Higher levels of no-till agriculture associated with lower PM2.5 in the Corn Belt

No-till approaches to agricultural soil management have been encouraged as a means of reducing soil erosion, reducing water pollution, and increasing carbon sequestration. An understudied additional benefit of no-till approaches may be improvements in local air quality. No-till approaches involve reductions in both machinery use and soil erosion, both of which could lead to improvements in air quality. We leverage recent advances in remote sensing and air pollution modelling to examine this question at a landscape scale. Combining data on daily PM2.5 levels with satellite measures of no-till uptake since 2005, we show a strong association between increasing adoption of no-till and reductions in county average PM2.5 pollution over more than 28 million hectares of cropland in the American Corn Belt. The reduction in local pollution implies substantial monetary benefits from reductions in mortality that are roughly one-fourth as large as the estimated carbon benefits. The benefits of mortality reductions are also, by themselves, nearly equal to the current monetary costs of subsidizing no-till practices.


Introduction
Agriculture is a major source of air pollution around the world.In some parts of the world, including parts of the U.S., emissions from agriculture are the largest anthropogenic source of particulate pollution. 1In the United States, for instance, agriculture accounts for approximately 16% of PM 2.5 emissions. 2Emissions from crops grown for export from the U.S. have been estimated to generate air pollution externalities whose costs exceed $36 billion annually, primarily due to ammonia emissions. 3More generally, agriculture is the largest single sector source of damages from air pollution in the United States. 4ile much of this pollution is due to fertilizer application and animal husbandry, soil management practices also contribute to localized increases in particulate pollution. 1,5Roughly 25% of air pollution damages attributable to agriculture in the United States is due to direct PM 2.5 emissions from "crop related activities." 4 Reducing this pollution could have sizeable health benefits. 6,7,8Improved soil management practices, in particular low or no-till approaches have been widely promoted as a way to reduce carbon pollution and increase carbon sequestration rates. 9,10These claims are controversial, with several recent studies suggesting that the carbon benefits may be overstated. 11,9But, as with many policies that have carbon benefits, no-till may also have co-benefits in the form of reduced local air pollution.
The potential benefits of no-till that derive from reducing local air pollutants have received less attention than carbon benefits. 12Low disturbance tillage systems may substantially reduce local particulate pollution from windblown dust, with exact results dependent on the no-till practice used, the type of cropping system, and the location of the farms. 13,14,15Much of this research has focused on the air pollution impacts of no-till over a small scale and in areas that are dry relative to the corn and soy growing regions of the American Midwest.Broadly this work has focused on the benefits of no-till in reducing wind erosion from a soil management perspective rather than examining the consequences for air pollution.One examination on the Columbia Plateau in the U.S. that does look specifically at the impact of no-till on particulate matter finds that no-till leads to significant reductions in PM 10 emissions from treated fields. 16e benefits of no-till for local air pollution, applied over a wide geographic region and one that typically has wet soils, remain understudied.
Understanding the magnitude of the improvement in local air pollution due to adoption of no-till is important in determining the appropriate level of policy support for no-till adoption.
In other settings, policies with carbon mitigation benefits have had local air pollution co-benefits whose value exceeded the cost of the policy without regard to the carbon benefits. 17,18Including the benefits of reductions in local air pollutants due to no-till in the determination of the proper level of policy support may be especially relevant given disagreement about the presence and size of the carbon benefit.

Our approach
To estimate the impact of adoption of no-till agriculture on local air pollution levels we leverage two recent advances in data collection and modeling.One obstacle to estimating the benefits of no-till agricultural has been poor data on the uptake of no-till over the entire U.S. Corn Belt growing region.Historically, analysis of the uptake of no-till practices has depended on five year surveys conducted by the Conservation Technology information Center and the National Resources Inventory database (e.g.ref. 19,20 ).These sources do not provide data that is spatially or temporally granular.Instead here we rely on advances in remote sensing and machine learning to measure uptake of no-till at approximately the field level and at an annual frequency. 21This enables us to conduct an examination of how year-to-year changes in no-till usage in a given county impact the air quality of that county.
A second obstacle has been the lack of data on PM 2.5 levels with sufficient spatial and temporal resolution.Data based on ground monitors generally offers daily, or even hourly, information on pollution levels but over a very limited geographic region.These monitors are rarely located in the rural regions that are most likely to be impacted by the changes in tillage practices that we examine here.The problem of limited geographic scope can be solved by using remotely sensed data on pollution levels that takes advantage of satellite data to provide estimates of pollutant levels across a large geographic area.However, the increase in geographic scope comes at the cost of low temporal frequency.Because we expect changes to local air pollutants due to the uptake of no-till to be concentrated at the times when no-till practices are most likely to occur, the lack of temporal frequency makes it difficult to detect impacts with existing remotely sensed pollution data.
In the current study, we utilize a recently released dataset of highly localized pollution estimates for the United States at a daily scale. 22These estimates combine data from multiple sources including satellite measures of aerosol optical depth (AOD), emissions inventories, and ground-based pollutant monitors.These data are combined in a multi-model ensemble approach to estimate a pollution surface that provides daily estimates of PM 2.5 on a 1km×1km grid for the entire U.S.
We estimate the relationship between changes in no-till adoption and pollution levels econometrically with a two-way fixed effects model.Our model examines how daily pollution levels in a county change during the period that farmers are most likely to be engaged in tillage in that county.We define the period when tillage is most likely to occur based on data on crop specific harvest times by state reported to the USDA (for full details see SI1.1). 23We examine how pollution levels in this "tillage period" change from year to year within a county as the level of no-till adoption changes in that county.We define no-till adoption as the share of the area in that county using no-till in a given year.. 21 By examining year to year changes in pollution within a county, our model accounts for time invariant features of counties that may impact both no-till uptake and pollution levels.We also include a variety of time fixed effects and trends to control flexibly for temporal patterns in pollution levels.In particular, we include state by year fixed effects to account for the general down-trend in pollution levels over the period we study. 24 also control flexibly for weather conditions at the daily level.

