📄 agriculture

Rice evapotranspiration estimation and irrigation optimization based on coupling UAV multispectral and thermal infrared imagery with the FAO-56 model

This study proposes a novel framework that integrates UAV multispectral and thermal infrared imagery with the FAO-56 model and NSGA-II optimization to accurately estimate rice evapotranspiration and identify optimal water-nitrogen management strategies that balance high yield, resource efficiency, and low greenhouse gas emissions in Northeast China's cold black soil region.

Original authors: Yan Sha, Tianyang Gao, Aizheng Yang, Xiaofang Wang, Shengbo Li, Lei Ma, Nan Chai, Mo Li

Published 2026-08-13

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yan Sha, Tianyang Gao, Aizheng Yang, Xiaofang Wang, Shengbo Li, Lei Ma, Nan Chai, Mo Li

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are a chef trying to cook the perfect bowl of rice. You know that water and fertilizer are the secret ingredients, but getting the balance right is tricky. Too much water, and the rice gets soggy; too little, and it's crunchy and sad. Too much fertilizer, and you waste money and pollute the air; too little, and the rice stays small. For decades, farmers have guessed the right amounts based on old habits and the weather, but it's like trying to hit a moving target with a blindfold on. This is especially hard in cold places where the growing season is short and the weather is unpredictable.

Scientists have been trying to solve this puzzle using a concept called "evapotranspiration." Think of this as the rice plant's personal sweat rate. Just like you sweat when you run or when it's hot, plants "sweat" water through their leaves to cool down and grow. If you know exactly how much a plant is sweating every single day, you can give it just enough water to keep it happy without wasting a drop. The big question is: how do you measure this invisible sweat across a whole field without walking through every single plant? This is where the story of high-tech farming begins, turning fields into data-rich playgrounds where drones act as the eyes of the farmer.

The Drone Detective and the Rice Sweat

In this study, a team of researchers in the cold, black-soil regions of Northeast China decided to upgrade rice farming from "guessing game" to "precision science." They wanted to figure out exactly how much water and fertilizer rice needs to grow big, save water, and keep the planet cool, all at the same time. To do this, they built a high-tech system that combines flying robots (drones), fancy math, and a classic farming rulebook.

The Flying Eyes
The researchers sent two types of drones over their rice fields. One drone carried a "multispectral" camera, which sees colors humans can't, like the specific shades of green that tell us how healthy the leaves are. The other drone carried a "thermal" camera, which sees heat. This is crucial because a thirsty plant gets hotter, just like your skin gets hot when you're running a fever. By looking at the heat and the greenness of the rice, the drones could tell exactly how much the plants were "sweating" (evapotranspiration) on any given day.

The Old Rulebook vs. New Data
Usually, farmers use a famous guide called the FAO-56 model to guess water needs. Think of this model as a standard recipe book. But standard recipes don't account for the fact that today might be hotter than yesterday, or that one patch of rice might be thirstier than the next. The researchers realized that instead of just following the recipe blindly, they could use the drone data to update the recipe in real-time. They used the drone's "greenness" readings to figure out how much of the ground was covered by leaves, and the "heat" readings to see if the plants were stressed. This allowed them to calculate the daily water needs with much higher accuracy than before.

The Balancing Act
Once they knew exactly how much water the rice was using, they faced a tricky math problem. They wanted to find the "Golden Mean" that would:

Grow the biggest harvest.
Use the least amount of water.
Use the least amount of fertilizer.
Produce the least amount of greenhouse gases (which warm the planet).

The problem is that these goals often fight each other. To get the biggest harvest, you usually need more water and fertilizer, which creates more pollution. To save water, you might get a smaller harvest. To solve this, the researchers used a super-smart computer algorithm (called NSGA-II) that acted like a referee, testing thousands of different combinations to find the best possible compromise.

What They Found
The study confirmed that water and fertilizer work together like a dance team; if one steps out of line, the whole performance suffers. They found that:

The Best Combo for Yield: Giving the rice 80% of the maximum water and 155 kg/ha of nitrogen fertilizer produced the most rice (about 11,883 kg per hectare).
The Best All-Rounder: However, the "perfect" solution wasn't the one with the most rice. The computer found a "Scenario S5" that was the best overall balance. This plan called for 669.94 mm of irrigation and 117.48 kg/ha of nitrogen.
The Result: This balanced plan produced a massive 11,473.43 kg/ha of rice (almost as much as the maximum!), but it used less water and fertilizer, and it kept the environmental impact lower.

Why It Matters
The researchers discovered that their drone-based system was incredibly accurate. When they compared their drone calculations to the actual water measured in the soil, the numbers matched up with a correlation of 0.89 (a very strong match). This means the drones could predict the rice's "sweat" almost perfectly.

They also proved that you don't have to choose between a big harvest and a clean planet. By using real-time data from drones to guide the water and fertilizer, farmers can stop guessing and start optimizing. Instead of pouring water and fertilizer on a schedule that might be wrong, they can give the rice exactly what it needs, exactly when it needs it.

In the end, this study suggests that the future of farming in cold regions isn't about working harder; it's about watching smarter. By letting drones be the eyes and computers be the brain, farmers can grow more food, save precious water, and keep the air cleaner, all without sacrificing the size of their harvest. It's a win-win-win for the rice, the farmer, and the planet.

