for the first two plots, while the remaining two plots were sampled after midday to early afternoon hours (before sunset). All sampling locations were recorded with geo-tagged photographs. A cap-view of 45^o under clear-sky conditions was used for LiCOR2200C measurements and atmospheric scattering correction was applied, following a 4A measurement sequence protocol as described in the instruction manual (https://licor.app.boxenterprise.net/s/fqjn5mlu8c1a7zir5qel). FV2200 software (https://www.licor.com/env/support/LAI-2200C/software.html) was used to process the LiCOR dataset. Statistical analyses were performed after excluding 10% of total acquired samples as outliers. The results show that VitiCanopy has greater accuracy compared to fLAIr with a correlation coefficient of 0.65 over 0.25, while producing overestimated LAI values (mean diff = 0.74, p<0.0001). On contrast, fLAIr generated slightly underestimated LAI values (mean diff=-0.24, F=0.0155). Per plot analysis showed that measurements acquired earlier during the day (first two plots) provided higher correlation values (0.39<r<0.64), while those acquired after midday scored lower (r<0.12). This comes in agreement with relevant literature, suggesting that the ideal light conditions for accurate LAI measurements (under clear-sky conditions) is the earliest possible after sunrise. Although correlation values remained low to moderate (0.07<r<0.64), findings indicate that VitiCanopy performs more accurately than fLAIr and can be used as an alternative to costly and sophisticated equipment, however caution should be taken while standardising the optimal atmospheric/lighting conditions. This insight can be useful for disease predictive models, as well as farmers and agronomists who seek an accessible way to monitor LAI, potentially leading to spatially variable spraying applications. Future plans include the integration of LAI measurements as an additional parameter within the gaiasense’s Smart Farming solution aiming to enhance information richness of the existing operational pest infestation risk index calculation algorithms for vineyards.

This work was supported by EU-H2020 project ‘Resilient farming by adaptive microclimate management’ (STARGATE – 818187).