Results
There has been a general decline in the average level of PM 2.5 across the entire United States over the period we study. 25,24,26In Panel A of Figure 1 we document that general pattern also appears in our data focusing on the four weeks of the year when tillage is most likely to occur.
From 2005 to 2016 average daily PM 2.5 levels in the counties we study during the four weeks that tillage is most likely to occur fell by roughly 35%, from nearly 10µg/m 3 to 6.5µg/m 3 .Panel B of Figure 1 shows the descriptive fact that motivated our analysis -over the same time period the adoption of no-till approaches have increased substantially.From 2005 to 2016 the average share of farmland in no-till across all counties in our sample increased more than 20% from just over 40% of fields to roughly 50% of fields.
The largest changes in no-till have occurred in the upper Midwest, namely North and South Dakota, Nebraska, and Iowa (Figure 2).The largest reductions in pollution have occurred in the southern part of our sample in Ohio, Indian, and Illinois.This reduction in pollution may have been driven by changes in coal-fired power generation in the Ohio River Valley.
We find that pollution is, on average, higher during the month in which we predict tillage to occur compared to the rest of the year.On average PM 2.5 during tillage times is 0.18 µg/m 3 higher across our sample than during non-tillage.This averages across all years and counties in our sample.
Our primary result is that in years with a higher share of crop pixels managed in no-till, the increase in PM 2.5 during tillage period relative to non-tillage is smaller (Table 1).Our results suggest that a percentage point increase in the share of land in no-till management in a given year is associated with a 0.01 µg/m 3 reduction the increase PM 2.5 during the tillage period relative to the non-tillage period.That reduction in PM 2.5 represents a 0.11% decline from the mean while a one percentage point increase in no-till is a roughly 2.22% increase from the mean level of no-till uptake.
Our result is robust to a variety of specifications, including various different sets of daily, weekly, and day of sample fixed effects (Table 1).Our results are also robust to dropping   Notes: Robust standard errors are clustered at the state level.Tillage time is calculated as the period in a four week window following the most active harvest week in each state according to the USDA."Avg.no-till pct" is the percent (0-100) of pixels in a county that are assessed as employing no-till during a given year.PM is measured using data from. 22All regressions include quadratic controls for daily precipitation and daily maximum temperature.Coefficients on the individual indicator for tillage time and the continuous measure of no-till are suppressed in the table.Additional controls are measures of the height of the boundary layer, a quadratic in wind speed and a set of indicators for which direction the prevailing wind each day.
Coefficients indicate the change in PM (µg/m 3 ) during the period of tillage compared to the non-tillage period in a given county for a percentage point increase in the amount of land in no-till in that county.The average level of PM and the standard deviation for the counties included in each sample is reported in the rows labelled mean and SD.Compiled 18 Apr 2022.
temperature and precipitation controls (Column 1), omitting the period immediately prior to spring planting during which some tillage may occur (Column 4), and including controls for the level of no-till up-take in upwind counties (Column 5).We also test whether dropping Ohio, Illinois, and Indiana -the states with the largest secular changes in PM 2.5 levels -change our results.It does not.Our results are also robust to alternative clustering of our standard errors including clustering at state × year and by state and year separately.
The large reduction in PM 2.5 levels across the entire country during our sample period raises a potential concern that our results are the consequence of the intersection of two unrelated trends: the decline in pollution nearly everywhere in our sample and the overall increase in no-till.Specifically, because pollution declined in most of the counties in our sample and most counties had an increase in the use of no-till our results are only detecting the correlation between these trends.An implication of this hypothesis is that the specific geographic pattern of no-till uptake does not matter as long as it is generally positive.Conversely, the causal interpretation of our results implies that changing the pattern of no-till uptake would result in a zero effect as the causal link is broken.We can test this hypothesis by conducting a placebo test, akin to a randomization inference test, where we randomly assign no-till rates across counties and re-estimate our primary specification.The hypothesis of a causal relationship between no-till up-take and changes in pollution implies that this exercise should result in a mass of estimates clustered around zero.In contrast, if we are only detecting non-causal correlation the estimates from this exercise should be clustered around our reported effects and away from zero.
We conduct our test in two stages.We first randomly assign no-till uptake levels to all counties in our sample in each year, maintaining the observed distribution of no-till uptake within each year.This maintains the general positive trend in no-till uptake across the whole sample.We then re-estimate our primary specification.We do this 10,000 times.In results reported in Figure 3 we show that this simulation results in a mass of estimates clustered around zero and substantially different from our main results.Notes: We report summary statistics for our key variables here.Planted pixels measures the number of pixels in a county that are recorded as planted, scaled to 100,000 pixels.Tillage share reports the share of those that are measured as being in no-till.PM2.5 reports the average annual level of PM2.5 in the county.We aggregate data on PM2.5 from [22] to the county level.
Using the estimates in Table 1 we can estimate the share of the reduction in PM 2.5 levels that is due to increased use of no-till.Collectively the average share of crop pixels in no-till has increased by 9.67 percentage points from the first two years of our sample to the last (Table 2).
Average PM 2.5 levels during the tillage period have fallen 3.29 µg/m 3 over that same period.
Combining our estimates of the impact of increasing no-till with the change in share of cropland in no-till suggests that the increase in no-till management can account for 3.8% of the 3.29 µg/m 3 decrease in PM 2.5 .

