A warm patch does not automatically mean more irrigation.
Use imagery to form a hypothesis
Combine thermal candidates with soil moisture, weather, irrigation history and crop stage. Different capture times and sunlight complicate temperature comparison. Check reflections, bare soil and shadow before interpreting crop temperature.
Inspect the water system too
Inspect pumps, valves, lines, blockages, pressure and drainage. Do not rule out nutrient, disease or soil causes from an image alone. Require field checks and a responsible decision before changing irrigation settings.
Track both water and production
Evaluate water use together with yield, quality, energy, labor and variability. When transferring results, preserve rainfall, soil, crop and irrigation-system context.
Do not assign a package's results solely to drones
IRRI's account of low-emission rice in Can Tho combines fertilizer and pesticide drones with water-level sensors and other agronomic measures. It illustrates integration, rather than a comparison isolating irrigation decisions based on drone imagery. This guide therefore proposes identifying changes in water management, nutrient management, work methods and weather when interpreting outcomes. A before-and-after difference following a technology package does not establish the causal contribution of one aircraft or analytical method.
Test water management separately from image information
Start with a focused question: does imagery improve decisions under the same water-management policy? Compare multiple reasonably matched units using imagery against units relying on the usual observations. If the water-management method also changes, distinguish the presence or absence of that change from the use of imagery and record the comparisons. Randomize allocation where feasible and document soil and elevation differences. Adjacent observation points influenced by the same waterway or water-level operation should not be counted as independent experimental units. Label small demonstrations without sufficient comparison units as exploratory observations, and seek confirmation across periods instead of extrapolating one favorable result.
Keep a decision trail that can be evaluated
Record image acquisition, field inspection and actual irrigation or drainage changes as separate events. Before acting, agree which condition in which area should improve and what would trigger a reversal; do not replace evaluation measures afterward with more favorable ones. State the measurement boundary for water intake or equipment runtime, and compare rainfall, yield, quality, energy and inspection time alongside area. Include the added cost of processing and checking imagery. Retain days when flights were impossible, sensors failed or the farmer declined the recommendation, rather than evaluating usefulness only on successful capture days.
| Evaluation question | Record | Interpretation boundary |
|---|---|---|
| Did information change a decision? | Decision before and after, with owner | Viewing an image is not itself an improvement |
| Did water use change? | Measurement boundary, volume, rain and actions | Pump runtime alone does not establish field water volume |
| Did farm outcomes improve? | Yield, quality, cost and work time | Do not define success only by water savings |
Include cases where more irrigation is not the answer
An interface should present a short list of hypotheses for a warm area rather than immediately offering an 'add irrigation' action. Route possible pipe or pressure problems to equipment inspection, and concerns about drainage or roots to the agronomic adviser. Avoid changing water delivery across the whole field while the cause remains unresolved. Stale comparison photographs or ground measurements should weaken the decision and return it to review. Tell the farmer what to inspect today, whom to contact and when to reassess. Agree fallback conditions and a contact route so an absent decision owner does not lead to an automatic change, and retain reasons for rejected suggestions to improve the service.