POLICY  •  NEWS  •  MARKETS

AG POLICY & MARKETS DAILY

WEDNESDAY, AUGUST 12, 2026   |   SPECIAL REPORT & ANALYSIS

SPECIAL REPORT  |  NASS MODERNIZATION

USDA Can Modernize NASS Fast — but a Full Survey Replacement Will Take Years

Satellites can sharpen crop estimates fast; full replacement is a longer lift

Analysis  ·  August 12, 2026

Tyne Morgan’s Aug. 11 report (link) lays out one of the most consequential potential changes to USDA agricultural statistics in decades. USDA Secretary Brooke Rollins has asked Indiana farmer and former U.S. ambassador Kip Tom to help overhaul the National Agricultural Statistics Service (NASS), beginning with corn and soybeans. Tom says he has already solicited technology proposals, delivered a roughly 20-page plan to Rollins and wants satellite, artificial intelligence and farm machine data tested alongside the current NASS system immediately. 

The most eye-catching possibility is Tom’s suggestion that some crop information eventually could be updated every two weeks rather than monthly, with data processed within a day or two. But there is a major distinction between having a new analytical system operating by the end of 2026 and replacing the official NASS crop estimation system by then. The first is realistic. The second is not. Tom himself says the immediate assignment is corn and soybeans, with wheat likely to follow, while the existing NASS operation spans hundreds of reports covering crops, livestock, prices and other agricultural statistics.

USDA is clearly serious about moving. Rollins formally announced data-modernization listening sessions Aug. 4, and that follows a department-wide statistical-data review launched through a February request for information. USDA’s RFI results showed broad support for modernization but also calls for better coordination, clearer methodology and greater transparency about forecasting and revisions — an important reminder that confidence in USDA will not be restored merely by changing the technology.

Figure 1. NASS at a glance: the scale of the system Kip Tom’s project would begin to modernize. Sources: USDA NASS; American Farm Bureau Federation.

Biggest Point: NASS Is Not Starting from Zero

The impression that USDA currently relies on clipboards while private companies use satellites is too simplistic. NASS already uses satellite remote sensing extensively. Its Cropland Data Layer relies on Landsat and Sentinel imagery and ground-reference data to identify crops and enhance acreage estimates. NASS also already uses remotely sensed vegetation measurements and NASA modeling to generate independent corn and soybean yield indications for the Agricultural Statistics Board in major producing states.

NASS’s Objective Yield surveys simultaneously provide physical measurements for monthly corn, soybean and wheat forecasts, while the Agricultural Yield survey collects farmer-reported expectations. Those multiple indications are combined and evaluated rather than having a single model generate the official answer.

That changes how Tom’s project should be viewed. The revolution is not introducing satellites to NASS. It is potentially moving satellite, administrative and machine-generated data from supporting evidence toward the center of the estimation process.

That is a much more achievable objective.

It also means claims about dramatically higher-resolution commercial satellites need to be kept in perspective. NASS says its Cropland Data Layer generally achieves 85% to 95% producer accuracy for major crop-specific land-cover categories. But that measures the ability to identify what crop is growing, not how many bushels that crop ultimately will produce. Higher spatial resolution can improve field identification without producing an equivalent improvement in national yield accuracy.

Twice the image resolution does not mean twice the yield accuracy.

Why USDA Is Willing to Move So Aggressively

The modernization campaign is partly technological, but it is also a response to the damage done to NASS credibility over the past year.

USDA in January ultimately estimated 2025 corn harvested acreage at 91.3 million acres, 5.2% above the June estimate, an extraordinary revision that helped push corn futures sharply lower when the final figures appeared. Reuters reported that reduced staffing and slower processing of Farm Service Agency acreage data contributed to the difficulties NASS faced in assembling a complete picture.

Figure 2. The credibility problem in one picture: USDA’s final 2025 corn harvested acreage came in 5.2% above the June estimate, an extraordinary revision that helped push corn futures sharply lower. Source: USDA NASS. 

Declining survey participation unquestionably compounds the problem. But there is an important qualification to the article’s discussion of response rates. At USDA’s April 2026 Data Users’ Meeting, NASS official Lance Honig confirmed a 37.6% response rate in the survey then being discussed, but he explicitly cautioned that response rate alone does not determine data quality; sample distribution, nonresponse bias and other statistical factors also matter.

