Science

Could AI help rescue coastal dead zones

Excess fertiliser is choking our waterways but help could be on its way through low-cost new developments.

Could AI help rescue coastal dead zones
The algal bloom in Lake Erie in 2011 could be viewed from above

Half of the global population is supported by nitrogen fertiliser. Their development and use in the Green Revolution hugely increased crop yields, but now, across the globe, only 35% of those applied are used by harvested crops. The excess fertiliser amounts to 75 million tonnes annually and has devastating consequences on the environment. Nitrogen sensors could allow growers to fine-tune the application of nitrogen fertilisers to be applied at the optimum time and quantity.

Overfertilisation has rendered 12% of once-arable land unusable. Fertiliser runs off into waterways, where nitrogen, which is essential for plant growth and often in limited quantities, allows plants such as algae to grow exponentially. Bacteria can then feed on this algae and consume the water’s oxygen, creating the dramatically named “dead zones”, of which there are now over 500 identified in coastal areas. Furthermore, as the oxygen runs out, bacteria turn to nitrogen compounds instead, and consume their friends to release nitrous oxide - a greenhouse gas 300 times more potent than carbon dioxide.

Fields Above Hawes 6965
Photo: Nilfanion, CC-BY-SA-4.0
Space missions and soaring profits?Testing nitrogen content in soils and crops has the potential to significantly raise profits, but results have so far been inconsistent. Measuring soil nitrogen is often a lengthy and costly process that requires samples to be sent to a lab. By the time the information gets to the growers, the soil nitrogen content is more than likely to have changed. Measuring crop nutrient status has either required farmers to painstakingly “scan” crop leaves or rely on low-resolution satellite pictures. But better techniques will soon be on the market: low-cost soil sensors should be available for commercialisation within 3-5 years, and NASA is already planning to send hyperspectral sensors into space. The market size for agricultural sensors is estimated to grow from $1.55 billion in 2021 to $3.7 billion by 2028, and expand at a compound annual growth rate (CAGR - the mean annual growth rate of an investment over time) of 13.6%. With innovations such as these, there’s hope we could soon see life return to our dead zones.

Growers are caught in a dilemma: too much fertiliser and the environment and their wallets suffer; too little and their crop yields may diminish. New sensors developed by researchers at Imperial College London monitor the soil for ammonium, pH, and conductivity. They combine this data with weather data and application time using machine learning to predict soil nitrogen levels over the following 12 days. The ammonium data can accurately predict nitrogen uptake by plants (microbes convert ammonia into nitrates, which plants can absorb). According to lead researcher Dr Max Grell, this technology should empower growers by allowing them to fine-tune fertilisation to the specific needs of the soil.

Techniques are advancing beyond soil nutrient measurement and even beyond this world, quite literally. A research team at the University of Illinois recently published their findings from a novel experiment. They flew a plane, equipped with powerful hyperspectral sensors, over a corn crop three times. The sensors were designed to pick up certain infrared and far-red light radiation, the type emitted by leaves under nutrient stress. Their sensors were consequently able to estimate the nitrogen content of the plants themselves, with an impressive 85% accuracy. This is a leap forward from the lower-resolution satellite techniques. However, satellites still have a role as they could be equipped with hyperspectral sensors, which would allow growers to determine the nutrient status of huge areas of crops at the start of the growing season.

From Issue 1793

Discover stories from this section and more in the list of contents

Explore the edition