Technologies for Improved Nutrient Analysis

Tacy CalliesNutrition, Technology

nutrient
Figure 1. Drone equipped with a multispectral camera

By Yiannis Ampatzidis and Ute Albrecht

Nutrient management is important for citrus production. Regular nutrient assessments should be conducted to optimize nutrient balance and prevent deficiencies or over-fertilization. Optimizing nutrition is important for tree health and can improve tolerance to stresses and diseases.

Good nutrient management requires regular field monitoring to identify problems and examine crop responses. Leaves need to be collected and sent to a specialized laboratory to get a detailed analysis of macronutrients and micronutrients, which is time-consuming and costly.

It is recommended to conduct nutrient analyses in July and August after the spring flush (see edis.ifas.ufl.edu/pdf/SS/SS53100.pdf), but more frequent analyses of leaf nutrients may be necessary to determine deficiencies associated with huanglongbing or other biotic and abiotic factors. Additional analyses may also be necessary where responses to novel management practices need to be monitored. However, it is not economically feasible to frequently collect and analyze leaves for plant nutrient status.

In addition to being time-consuming and costly, leaf nutrient analysis is prone to human error because of inconsistencies and bias during leaf sampling and the analysis process, which can compromise the interpretation and relevance of the data. Faster and cheaper alternatives to conventional nutrient analysis methods are being developed at a rapid pace.

DRONES AND ARTIFICIAL INTELLIGENCE
New technologies like unmanned aerial vehicles (UAVs or drones) and artificial intelligence (AI) can be utilized to develop a more efficient methodology to determine leaf nutrient concentrations and improve the speed of data collection and consistency. Researchers at the University of Florida Institute of Food and Agricultural Sciences (UF/IFAS) Southwest Florida Research and Education Center (SWFREC) developed a non-destructive method that can be quickly and efficiently used to determine citrus leaf nutrients and create fertility maps that are compatible with variable-rate fertilizer applicators. This novel method can help overcome or complement some of the limitations of traditional leaf nutrient analysis methods.

Spectral reflectance (i.e., the energy a surface reflects at a specific wavelength) of citrus canopies in five bands of light (red, green, blue, red edge and near-infrared) were used to create an AI-based model to determine plant nutrient concentrations. The data were collected with a quadcopter UAV equipped with a multispectral camera (Figure 1).

A large dataset with good variability was developed by analyzing four large-acreage commercial field trials in two different citrus production areas (Central Ridge and Southeast Florida) with two different scions (Hamlin and Valencia orange) and a diversity of rootstocks. The differences in location, grove management and scion−rootstock combination affect nutrient uptake and distribution in the tree canopy, which made the dataset sufficiently robust to develop a precise predictive model.

The framework of this study was divided into two main phases. In the first phase (data acquisition), researchers acquired the spectral measurements of the canopy reflectance with a UAV-based multispectral camera. Leaf samples were collected, and nutrients were analyzed in the laboratory to generate the dataset. The second phase (model building and validation) consisted of: 1) a pre-analysis to evaluate the dataset for each nutrient and 2) model development and evaluation to ensure the repeatability of the methodology used.

FERTILITY MAPS
The developed AI model provides nutrient concentrations for individual trees. A novel cloud-based application, Agroview (also developed at the UF/IFAS SWFREC), was used to create fertility maps with discrete management zones to help visualize the data (plant nutrient concentrations).

nutrient
Figure 2. Potassium map based on University of Florida Institute of Food and Agricultural Sciences guidelines with deficient (red), low (brown), optimum (green) and high (blue) zones, developed by Agroview.

An example of a fertility map for potassium can be seen in Figure 2. Five zones (deficient, low, optimum, high and excess) were determined based on UF/IFAS guidelines. Figure 2 depicts the range values for each zone.

The tree ratio for each zone is the number of trees divided by the number of tree spaces (trees + gaps; a gap is a place with no tree). If the tree ratio is 100%, it means that the entire zone has zero gaps. As can be seen in Figure 2, the tree ratios for the deficient and low zones are lower (60.8% and 82.6%, respectively) than the optimum and high zones (89.9% and 87.8%, respectively) in this grove. Similar patterns can be observed for the nitrogen and phosphorus in this area. This indicates that this area of the grove might be problematic and require different management.

Figure 3. Agroview is a cloud-based application developed to convert data into practical information for growers.

PRACTICAL INFORMATION
Agroview (www.agroview.ai) converts data collected by UAV and from the ground into practical information. For example, Figure 3 shows a large citrus production area with specific information for a selected area (outlined in red) displayed in the window on the left. This information includes the total number of trees and gaps, the UAV flight date, the total size of the area, average tree size and density and average plant nutrient concentrations.

The main advantage of this AI-based methodology is that large populations of citrus trees can be assessed quickly and at a low cost while reducing inaccuracies that result from sampling a small subset of plants. The AI model developed with this technology had an error of less than 15% for most of the nutrients analyzed, but the accuracy is expected to improve as evaluations continue.

This new technology generates prescription maps for variable-rate application of fertilizers based on UAV imagery. Although this model was tested in commercial citrus production systems, it can be easily adapted to other crop and production systems.

Acknowledgment: This material was made possible, in part, by a cooperative agreement from the U.S. Department of Agriculture’s (USDA) Animal and Plant Health Inspection Service, the U.S. Department of Agriculture’s Agricultural Marketing Service through grant AM190100XXXXG036 and the Citrus Research and Development Foundation (CRDF). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the USDA and CRDF.

Yiannis Ampatzidis is an associate professor and Ute Albrecht is an assistant professor at the UF/IFAS SWFREC in Immokalee.

Share this Post