That is critical. A low response rate does not automatically mean a bad estimate, just as millions of satellite pixels do not automatically produce a good one. The central issue is whether the observations are representative, unbiased, measurable and statistically defensible.

Figure 3. The pressure behind modernization: response to NASS row-crop production surveys has slid from 80–85% in the early 1990s to 46% in 2024. Sources: American Farm Bureau Federation; farmdoc daily; USDA Data Users’ Meeting, April 2026.

What USDA Really Can Accomplish by the End of 2026

A great deal can happen quickly if “system in place” means an operating pilot rather than replacement of official NASS estimates.

USDA can have commercial satellite and AI vendors running shadow estimates for corn and soybeans almost immediately. Their models can be reconstructed back to July, as Tom proposes, then compared against August, September, October and final NASS estimates as well as harvest results. That is exactly the sort of parallel testing Tom describes.

USDA also can generate internal estimates much more frequently than it publishes official statistics. Daily imagery can be processed into weekly or biweekly yield indications without changing a single official NASS release. The logical near-term product would be an experimental corn-and-soybean crop monitor that updates more frequently but is clearly separated from the official Crop Production estimate.

Machine data also can be piloted this harvest. Yield monitors provide something satellite systems desperately need: ground truth. If USDA can compare actual harvested yields with the spectral signatures observed throughout the season, models should become better calibrated. Tom correctly identifies that feedback loop as potentially one of the most powerful parts of the system.

Cross-agency administrative information offers another large opportunity. USDA’s “One Farmer, One File” initiative is already intended to unify FSA, NRCS and Risk Management Agency systems. But it is worth noting that NASS is not the primary agency named in that initiative. One Farmer, One File can make administrative acreage information easier to access and reconcile, but it does not itself replace NASS’s independent statistical architecture.

The most credible year-end achievement, therefore, would be: a functioning hybrid corn/soybean pilot, commercial vendors under contract, back-tested satellite models, a limited machine-data calibration program and a framework for incorporating more administrative data.

That would be meaningful progress.

What Is Not Likely by Dec. 31

A wholesale abandonment of farmer surveys or Objective Yield measurements is highly unlikely. Even Tom says enumerators and physical field checks would be reduced gradually rather than eliminated.

Nor is it likely USDA will suddenly replace the established monthly Crop Production report with an official biweekly crop production number this fall. NASS’s Agricultural Statistics Board currently produces the official estimates, and market-sensitive reports are compiled and released under special security procedures.

There is also a major institutional guardrail that receives little attention in the article. NASS is one of OMB’s recognized federal statistical agencies. Current federal regulations require USDA to give the statistical agency sufficient autonomy to set data-quality standards and determine whether statistical products are good enough for dissemination. In other words, the Secretary can order modernization, but an adviser or vendor cannot simply dictate what the official corn yield will be.

Interestingly, those OMB regulations contain a Dec. 10, 2026, deadline for parent agencies to revise policies or organizational arrangements that interfere with their recognized statistical agencies’ fundamental responsibilities. Thus, at roughly the same time USDA wants Tom’s modernization framework completed, it also must be protecting NASS’s statistical autonomy.

That makes a fast technology rollout possible but argues strongly against a political or administrative shortcut around NASS methodology.

Table 1. Separating realistic year-end deliverables from the aggressive reading of “system in place.” Source: Ag Policy & Markets Daily analysis of the Tom proposal, OMB regulations and the FY 2026 NASS appropriation.

There Is Another Hurdle: The Paperwork Reduction Act

If USDA obtains information passively from satellites, there is essentially no farmer reporting burden. But if the new system asks large numbers of producers to transmit yield-monitor or other farm information directly to USDA on a recurring basis, additional federal information-collection requirements come into play.

OMB rules generally do not approve information collections requiring respondents to report more frequently than quarterly unless the agency demonstrates a statutory requirement or other substantial need. New or materially changed federal information collections also generally require OMB clearance.

That does not make biweekly crop statistics impossible. It does suggest that the easiest path to biweekly reporting is satellite imagery, existing administrative records and automated data already available to USDA — not asking farmers to complete another form every two weeks.

That distinction could become very important.

Farmer Data Ownership May Be the Hardest Part

Tom repeatedly stresses that producers must remain in control of their machine data. That is politically essential, but the government will need something more precise than a general assurance that information will be “anonymized.”

