SUMMARY OF SITE-SPECIFIC SAMPLING
D.W. Franzen and V.L. Hofman
NDSU Extension Soil Specialist
Extension Ag and Biosystems Engineering
In 1994, commercial sampling for variable-rate fertilizer application began in the Red River Valley for sugarbeet production. Variable-rate P, K and limestone application in the central Corn Belt was directed by grid sampling, a systematic method designed to eliminate sampling bias. Field nutrient patterns are formed from analysis of the grid samples. The density of grid sampling influences how closely field nutrient maps reflect actual levels.
Research in Illinois and Wisconsin suggested that grids be about one sample per acre ( a 200-220 foot grid), which is reasonable in these areas considering sampling is conducted only once every four years with a hand probe and an ATV or unmodified pickup truck. However, sampling for NO3-N to 4 feet in depth for sugarbeet requires much more time and much higher equipment costs than surface P, K and pH sampling. Sampling must also occur each year for N responsive crops (sugarbeet, wheat, barley, canola, sunflower, corn, potatoes).
Based on economics alone, samplers in the Red River Valley began to sample in 5 acre grids. At that time, samplers were told that this method would reveal some of the nutrient variability within fields, but the grids were not tight enough to consistently show actual field patterns to direct variable-rate application long-term. Samplers were asked to be flexible enough that following sufficient sampling research, adjustments may have to be made in sampling strategies. Based on information from these studies, many samplers are already modifying their earlier sampling methods.
Sites were established near Gardner from 1994-96, Colfax from 1995-98, east of Hunter from 1997-98, and sister sites were located outside the Red River Valley near Valley City and Mandan. All sites were densely gridded, with cores to 4 feet taken when possible in the Valley locations and to 2 feet outside the Valley, or when the subsoil was so saturated with water that deep sampling was impossible (Hunter site, 1998).
Grid samples were collected in a 110 foot grid (1/4 acre), which helped reveal underlying fertility patterns within the fields for comparisons with less dense sampling methods. Grain yields were collected at Colfax, to determine profitability in various crops such as corn and wheat from variable-rate application of N. At many of the sites, additional samples were collected in 10 foot and 2 foot grids to determine how many sample cores might be needed to represent each grid or zone. The methods for comparison of grids and zones have been given in previous Sugarbeet Research and Extension Reports from 1995-1997. Topography was measured in each field using a laser-surveying device with recording taken in a 110 foot grid.
The Colfax site as described by the Richland county soil survey consists of Tiffany/Embden and Glyndon/Wyndmere soils, with the Glyndon/Wyndmere soils in the east 2/3 and Tiffany/Embden in the west.
Figure 1. Soil series as described by soil survey
Figure 2. Colfax site as mapped by M.D. Sweeney, 1998
The Colfax site as described in a detailed soil survey conducted in 1998 consists of a natural Tiffany drainage way bordering the east 1/3 of the field in a north-south direction. This channel was widened to allow surface drainage of water just prior to the 1998 sampling. Some grading has been conducted in the past prior to 1995 in the same area. To the east of the drainage way are Wyndmere and Hecla soils. To the west are Wyndmere and soils, with some disturbed Maddock along the southwest corner.
Changes in NO3-N over time
The NO3-N patterns changed little between 1995 and 1998. The Tiffany soils shown in Figure 1 were consistently low each fall, probably as a result of denitrification from excessive water during each summer. The Hecla soils and parts of the Wyndmere soils were higher in N.
Figure 3. Colfax NO3-N levels, 0-2 feet, 1995-1998
In 1995, the field was in corn, in 1996 it was seeded to spring wheat, with soybeans planted late into the Tiffany area because of early season wetness. In 1997, the field was seeded to corn, and in 1998, the field was planted to soybeans. As each crop was harvested, similar patterns of yield were revealed (Figure 4).
Figure 4. Wheat and corn yields, 1996 and 1997
The patterns of yield and nutrients corresponded to patterns of topography
Figure 5. NO3-N, P and corn yield over topography
Grid and topography comparisons with 110 foot grid values
Area-based topography sampling represented 110 foot grid values similar to the 220 foot grid in most years. Relatively low correlations of chloride would not cause concern, considering that chloride ranged from 100 to 400 lb/acre in all study years. The recommendation would be the same whether the value sampled was 100 or 400 lb/acre; no additional chloride would be needed. Area-based topography correlation with P was probably low in 1995 because the samples were obtained with an automatic probe under wet conditions, which may have degraded the quality of the 0-6 inch sampling depth.
Table 1. Comparison of topography-based and grid sampling strategies, Colfax, 1995-1998
The 220 foot grid tended to have a better correlation than area-based topography sampling, however, it would take 36 samples to construct a map on the 40 acre field, compared to 5 samples for the area-based topography sampling. The tendency for N correlation to decrease with time is probably a result of variable-rate N applications in 1996 and 1997. In 1996, there was a significant increase in net return due to topography-based variable N and S application, compared to both a uniform N application and grid sampling at a 440 ft. grid. In 1997, there were no significant differences between treatments, partially due to treatment overlap through a misapplication of the fertilizer on the west side of the farm, and partially due to a generally low range in N values. Most sample N values were within 40 lb/acre of each other.
The Gardner site was a 40 acre farm split into a north 15 acre field and a south 25 acre field. The north 15 acres was in alfalfa from 1994-1996. During the winter of 1994-95, some of the alfalfa suffered winter kill due to ice buildup. The stand continued to decline until after the 1996, it was plowed under. The south field was in spring wheat in 1994, barley in 1995, and was seeded to alfalfa in 1996. Although the soil survey from Cass county describes the field as one soil mapping unit, Fargo/Hegne, a detailed soil survey conducted during 1998 shows more interesting features (Figure 6).
