Agriculture Smart Data Lab

Embracing smart irrigation management techniques empowers growers to irrigate with greater efficiency, promoting sustainable agricultural production. Crop evapotranspiration is a key quantity in enabling informed irrigation decisions. The emergence of Unmanned Aerial Vehicles (UAVs), has brought about a revolutionary shift in agricultural monitoring. This project introduces a robust method for classifying water stress in cotton using a compact UAV platform and convolutional neural networks (CNN).

This research was conducted at the USDA-ARS Cropping Systems Research Laboratory (CSRL) in Lubbock, Texas, where the cotton field was divided into 12 drip zones. The study included three replications to evaluate four irrigation treatments: “rainfed”, “full irrigation”, “percent deficit of full irrigation”, and “time delay of full irrigation”. The CNN model was able to classify the cotton water stress using UAV-based RGB images with a 91% overall best prediction accuracy.

Limited water supplies, increasing pumping costs, variable rainfall patterns, and competition with municipalities are some of the challenges faced by agricultural communities across the state. Consequently, timely delivery and efficient use of irrigation water are critical to the sustainability and long-term stability of agricultural production in Texas. Presently, irrigation scheduling decisions based on reference evapotranspiration (ET) are limited due to the lack of reliable and readily available in-field weather data and updated crop coefficients. Additionally, information about in-field variability of crop water demand is seldom considered in the decision-making process. As a result, the efficiency of irrigation is often low, which leads to wasteful use of water in areas that are already plagued by scarce water resources.

In this project, we propose the development of an Unmanned Aerial System (UAS) based crop monitoring system that calculates crop water use for irrigation scheduling and increased water use efficiency. To do this, our system will take advantage of big data analytics and Artificial Intelligence (AI)on the UAS-derived phenotypic data and in-field weather data to calculate actual crop evapotranspiration and biomass accumulation, and determine the timing and quantity of irrigation water needed.

In this project, we modified the CNN as a regression model for cotton yield prediction. Adapting the CNN architecture for cotton yield prediction while maintaining the image as the input introduces an intriguing approach to leveraging spatial patterns within the visual data to forecast agricultural outcomes. By transforming the output layer into a single-neuron regression model, the network can learn to interpret intricate visual cues within the images to predict the quantitative yield values accurately. This unique fusion of image-based inputs and regression modeling expands the capabilities of CNNs, enabling them to decipher nuanced features within the visual data directly related to cotton yield prediction. 

There is a growing scientific and public consensus about the impact of changing climatic patterns on the intensification of extreme rainfall events. Existing stormwater infrastructure systems fail to adequately respond to the changing extreme storm events since they are designed with older manuals based on historical rainfall records. Thus, there needs to be a re-examination of the frequency and intensity of rainfall projections. A comprehensive, reproducible, and consistent methodology to process the projections and develop updated Intensity-Duration-Frequency (IDF) curves is needed for infrastructure design and management.

This project is the first to propose such a comprehensive framework to apply multiple approaches to the precipitation projection for Texas. The study will apply an array of traditional statistical methods and machine learning (ML) models to update the IDF curves for Texas for the next 80 years. The method will be scalable and transferable to other locations, thus enabling the scientific community and engineers to better adapt to future rainfall patterns.

According to the 2020 Texas State Water Plan, irrigation will remain the largest water consumer in Texas for the next 30 years. However, many agricultural regions in the state are facing severe water shortages for irrigation, creating a need for effective water conservation strategies. The challenge lies in motivating farmers to adopt irrigation scheduling practices and transition to more efficient methods, which can be expensive. What if we could incentivize farmers through financial compensation for implementing water conservation practices, using emerging carbon markets? Would these incentives encourage them to make the necessary changes that would benefit water conservation efforts in Texas? This project explores the relationship between optimal irrigation practices, carbon intensity, and financial benefits. Our goal is to establish Texas A&M as a leading institution in evaluating agricultural water management practices for carbon markets.