Technical Summary: Rice Evapotranspiration Estimation and Irrigation Optimization via UAV-FAO-56 Coupling

Problem Statement
Rice production in the cold black soil regions of Northeast China faces a critical challenge: balancing food security with environmental sustainability. Current water and nitrogen management practices rely heavily on empirical, experience-based decisions and uniform inputs. These static approaches fail to account for the highly dynamic spatial and temporal variations in soil conditions, crop growth, and water-nitrogen requirements, particularly in cold regions where early-stage soil heat is insufficient and meteorological factors fluctuate. This disconnect leads to water waste, low nitrogen use efficiency, and elevated greenhouse gas (GHG) emissions. Furthermore, existing evapotranspiration (ET) estimation methods often struggle with accuracy in paddy ecosystems due to the complex interplay of soil evaporation, crop transpiration, and surface water, and many studies fail to bridge the gap between remote sensing data retrieval and actionable, multi-objective irrigation optimization.

Methodology
This study proposes a novel framework that integrates Unmanned Aerial Vehicle (UAV) remote sensing with the FAO-56 dual crop coefficient model and the NSGA-II multi-objective optimization algorithm.

Experimental Design: Field experiments were conducted at Farm 856 in Heilongjiang Province (2025) using a split-plot design. The main plots involved four controlled irrigation lower-limit thresholds (60%, 70%, 80%, 90% of soil moisture), and subplots involved four nitrogen application rates (0, 110, 155, and 200 kg/ha).
Data Acquisition: UAVs (DJI Mavic 3 Multispectral and DJI Matrice 30T) captured high-resolution multispectral and thermal infrared imagery at a 30 m altitude during key growth stages. Ground truth data, including soil moisture, leaf area index (LAI), and greenhouse gas fluxes, were collected synchronously.
ETc Estimation Framework: The study utilized the FAO-56 dual crop coefficient approach ( $ET_c = K_s \times K_{cb} \times ET_0$ $E T_{c} = K_{s} \times K_{c b} \times E T_{0}$ ).
- Basal Crop Coefficient ( $K_{cb}$ ): Retrieved using multispectral vegetation indices (NDVI, SAVI, EVI) to estimate fractional vegetation cover.
- Water Stress Coefficient ( $K_s$ ): Quantified using the Crop Water Stress Index (CWSI) derived from thermal infrared canopy temperatures.
- Validation: The estimated daily $ET_c$ was validated against actual evapotranspiration ( $ET_a$ ) calculated via the soil water balance equation.
Multi-Objective Optimization: An NSGA-II model was developed to optimize water-nitrogen management. The model aimed to simultaneously:
- Maximize grain yield.
- Maximize Irrigation Water Use Efficiency (IWUE).
- Maximize Partial Factor Productivity of Nitrogen (PFPN).
- Minimize Global Warming Potential (GWP).
- Decision variables included daily net irrigation amounts and nitrogen application rates, constrained by crop water requirements, available water resources, and local fertilizer limits.

Key Results

Model Performance: The coupling of UAV multispectral and thermal data significantly improved ET estimation accuracy. The NDVI-CWSI coupled model achieved the highest performance with an $R^2$ of 0.89 and a Root Mean Square Error (RMSE) of 0.39 mm/day, representing a 10.3% reduction in RMSE compared to traditional methods. NDVI showed the strongest correlation with the basal crop coefficient ( $R^2 = 0.85$ ).
Treatment Effects: Water, nitrogen, and their interactions significantly affected yield, IWUE, PFPN, and GWP. While treatments W3N2 (80% moisture + 155 kg N/ha) and W3N3 (80% moisture + 200 kg N/ha) produced the highest yields (~11,883 and ~11,436 kg/ha, respectively), the W2N1 treatment (70% moisture + 110 kg N/ha) demonstrated the best comprehensive performance with a TCQ value of 0.65.
Optimization Outcomes: The multi-objective optimization revealed distinct trade-offs between yield, efficiency, and emissions. Scenario S5 was identified as the optimal solution, offering the highest coordination degree (0.86). This scenario recommended an irrigation amount of 669.94 mm and a nitrogen rate of 117.48 kg/ha, predicting a yield of 11,473.43 kg/ha with an IWUE of 1.71 kg/m³ and a GWP of 12,473.54 kg CO₂-eq/ha.

Significance and Claims
The authors claim that this study provides a robust methodological framework for shifting rice management from static, empirical schedules to dynamic, data-driven precision agriculture. By coupling UAV-derived evapotranspiration with multi-objective optimization, the study demonstrates that it is possible to reconcile competing goals of high yield, resource efficiency, and emission reduction in cold-region rice production.

Specifically, the paper asserts that:

Accuracy Improvement: The NDVI-CWSI coupled approach effectively overcomes the limitations of traditional models in paddy fields, providing the daily-scale precision necessary for dynamic water-nitrogen regulation.
Decision Support: The integration of real-time crop water consumption data into the NSGA-II model allows for the generation of actionable irrigation schedules that balance economic and environmental benefits, avoiding the pitfalls of single-objective optimization.
Regional Applicability: The framework offers a viable pathway for improving water-nitrogen coordination in cold regions, where short growing seasons and specific climatic constraints make precise management critical for sustainable production.

The study concludes that while uncertainties remain regarding parameter transferability and soil evaporation partitioning, the proposed UAV-FAO-56-NSGA-II framework provides reliable data support for developing green, high-yield, and precise irrigation strategies for cold-region rice.

The Drone Detective and the Rice Sweat

Technical Summary: Rice Evapotranspiration Estimation and Irrigation Optimization via UAV-FAO-56 Coupling

More like this