Weather-Driven Agricultural Decision-Making Using Digital Twins Under Imperfect Conditions CCS: Applied computing Agriculture

Tamim Ahmed, Monowar Hasan · 2025

Read Weather-Driven Agricultural Decision-Making Using Digital Twins Under Imperfect Conditions CCS: Applied computing Agriculture by Tamim Ahmed, Monowar Hasan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-08326.txt.

Introduction

Tamim Ahmed and Monowar Hasan build a farm decision tool in a 2025 paper around digital twins. A digital twin is a live virtual copy of a real system, fed by its sensors. In farming, weather data drives choices about water and timing, but field sensors often report flawed values.

Their framework, Cerealia, checks incoming weather data for inconsistencies when no perfect feed exists. Neural network models flag the odd readings, so growers can judge which data to trust. The team prototyped the system on a small edge computer and tested it in a working orchard and on public datasets. Cleaner inputs should mean sounder automated farm decisions.

Catalog record

CREATORTamim Ahmed, Monowar Hasan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-08326

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Weather-Driven Agricultural Decision-Making Using Digital Twins Under Imperfect Conditions CCS: Applied computing Agriculture Tamim Ahmed, Monowar Hasan Abstract . By offering a dynamic, real-time virtual representation of physical systems, digital twin technology can enhance data-driven decision-making in digital agriculture. Our research shows how digital twins are useful for detecting inconsistencies in agricultural weather data measurements, which are key attributes for various agricultural decision-making and automation tasks. We develop a modular framework named Cerealia that allows end-users to check for data inconsistencies when perfect weather feeds are unavailable. Cerealia uses neural network models to check anomalies and aids end-users in informed decision-making. We develop a prototype of Cerealia using the NVIDIA Jetson Orin platform and test it with an operational weather network established in a commercial orchard as well as publicly available weather datasets. 1. Introduction Modern agriculture, especially high-value specialty crop production, relies heavily on data-driven automation technologies for production management. Agricultural decision-making increasingly relies on precise and consistent environmental data to optimize crop health, resource utilization, and yield predictions. Regional as well as localized weather critically drives several on-farm decision-support models and intelligent automated systems, including irrigation, abiotic stress (e.g., cold, heat) mitigation, and pest management, among others ( boomgard2022machine ) . Weather data measurements, which include essential metrics such as temperature, humidity, and precipitation, form the foundation of such decisions. However, weather networks, like any cyber-physical and Internet-of-Things (IoT) systems, are often plagued by inconsistencies 1 1 1 In this work, we focus on effects on inconsistent/irregular weather streams rather than the causes that trigger the anomalies. due to sensor faults, calibration drift, communication errors, environmental interference, or cyber breaches. Imperfect measurements can significantly undermine the reliability of downstream applications in agriculture, ranging from predictive modeling to real-time interventions. Such discrepancies can lead to inaccurate predictions, suboptimal resource allocation, and reduced agricultural productivity. Any inconsistent weather data input could lead to inadequate decision-making and wrongful actuation of precision management scenarios, resulting in considerable crop yield and quality losses. To address this problem, we adopt the digital twin technology ( barricelli2019survey ) for analyzing and understanding agricultural weather networks that are critical for grower decision support. We build dynamic virtual data streams that mirror physical stations—to monitor, validate, and enhance the reliability of weather data in the face of anomalies. While digital twin technology is used in other cyber-physical domains (e.g., control systems, manufacturing, robotics), its benefits are not fully explored for digital agriculture. By leveraging the concept of digital twins, we show that it is feasible to create a virtual replica of weather networks for real-time inconsistency detection, data imputation, and predictive analytics. We name our framework Cerealia . 