Local air pollutant benefits compared to carbon benefits of no-till
Whether no-till reduces carbon emissions is controversial.For example, a recent summary of nature-based carbon mitigation approaches 11 intentionally did not include no-till because of questions about its effectiveness in reducing carbon emissions.They argue the available evidence suggests that carbon benefits of no-till only occur when fields have been in no-till management continuously for more than a decade, which is not the case in much of the U.S. Corn Belt.
On the other hand, a global review of studies on low disturbance soil management suggests that the median rate of carbon sequestration that can be achieved after adopting no-till approaches is roughly 0.2 t/ha/yr. 27Applying global model-based approaches to the U.S. suggests much of the Midwest would have sequestration rates in the range 0.05-0.1 t/ha/yr. 28A modelbased approach specific to the United States suggests that median accumulation rates are closer to the globally observed 0.2 t/C per ha. 19Monetizing these various estimates suggests that the annual benefits of carbon reductions, if the entire U.S. Corn Belt adopted no-till, would be roughly 5 billion (USD2020) annually. 19Using the value of a statistical life used by the U.S.
EPA for cost-benefit analysis, adjusted to 2020 dollars, of roughly 9.5 million, the reductions in air pollution from adopting no-till would need to save roughly 500 statistical lives per year to equalize these benefits.
Existing work examining the mortality benefits of reducing PM 2.5 estimates a 1 µg/m 3 decrease in annual PM 2.5 results in 7.17 fewer deaths per 100,000 individuals. 29Scaling this to a single month reduction in PM 2.5 suggests reductions of 1 µg/m 3 for a month would save 0.6 lives.There are roughly 33 million individuals in our study area.Full adoption of notill agriculture suggests an increase of 50 percentage points relative to current levels, which suggests a 0.65 µg/m 3 decline in PM 2.5 levels across the study area for the month that tillage occurs.Putting all this together suggests a decline of roughly 130 deaths per year.That indicates the air pollution benefits of adopting no-till might be 25% of the carbon benefits or 1.25 billion (USD2020) annually.This ignores any other benefits of reducing pollution, which could be sizeable as reductions in pollution have been shown to have meaningful consequences for students' academic performance, morbidity, and labor productivity. 30,31,32igure 3: Placebo test-This figure reports the density plot of point estimates generated by creating a placebo distribution of no-till shares that matches the observed distribution in each year of the sample and randomly assigning counties no-till shares from this distribution.We then re-estimate our primary specification with the randomly assigned no-till shares and record our coefficient of interest.We do this 10,000 times to generate the density of estimates.The actual impact that we estimate is show in the dashed red line.

