agriculture · environment · family: it worked in the lab
thirty-five years and thesensor still doesn't work
Real-time on-the-go soil organic matter sensing still cannot deliver reliable variable-rate application
Problem statement
Precision agriculture requires adjusting herbicide, fertilizer, and seeding rates in real-time as equipment moves across a field, because soil organic matter content varies dramatically even within a single field. A patent filed in 1989 by the Purdue Research Foundation described an optical sensor mounted on a vehicle to measure soil reflectance and infer organic matter content on-the-go. The patent reports a laboratory study with a prototype probe and does not report field results. Over 35 years later, the patents cited here do not show that the problem has been solved: a related patent filed in 2014 (US9585307B2) still says that progress on such sensors has been unsatisfactory and that problems remain in calibrating them, and a 2004 review of on-the-go soil sensors found that only electrical and electromagnetic sensors were in wide use.
Why this matters
Both patents say that conventional equipment applies chemicals and plants crops at uniform rates regardless of changes in soil type or organic matter content, which puts too much chemical on some parts of a field and too little on others. The 1989 patent calls soil organic matter a significant variable in soil management and says a soil's adsorption of pesticides, its water holding capacity, and its yield potential are often related to it. The later patent adds that sensing organic matter in real time would let producers apply products without preprocessed maps, and that applying fertilizer only where the crop needs it, based on organic matter content, lowers the producer's cost and reduces run-off and leaching into the watershed.
What’s been tried and why it hasn’t worked
The Purdue sensor used red light-emitting diodes and a photodiode to estimate organic matter content from reflected light, reading soil three to four inches below the surface. In a laboratory study with a prototype probe, reflectance followed organic matter content in a predictable way, though readings became hard to interpret above 6% organic matter. The patent does not say how the sensor performed in real fields. The inventors' 1991 paper in Transactions of the ASAE reports that preliminary field tests were successful for soils from below 1% to over 6% organic matter. A 2004 review of on-the-go soil sensors says the sensor read reflectance at 660 nm, correlated with organic matter at r² = 0.71 for moist soil, and was used to control soil-applied herbicide rates in real time; it lists soil moisture as an interference and says the need for soil-specific calibration was the main limit on wider use. The patent does name problems that such sensors face: surface roughness from tillage practices changes reflectance; crop residue on the soil surface has to be kept out of the reading; soil moisture varies at the surface and affects reflectance; and a single universal relationship between organic matter and soil color does not hold across different soils. The Purdue design tried to reduce these problems by reading below the surface, smoothing the soil ahead of the sensor, shielding it from ambient light, and calibrating for each landscape. A later patent by a different inventor (US9585307B2, Kyle H. Holland) repeats much of the same list of problems and adds auto-calibration, which sets the calibration from the statistics of the readings taken as the sensor moves through the field. The patents do not show whether this solved the problem. The Google Patents record for the Purdue patent lists Veris Technologies patents for mobile soil optical mapping among the later patents that cite it. The 2004 review found that only electrical and electromagnetic sensors were widely used in precision agriculture, and that the readings they give, such as electrical conductivity, cannot be used directly because the value depends on texture, organic matter, salinity, moisture content and other soil properties.
What would unlock progress
A robust multi-spectral or hyperspectral sensing approach combined with machine learning models trained on diverse soil conditions could potentially separate the organic matter signal from moisture, texture, and residue confounders. Both patents already describe a sensor that reads below the soil surface to avoid surface interference; combining such a subsurface probe with rapid spectroscopic analysis could provide more reliable readings. Advances in low-cost hyperspectral cameras and edge computing make previously impractical sensor fusion approaches increasingly feasible.
Entry points for student teams
A student team could design a benchtop experiment comparing optical reflectance, electrical conductivity, and capacitance-based measurements of soil organic matter across multiple soil types with varying moisture levels, quantifying the sensitivity of each method to confounding variables. Alternatively, a team could develop a machine learning model using publicly available USDA soil spectral libraries to predict organic matter content from reflectance data and evaluate its generalization across soil types. Skills in sensor design, soil science, signal processing, and machine learning would be most relevant.
