Jiyul Chang

Factors influencing spatial variability of soil apparent electrical conductivity

ABSTRACT Soil apparent electrical conductivity (ECa) can be used as a precision farming diagnostic tool more efficiently if the factors influencing ECa spatial variability are understood. The objective of this study was to ascertain the causes of ECa spatial variability in soils developed in an environment with between 50 and 65 cm of annual rainfall. Soils at the research sites were formed on calcareous glacial till parent materials deposited approximately 10,000 years ago. Soil samples (0–15 cm) collected from at least a 60 by 60 m grid in four fields were analyzed for Olsen phosphorus (P) and potassium (K). Elevation was measured by a carrier phase single frequency DGPS and ECa was measured with an EM 38 (Geonics Ltd., ON, Canada) multiple times between 1995 and 1999. Apparent electrical conductivity contained spatial structure in all fields. Generally, the well drained soils in the summit areas and the poorly drained soil in the toeslope areas had low and high ECa values, respectively. The landscape differences in ECa were attributed to: (i) water leaching salts out of summit areas and capillary flow combined with seepage transporting water and salts from subsurface to surface soils in toeslope areas; (ii) lower water contents in summit than toeslope soils; and (iii) water erosion which transported surface soil from summit/shoulder areas to lower backslope/footslope areas. A conceptual model based on these findings was developed. In this model, topography followed a sine curve and ECa followed a cosine curve. Field areas that did not fit the conceptual model were: (i) areas containing old animal confinement areas; (ii) areas where high manure rates had been applied; and (iii) areas where soils were outside the boundary conditions of the model, i.e., soils not developed under relatively low rainfall conditions in calcareous glacial till with temperatures ranging between mesic and frigid. This research showed that the soil forming processes as well as agricultural management influenced ECa and that by understanding how landscape position influences salt loss and accumulation, water redistributions following precipitation, and erosion areas that do not fit the conceptual model can be identified. This information can be used to improve soil sampling strategies.

Precision Farming Protocols: Part 1. Grid Distance and Soil Nutrient Impact on the Reproducibility of Spatial Variability Measurements

To determine temporal changes in soil nutrient status, reproducible results must be obtained at each time step. The objective of this paper was to determine the impact of grid distance on the reproducibility of spatial variability measurements. Soil samples from the 0 to 15 cm depth were collected from a 30 by 30 m grid in May 1995 in a 65 ha notill corn (Zea mays L.) field. Each bulk sample contained 15 individual cores, collected at sample points located every 11.4 cm along a transect that transversed 3 corn rows (57 cm). At each sampling point latitude, longitude, elevation, landscape position, and soil series were determined. The 30 m grid was used to develop 4 and 9 independent data sets having a 60 and 90 m, grids, respectively. Semivariograms, nugget to sill ratios, and mean squared errors were calculated for each data set. At 60 m: (i) the total N, total C, and pH semivariograms, of different start points, were similar, while semivariograms for Olsen P, K, and Zn were different; (ii) the spatial dependence ratings, based on the nugget to sill ratio, for total N, total C, and pH semivariograms were consistent and suggested moderate spatial dependence; (iii) the spatial dependence rating for Olsen P, K, and Zn for the 4 semivariograms were not consistent and ranged from weak to moderate spatial dependence. At 90 m all soil nutrients had different semivariograms for each start point, while the spatial dependence rating for each total N, total C, and pH start point were consistent and showed moderate spatial dependence. The total C, P, K, Zn, and pH MSE values at 60 m, were 30, 30, 41, 28, and 72% lower than the variance, respectively. This study showed that semivariogram, semivariance, MSE, and nugget to sill ratios reproducibility were dependent on soil nutrient and grid distance.

Clouds Influence Precision and Accuracy of Ground‐Based Spectroradiometers

Abstract The objectives of this study were to determine the precision and accuracy, under field conditions, of two commonly used ground‐based spectroradiometers and to propose guidance on how to minimize system errors. Sunlight irradiance and reflected radiance were measured on calibration tarps (3.6% and 52% reflectance) on 6 days using a CropScan MSR 16 handheld multispectral radiometer and a Fieldspec model FR hyperspectral radiometer during 2002. Radiance and irradiance were corrected for temperature and sun angle and converted to percent reflectance. Analysis showed that variances of the reflectance values for both radiometers increased with cloud cover. These results were attributed to several factors. First, cloud cover produced atmospheric conditions that made irradiance highly variable. Under these conditions, if reflected light is calculated by dividing radiance from target by radiance from a known standard, which is only periodically measured, then the calculated reflectance value may contain errors. Second, the reduction of diffuse irradiance by increasing cloud cover may introduce errors into reflectance calibration. Third, the relationship between incident irradiance, reflection of surface, and sensor efficiency may not be linear, and therefore, calculated reflectance can be variable when incident irradiance is variable. Results from this study showed that 1) the field measurements must be conducted under similar conditions at a similar time, 2) both sensors must be calibrated before and after measurements with reference panel, with ample time for device warm up, 3) measured reflectance should be corrected with reflectance from a reference panel, and 4) for the FieldSpec, reflectance measurements can be improved by simultaneously measuring radiance from the target and a known standard.