Cost benefit analysis
The USDA subsidizes the uptake of no-till agriculture through the Environmental Quality Incentives Program (EQIP).This program is designed to encourage farmers to take up conservation farming practices through technical and financial assistance.No-till agriculture is one practice that it encourages but reductions in air pollution are not given as a reason supporting the subsidization of no-till specifically. 33Despite disagreement about no-till's carbon benefits, sequestration is offered as a benefit of the subsidies.Comparisons of the costs of subsidizing no-till and the potential carbon benefits have suggested that EQIP payments may be a cost-effective way of encouraging carbon sequestration relative to other options. 19 use our estimates of the potential mortality benefits of reduced air pollution associated with the uptake of no-till to estimate whether increases in EQIP payments could be justified due to air pollution benefits.The USDA NASS reports that in 2017 there were just over 28 million hectares of land in corn and soybean production in the states we study.We focus on these crops as they are the largest crops by area planted in the region.If we assume that every hectare planted in these crops utilized no-till approaches and received an equal subsidy payment our estimated benefits of 1.25 billion USD suggest that the air pollution benefits alone would justify an EQIP payment of $43 per hectare.That represents somewhere between 50% and 120% of the current EQIP payments for using no-till in these states. 19ese results suggest that the air pollution benefits of adopting no-till may represent a substantial portion of the monetary benefits of adopting the practice.Our estimated benefits implicitly assume that all farmers would take-up no-till agriculture at these higher subsidy rates.
This may not be true, as the share of farmers using no-till increases, the cost of encouraging the marginal farmer to adopt it is likely to increase and may surpass these subsidy rates.We leave estimation of the total cost of achieving 100% adoption of no-till to future research.Even if 100% adoption cannot be achieved our results suggest that inclusion of air pollution benefits would support higher EQIP payment rates.

Discussion
Adoption of no-till agricultural practices has benefits for erosion and soil health, and it may have carbon benefits as well.Less well understood are the potential benefits from reducing local air pollutants.We show here that adoption of no-till throughout the U.S. Corn Belt is associated with declines in local PM 2.5 levels.The implied reductions in mortality due to these reductions in pollution suggest that the monetary benefits of adopting no-till may be substantially higher than previously estimated, which have focused primarily on the environmental and carbon reduction benefits.Including benefits from reduced mortality due to lower air pollution could raise these benefits substantially.
Our analysis is subject to several limitations.The adoption of no-till agricultural approaches that we study is not randomly assigned.Although there are clear theoretical and practical reasons why no-till would reduce pollution levels (e.g.reductions in wind erosion, soil disturbance, and emissions from machine operation), it remains possible that the reductions in pollution we observe are due to other changes in these counties.Notably, to affect our estimates these changes would have to both be not fully captured by the county and year fixed effects, and correlated with the residual variation in tillage and PM 2.5 .
One possibility is that our results are attributing reductions in local pollutants due to broad shifts in power generation to the adoption of no-till agriculture.We have attempted to address this by examining how our estimates change under random assignment of no-till shares.If our estimates were due simply to general declines in pollution the results from this random assignment should be similar to our estimated effects.They are not and so we believe that our estimates are capturing the impact of adoption of no-till approaches.But more examination of this question with better econometric identification is warranted.
Another potential concern is that our calculation of the benefits of reduced air pollution assumes the mortality benefits are uniform across our study area.The impact that air pollution has on mortality can be highly variable across space. 34Because we study reductions in pollution due to changes in agricultural practices the largest reductions in emissions will be in rural areas that have lower populations and so may have lower benefits from reductions in mortality.
However, emitted PM 2.5 can travel long distances and there are several large population centers in our study area.If rural reductions in emissions reduce pollutant levels in these population centers, our mortality estimates are more likely to be accurate.
Much of the existing research on no-till has focused either on highly local benefits to soil health or water quality or global benefits of higher carbon sequestration.Our estimates here offer evidence that there may also be regional benefits in the form of reductions in local air pollutants.
Like in other settings the benefits of these reductions in air pollutants may be substantial.
More research is needed to provide additional insight into the size of these potential benefits and identify a clear causal link between adoption of no-till approaches and changes in air pollution.Our estimates also point to another channel through which agriculture influences local air pollution and highlight the importance of research on the links between air pollution and agricultural practices.

Figure 1 :
Figure 1: Change in PM 2.5 and land in no-till-Averages are calculated over the full set of counties in our sample.Average PM 2.5 is calculated during the four weeks of the year tillage is most likely to occur.

Figure 2 :
Figure 2: Average share of no-till and PM 2.5 -We average the share of crop pixels in no-till and PM 2.5 by county across the initial years of our sample.Grey counties are those for which we have no data or are dropped because they contain high levels of urbanized area (the I80 corridor and area around Chicago for example).

Table 1 :
Impact of no-till management of fields on PM 2.5

Table 2 :
Summary statistics