Genes
Structural cousins: stuck the same way, in other fields
Sources
US5044756A, "Real-time soil organic matter sensor," Purdue Research Foundation, Google Patents, accessed 2026-02-10. Expired – Fee Related. Related: US9585307B2, "Optical real-time soil sensor and auto-calibration methods."; J. L. Shonk, L. D. Gaultney, D. G. Schulze, G. E. Van Scoyoc, "Spectroscopic sensing of soil organic matter content," Transactions of the ASAE, 34(5):1978-1984, 1991, (abstract read at ), accessed 2026-09-29; V.I. Adamchuk, J.W. Hummel, M.T. Morgan, S.K. Upadhyaya, "On-the-go soil sensors for precision agriculture," Computers and Electronics in Agriculture, 44(1):71-91, 2004, (read at ), accessed 2026-09-29; CropX, "CropX Hardware," accessed 2026-09-29; Teralytic, accessed 2026-09-29; AgroCares, accessed 2026-09-29. go to source 1 ↗ go to source 2 ↗ go to source 3 ↗ go to source 4 ↗ go to source 5 ↗ go to source 6 ↗ go to source 7 ↗ go to source 8 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
The original inventors (Gaultney, Van Scoyoc, Schulze, Shonk) assigned the patent to the Purdue Research Foundation. The patent expired due to fee non-payment, suggesting the university did not find a commercial licensee. Related to: USDA soil spectral library research, Veris Technologies EC mapping systems. Company pages read on 2026-09-29 describe what three current soil-sensing products are: CropX sells soil sensors installed in the field that read moisture, temperature and salinity at several depths; Teralytic sells a probe installed in an augered hole that reads NPK, pH and moisture; AgroCares sells a handheld scanner that reports on a soil sample in about five minutes, and its page also advertises a product for measuring the organic carbon stock in soil. The CropX and Teralytic pages do not mention organic matter.
Reconciliation 2026-09-29: Fact repair after a site review read this brief against the two patents it cites. Both were fetched on 2026-09-29 from Google Patents: US5044756A (https://patents.google.com/patent/US5044756A/en) and US9585307B2 (https://patents.google.com/patent/US9585307B2/en), both read in full. (1) The brief said the Purdue sensor "used near-infrared reflectance" to read "the soil surface." The patent describes eight red light-emitting diodes and a photodiode, held in a housing that is dragged three to four inches below the surface; the text now says that. (2) The brief said the sensor worked in controlled conditions and "failed" in real fields because of surface roughness, crop residue, soil moisture and calibration that did not transfer. The patent reports only a laboratory study with a prototype and gives no field results. It gives the search for one universal relationship between organic matter and soil color as the reason earlier sensors were unsatisfactory, and it describes roughness, residue and moisture as problems its own design was meant to reduce. The Problem Statement and the history section now follow the patent. (3) The brief called US9585307B2 a "follow-on patent." It was filed in 2014 by a different inventor, Kyle H. Holland, and cites the Purdue patent; the text now says that. The statement that the underlying challenge "persists" was replaced with what the later patent says, which is that progress has been unsatisfactory and calibration problems remain. (4) Claims that are in neither patent were removed rather than re-sourced: the USDA estimate that precision nutrient management could cut fertilizer use by 15–20% and save billions of dollars; dead zones; lower crop yields from uniform application; organic matter as "the single most important parameter"; that there is no commercially dominant low-cost sensor today; that Veris Technologies systems measure only electrical conductivity and need post-processing; that CropX, Teralytic and AgroCares have not solved the problem; the "21+ bankruptcies" figure for 2025; and the inventors' Purdue departments. The statement about fertilizer runoff was kept in the later patent's words: applying fertilizer only where the crop needs it lowers the producer's cost and reduces run-off and leaching into the watershed. (5) The suggestion of a subsurface probe now notes that both patents already describe one. Confirmed against the record and left as written: the patent title, the assignee (Purdue Research Foundation), the four inventors' names, the "Expired - Fee Related" status (the record shows a lapse in 1999 for failure to pay maintenance fees), and the title of US9585307B2. NOT checked in this pass: whether a reliable on-the-go organic matter sensor is sold today, which the two titles of this brief assert and the patents cannot settle; and whether the USDA soil spectral libraries named in the entry points are open for download, because the USDA page timed out. Genes untouched.