NASS currently operates under strong statutory confidentiality restrictions. Information supplied for statistical purposes is protected under CIPSEA and federal agriculture law, and NASS says identifiable information can be accessed only by authorized employees or sworn agents and cannot be publicly disclosed.

A workable machine-data system therefore needs answers to several questions: Who receives the raw yield-monitor file? Does it go directly to NASS or first to a commercial vendor? Can that vendor retain it? Can it be reused to train commercial models? What happens when a farmer withdraws consent? How are field boundaries stripped from published information? And who owns the derived model output?

The technology is the easy part compared with establishing enough trust that thousands of growers voluntarily allow government-linked systems to ingest extremely detailed production records.

Administrative Data Will Help — but It Cannot Simply Replace Surveys Either

There is an attractive argument that farmers already report acreage to FSA and RMA, so USDA should simply use those records instead of asking again.

USDA should make much greater use of those data. But NASS research has long found that FSA and RMA information is not a perfect probability sample of U.S. agriculture. Program participation varies by commodity, geography and operation type, meaning administrative records can have under coverage. NASS research therefore treats such information as extremely valuable auxiliary data rather than assuming it represents the entire farm universe.

That points toward the strongest future architecture:

Satellite observations + FSA/RMA administrative acreage + weather data + probability surveys + Objective Yield ground checks + voluntary machine harvest data.

The important word is plus, not “instead of.”

The biggest modernization opportunity is using each source to correct the weaknesses in the others.

Table 2. The strongest future architecture is additive: each source corrects the weaknesses of the others. The important word is plus, not “instead of.” Source: Ag Policy & Markets Daily analysis; USDA NASS methodology reports.

Corn and Soybeans Are the Easy Test — Livestock and Specialty Crops Are Not

Tom’s decision to begin with corn and soybeans makes sense because they are large-area crops with millions of relatively uniform acres, strong satellite signatures, substantial crop-insurance participation and extensive yield-monitor adoption.

Wheat is a logical next step.

But this approach becomes progressively harder for vegetables, fruits, tree nuts and other specialty crops, and much of NASS’ remaining portfolio cannot be observed from space at all. Satellites cannot determine grain stocks inside a commercial elevator, count hogs in a building, measure milk production, determine cattle placements or calculate prices received by farmers.

NASS publishes roughly 450 national statistical reports across more than 120 crops, 45 livestock items and numerous economic categories. Remote sensing may transform part of that system, but there is no plausible satellite replacement for NASS as an institution.

This is why the article’s body is more realistic than the broadest interpretation of its premise: Tom’s immediate project is really a corn-and-soybean production-estimation modernization project, not a replacement for everything NASS does.

Figure 4. NASS Objective Yield states: corn and soybeans — Tom’s starting point — are measured in the same Midwest core, while wheat (the likely next phase) adds Plains and Western states. Livestock, stocks, prices and specialty crops cannot be observed from space at all. Source: USDA NASS Objective Yield survey program.

Congress Also Has a Say

USDA cannot quietly turn off existing statistical programs. The fiscal 2026 NASS appropriation requires the Secretary to give House and Senate Appropriations committees written notice at least 30 days before discontinuing data-collection programs or reports.

Congress therefore becomes more important once USDA moves from “adding new technology” to eliminating surveys, reports or field operations.

Adding a satellite indication is relatively straightforward. Eliminating the June Agricultural Survey, removing an Objective Yield component or dropping a long-established report is another matter altogether.

Table 3. The institutional guardrails around market-moving federal statistics. Source: Ag Policy & Markets Daily analysis of OMB regulations, CIPSEA and the FY 2026 NASS appropriation.

More Frequent Reports Could Change the Grain Market in Unexpected Ways

Biweekly crop production estimates sound inherently superior to monthly reports, but frequency and quality are separate issues. If each update contains useful new information, more frequent estimates could reduce the amount of uncertainty accumulating between monthly reports and potentially lessen some giant one-day USDA surprises. Markets would receive smaller pieces of information more often.

But there is another possibility: an immature model could cause estimates to bounce around every two weeks as vegetation signals change, weather models update and machine data arrive. Instead of one large monthly USDA event, the market could get two smaller — or occasionally contradictory — ones.

The worst outcome would be two competing USDA numbers, an experimental biweekly estimate and an official monthly estimate, without traders understanding why they differ.

That makes transparency almost as important as accuracy. USDA should publish model-error statistics, revision histories and explanations of the weight given to satellite, survey, administrative and machine data. The department’s own 2026 RFI found strong demand for precisely that kind of methodological transparency.