Figure 6. A detailed soil survey of the Gardner site. M.D. Sweeney, 1998
The Hegne soil tends to be a little higher in elevation than the Fargo, and the Enloe areas are particularly poorly drained. The line of Enloe soils in the south field tend to form a natural drain together from west to east, before reaching the man-made drain in the east. A road borders the west boundary.
Change in nutrients over time
Patterns of NO3-N changed over time. N levels were spatially variable in the south field in 1994, but not the north field where the vigorous alfalfa was growing. In 1995, both fields were spatially variable, and N levels were more related to landscape than any other year studied.
Figure 7. NO3-N levels, Gardner, 1994-1996
Figure 8. P levels, Gardner, 1994-1996
In 1996, following the new alfalfa seeding, and the continued decline of the alfalfa stand in the north field, there was no spatial variability. This phenomena is very similar to that following sugarbeet at the Hunter site from 1997-1998, and also has been noted following sunflower at the Mandan site-specific farms from 1995-1998.
When N levels were spatially variable, high levels tended to be in the upland positions in the south field, while high levels tended to be in the depressions in the alfalfa. In the south fields, the depressions tended to be wet, but in the alfalfa field, even following heavy rains, it was possible to drive in the field within 24 hours, indicating soil moisture levels were relatively low. Denitrification was probably not proceeding at as high a rate as in the south, annually cropped field.
The P levels varied between years. P levels were lower in 1995 than in 1994, then recovered again in 1996. Patterns within fields were relatively similar, but the values changed. Canadian researchers have also noted that during alfalfa degradation, P levels decrease. It may be the case that our extracting solutions do not extract organic P very well.
Table 2. Comparison of topography-based and grid sampling strategies, Gardner, 1994-96
Area-based topography sampling was based on five sampling zones, compared to 36 samples for the 220 ft. grid. Topography sampling, both area and point compared well to the 220 ft. grid.
The Gardner site was important not only because it supported topography-based zone sampling to gather nutrient information meaningfully and cheaply, but also pointed out the impact that certain crops may have in altering the spatial patterns of nutrients within fields. A combination of soil sampling in the winter-killed areas, combined with previously crop credits from the remaining areas of alfalfa would have been necessary to describe the effects of the crop on subsequent crop N requirements.
The Hunter site is about 5 miles west of Gardner, and consists of Bearden and Perella soils. Because of the wet weather in the fall of 1998, the field has not been subjected to a detailed soil survey, but a survey is planned for spring or fall, whichever is most practical. The wet weather also prevented the 2-4 foot depth sampling this fall. Only the 0-6 inch and 6-24 inch depths were taken. The field was in spring wheat in 1997 and sugarbeet in 1998. The N level was spatially variable following spring wheat, but was not spatially variable following sugarbeet.
The field slopes gently from west to east, with a few locally upland positions near the field center. In 1997, N levels were highest along the east slopes of the central upland positions and lowest near the center and in the southeast where water tends to stand (Figure 8).
Figure 8. NO3-N levels to 4 feet draped over topography, 1997
The pattern of NO3-N completely changed following sugarbeet (Figure 9). Instead of high areas of N in the central uplands, the high N remained on the south edge of the field, but levels in other areas were low and uniform.
The difference in spatial variability between NO3-N levels following spring wheat and sugarbeet is reflected in the correlation between sampling strategies and the 110 foot grid. There was low correlation between the 110 foot N levels and sampling strategies in 1998, compared to higher correlation values in 1997 (Table 3). This is the same problem observed at the Gardner site when the field was in a vigorous stand of alfalfa. Sugarbeet and alfalfa are accumulating large amounts of soil N, and soil N levels alone may not be sufficient to reveal the true N status of the field. Following alfalfa it has been common to use a previous crop credit to reduce N recommendations below that directed solely by soil test. Based on work by Moraghan and an ongoing study in Pembina county in a sugarbeet/potato rotation, the level of N in sugarbeet tops is a much better predictor of field N status than soil testing alone.
Figure 9. NO3-N levels draped over topography, 1998
Table 3. Correlation of grid and topography sampling strategies, Hunter, 1997-98
One of the problems with topography sampling alone in determining management zones is determining boundaries on slopes. Satellite images of the Hunter site will be made available this winter to further define boundaries beyond estimates made from topography data. When this is done, any adjustments in zone sampling compared to grids will be reevaluated using both the 1997 and 1998 data.
This study has shown that the relatively flat soils of the Red River Valley can be extremely variable in nutrient levels. The variability may be associated with landscape and/or the activities of man. If sampling by grid, the 220 foot grid established in the central Corn Belt as the recommended grid strategy is valid here as well. However, topography-based zone sampling was also shown to be as effective as the 220 foot grid in representing field values and often superior to the 220 foot grid in fertility boundary definition. A 4-5 acre grid is a generally poor method of directing variable-rate fertilizer application, useful mostly in revealing some range of variability to a grower new to site-specific farming. The topography-based zones require fewer samples and lower sampling and analysis costs than grid, while providing a high level of quality information to the producer. Sampling by zone should use several sources of information, including aerial images of crops and bare soil, satellite imagery, topography, yield maps, electrical conductivity detection, and detailed soil survey information (not published survey information).