Reconciliation 2026-09-29 (re-sourcing): The editor asked for the claims removed in round 5 to be re-sourced, so this pass looked for primary sources outside the two patents. Two were found and read on 2026-09-29: the inventors' own paper, Shonk, Gaultney, Schulze and Van Scoyoc, "Spectroscopic sensing of soil organic matter content," Transactions of the ASAE 34(5):1978-1984, 1991 (abstract read at https://elibrary.asabe.org/abstract.asp?JID=3&AID=31826&CID=t1991&v=34&i=5&T=1; the full text is behind the society's paywall), and the review by Adamchuk, Hummel, Morgan and Upadhyaya, "On-the-go soil sensors for precision agriculture," Computers and Electronics in Agriculture 44(1):71-91, 2004 (read in full at https://www.ars.usda.gov/ARSUserFiles/50701000/cswq-0170-adamchuk.pdf). Restored, at the sources' scope: (1) field performance. The original text said the sensor worked in controlled conditions and failed in real fields. The 1991 abstract says the opposite of the second half of that claim: "Preliminary field tests have proven successful for soils ranging from below 1% to over 6% organic matter." The 2004 review says the sensor read reflectance at 660 nm, correlated with organic matter at r² = 0.71 for moist soil, was used to control soil-applied herbicide rates in real time, and that "the need for soil-specific calibration of the sensor was the main limitation to more widespread use"; its table lists "interference of soil moisture." The history section now says this. Of the four failure reasons in the original text, only calibration that does not transfer between soils and moisture interference are supported, and neither source calls the sensor a failure. (2) The state of the market. The original said no commercially dominant low-cost sensor exists today. The 2004 review says "only electrical and electromagnetic sensors have been widely used in precision agriculture" and that their readings "cannot be used directly since the absolute value depends on a number of physical and chemical soil properties, such as texture, organic matter, salinity, moisture content, etc." The Problem Statement and the history section now carry that finding, dated to 2004; no source was found for the state of the market today. (3) Current products. The original said CropX, Teralytic and AgroCares have not solved the problem and mostly need stationary insertion or soil sampling. Their own pages, read on 2026-09-29 (https://cropx.com/cropx-system/hardware/, https://teralytic.com/, https://www.agrocares.com/), describe an installed in-field sensor for moisture, temperature and salinity, a probe installed in an augered hole for NPK, pH and moisture, and a handheld scanner that reports on a soil sample in about five minutes. The CropX and Teralytic pages do not mention organic matter; the AgroCares page advertises a product for measuring "the organic carbon stock in your soil" (an independent check on 2026-09-29 caught this; the first draft of this note said none of the three pages mentioned organic matter). The Source Notes now say what the products are and nothing about whether they work. Searched for and left out: the USDA estimate that precision nutrient management could cut fertilizer use by 15-20% and save billions (searched USDA ERS and NRCS; no USDA document with that figure was found); the statement that Veris Technologies systems measure only electrical conductivity and need post-processing (the Veris site was fetched three times and gives no product specifications, and the Google Patents record shows Veris holds optical soil-mapping patents, which is the opposite of "EC only"); the inventors' Purdue departments (the paper's full text, which would list them, is paywalled, and a Wiley chapter page was blocked); and the "21+ bankruptcies" figure for 2025, for which no primary source exists. Genes untouched.