2 2 2 In ancient Roman religion, the Cerealia was a major festival dedicated to the sowing, growing, and harvesting of crops. As our work aims to assist in better decision-making for agricultural stakeholders, we name our framework Cerealia . Figure 1 . High-level schematic of this work. Cerealia stores historical weather traces and simulated anomalous data that is used to detect runtime inconsistent data generated by the remote weather stations deployed in the field. The end users can use Cerealia to analyze how inconsistent data impacts the target agriculture decision-making process. Figure 1 presents a high-level overview of Cerealia . The “twin” of weather stations stores a historical database of past weather measurements and uses a synthetic data generator engine that is capable of emulating “noisy” (viz., inconsistent) data. Based on these observations, Cerealia uses machine learning models to check for inconsistencies in the real-time weather feed coming from the in-field weather stations. Our current integration of Cerealia integrates nine neural network-based machine learning models (see § 3.3 ). These models are optimized for detecting and addressing inconsistencies in weather data. A key feature of Cerealia is to design it as a modular tool that end-users can utilize for several in-farm decision-making tasks when perfect physical measurements are not available. As a user-inspired scenario, we study inconsistency implications, i.e., how erroneous weather data impacts in-farm decision-making. Specifically, we conduct two comprehensive case studies that demonstrate its applicability and impact. Our first case study predicts weather data . This is useful for weather data imputation applications, where growers are required to predict missing or inconsistent weather data, perhaps due to system malfunction or lack of stations in the target regions. The second case study involves predicting fruit surface temperature , which is a crucial indicator of fruit stress. Growers use temperature prediction tools to initiate automated cooling precautions to reduce crop losses. We show that Cerealia is useful to observe whether inconsistent data affects the surface temperature prediction. Our Contributions. By bridging computing and agriculture, Cerealia represents a significant step toward building resilient, data-driven systems capable of withstanding the uncertainties in weather measurements. The contributions of this paper are as follows: • We introduce Cerealia , a digital twin-based framework that addresses data inconsistencies using neural network models in agricultural weather networks [§ 3 ]. • By leveraging Cerealia , we demonstrate the usefulness of digital twins for agricultural decision-making using two realistic case studies (predicting missing weather data and fruit surface temperatures) [§ 6 ]. Cerealia is implemented and tested by using a commercial orchard located in Central Washington, United States (§ 4 –§ 5 ). 3 3 3 This research uses weather stations installed in a commercial orchard maintained by Washington State University, United States. While this is a closed system, access can be made available for research purposes by request, as we did in this work. We use this platform to demonstrate the efficacy of our approach in a real, live setup. Our use of this platform, however, by no means implies our affiliation with this infrastructure/institution. The digital twin modules are deployed on an NVIDIA Jetson Orin platform ( karumbunathan2022nvidia ) . In addition to live weather feed from the orchard, Cerealia is also tested using a publicly available European weather dataset ( jena_weather_dataset ) . We conduct a comprehensive evaluation of nine state-of-the-art neural network models to analyze their effectiveness in detecting weather data inconsistencies. We measure the overheads of running learning models on our evaluation platform and test the scalability of Cerealia . We find that the overheads of checking (in terms of timing and memory usage) are minimal (less than 1 s and 0.5 MB of memory use). Our research highlights the potential of digital twin tools like Cerealia to improve agricultural decision-making under imperfect conditions. Cerealia implementation is publicly available: https://anonymous.4open.science/r/ag-dt-C7BC . 2. Background and Context We now start with a background on weather implications for agriculture and the rationale of this work (§ 2.1 ). We then provide a brief background on how a digital twin architecture like Cerealia can aid the weather-driven agricultural decision-making process (§ 2.2 ). 2.1. Weather Data Impacts on Agriculture Weather data has a significant impact on agriculture. For example, in early spring and late winter, frost damage in fruit crops can hurt fruit production. Fruit growers attempt to mitigate frost using wind machines, propane heaters, and under-tree sprinklers ( evans1999frost ) . To activate those mitigation tools, growers use weather data-driven models and sensing systems that automate and improve frost management with resource conservation

Read the complete text