A Realistic Timeline

Figure 5. Building the new engine without throwing away the old one: pilot and validate in 2026, score and weight in early 2027, run a full season in parallel, then let the hybrid system carry more official weight from late 2027. Source: Ag Policy & Markets Daily analysis of the Tom roadmap.

August through December 2026: This should be the pilot and validation phase. Vendors can begin running corn and soybean models in parallel with NASS, backcast July conditions, compare estimates with the Aug. 12 and subsequent Crop Production reports, and begin ingesting actual 2026 harvest information. USDA can establish farmer-consent, confidentiality and vendor-data rules while holding the listening sessions Rollins has announced. The regular NASS system continues operating throughout this period. NASS’s current calendar and Aug. 12 Crop Production release remain in place.

By December 2026: Tom’s goal of having a “system in place” is credible if that means the hybrid technology platform and methodology have been selected and tested. It is considerably less credible if interpreted as replacing the official corn and soybean forecasting machinery. Tom’s own roadmap calls for identifying contractors, beginning tests and refining the approach before final implementation.

January through spring 2027: USDA should be able to conduct its first rigorous post-harvest scorecard: How accurately did each satellite/vendor model forecast final 2026 yields? At what point in the growing season did it outperform or underperform the traditional indications? Was there systematic state-level bias? USDA could then establish formal weights for a hybrid system and complete any necessary OMB data-collection changes.

The 2027 growing season: This is the first realistic opportunity to operate an enhanced system through an entire crop cycle — planting, acreage determination, vegetative development, reproductive stages and harvest. USDA could plausibly make a public experimental biweekly corn/soybean indicator available while continuing established official reports.

Late 2027 into 2028: After two harvests of parallel testing, USDA would have a much stronger statistical basis for materially reducing some survey and Objective Yield workload and allowing remotely sensed and administrative information to carry more weight in the official estimate. Wheat could move into the system during this period.

2028 and beyond: Specialty-crop applications, RMA loss-adjustment uses and broader integration could expand, but each commodity will require its own approach. Livestock, stocks, prices and many economic surveys will remain fundamentally different statistical problems.

Bottom Line

Tom is right about the central premise: USDA can obtain far more agricultural information passively, continuously and cheaply than it does today, and declining survey participation makes modernization increasingly necessary. NASS itself has already been moving in that direction, including remote sensing, AI-assisted survey processing and other modernization work. 

But the strongest version of the proposal is not “satellites replace NASS.” It is NASS evolves from a survey-dominated system into a hybrid statistical system in which surveys become one source among several.

The most likely outcome is therefore substantial rather than revolutionary: fewer repetitive farmer surveys, fewer physical field observations, much heavier use of satellite and administrative information, machine data providing new ground truth, faster internal estimates and eventually more frequent public information.

What is not likely is a fully satellite-driven official crop-reporting system replacing traditional NASS corn and soybean statistics by the end of 2026. The statistical validation, farmer-data rules, OMB requirements, congressional oversight and institutional safeguards surrounding market-moving federal statistics make that timeline too aggressive.

A veteran USDA watcher emails: “I personally have some reservations about some of the premises for the change. I am not quite understanding why moving away from objective yield — on the ground samples —is favored, yet on-the-ground data is needed to verify any AI/satellite information. Yes, there are limits to humans. But there are limits to machines too. Are all the yield monitors that farmers use calibrated and calibrated accurately? And releases more often may not be better —as you point out —when they conflict with the official estimate. Seems like this is almost being ‘rushed’ along like some other Trump administration efforts. NASS estimates do need to change. And at least Kip Tom is heading this. He will make sure that it is done right. I am just uneasy about what seems to be a rushed effort. How many private satellite firms have tried and failed to come up with satellite-based estimates? Many. Farmers are some of the ones complaining, but they are also part of the problem for not participating in the NASS surveys. Kind of like bitching about an election result if you don’t vote.”

Upshot: The end-of-year target should be viewed as a deadline for building the new engine — not for throwing away the old one.

AG POLICY & MARKETS DAILY   |   SPECIAL REPORT  |  NASS MODERNIZATION — WEDNESDAY, AUGUST 12, 2026

AG POLICY & MARKETS DAILY — WEDNESDAY, AUGUST 12, 2026   |   PAGE 1