Detecting weed-free and weed-infested areas of a soybean field using near-infrared spectral data

Abstract Weed distribution maps can be developed from remotely sensed reflectance data if collected at appropriate times during the growing season. The research objectives were to determine if and when reflectance could be used to distinguish between weed-free and weed-infested (mixed species) areas in soybean and to determine the most useful wavebands to separate crop, weed, and soil reflectance differences. Treatments included no vegetation (bare soil), weed-free soybean, and weed-infested soybean and, in 1 yr, 80% corn residue cover. Reflectance was measured at several sampling times from May through September in 2001 and 2002 using a handheld multispectral radiometer equipped with band-limited optical interference filters (460 to 1,650 nm). The spatial resolution was 0.8 m2. The reflectance in the visible spectral range (460 to 700 nm) generally was similar among treatments. In the near-infrared (NIR) range (> 700 to 1,650 nm), differences among treatments were observed from soybean growth stage V-3 (about 4 wk after planting) until mid-July to early August depending on crop vigor and canopy closure (76-cm row spacing in 2001 and 19-cm row spacing in 2002). Reflectance rankings in the NIR range when treatments could be differentiated were consistent between years and, from lowest to highest reflectance, were soil < weed-free < weed-infested areas. Increased reflectance from weed-infested areas was most likely due to increased biomass and canopy cover. Residue masked differences between weed-free and weed-infested areas during the early stages of growth due to high reflectance from the residue and reduced weed numbers in these areas. These results suggest that NIR spectral reflectance collected before canopy closure can be used to distinguish weed-infested from weed-free areas. Nomenclature: Soybean, Glycine max (L.) Merr.

Field scale variability of nitrogen and δ15N in soil and plants

Abstract Understanding the factors that influence soil and plant nitrogen (N) spatial variability may improve our ability to develop management systems that maximize productivity and minimize environmental hazards. The objective of this study was to determine the field (65 ha) scale spatial variability of N and δ15N in soil and corn (Zea mays). Soil, grain, and stover samples were collected from grids that ranged in size from 30 by 30 m to 60 by 60 m. Plant samples, collected following physiological maturity in 1995, were analyzed for total N and δ15N. Soil samples, collected prior to planting in the spring of 1995 and 1996, were analyzed for inorganic‐N, total N, and δ15N. All parameters showed strong spatial relationships. In an undrained portion of the field containing somewhat poorly and poorly drained soils there was a net loss of 95 kg N ha‐1, while in an adjacent area that was tile drained there was a net gain of 98 kg N ha‐1. Denitrification and N mineralization most likely were responsible for lo...

Precision farming protocols. part 2. comparison of sampling approaches for precision phosphorus management

Abstract Research is needed to compare the different techniques for developing site‐specific phosphorus (P) recommendations on a field‐wide basis. The objective of this study was to determine the impact different techniques for developing site‐specific P recommendation maps on yield and profitability. Enterprise analysis combined with a crop simulation model and detailed field characterization was used to estimate the value of spatial P information in a system where N was not limiting. The systems evaluated were continuous corn (Zea mays) and corn and soybean (Gfycine max) rotations where sampling and fertilizer applications were applied annually and semi‐annually, respectively. The sampling techniques tested were: (i) an unfertilized P control; (ii) whole field; (iii) whole field plus historic information (feedlot); (iv) landscape positions; (v) soil type; (vi) soil type plus historic information (feedlot); and (vii) 90‐m grid sampling. The finding of this study were based on soil samples collected from a 30 by 30‐m grid. The value of the spatial information was dependent on the crops response to P, the accuracy of the different sampling techniques, crop rotation, and the length of time between sampling dates. All of the sampling techniques produced different application maps. The recommendation map based on a single composite sample under fertilized 56.5% of the field. Increasing the sampling density reduced the percentage of under‐fertilized land. If corn had a low P response, then simulation/enterprise analysis indicated that applying P did not increased profits. For all scenarios tested: (i) the soil type + historic sampling approach had higher potential profits than the 90 m grid sampling approach; and (ii) there was no economic benefit associated with the 90‐m grid sampling. However, if research shows that amortization of sampling and analysis costs over 3 or 4 years is appropriate, then it may be possible to derive economic benefit from a 90‐m grid sampling. For a corn/soybean rotation, where fertilizer was applied when corn was planted and N and P was not applied to soybeans, enterprise/ simulation analysis (2.8 Mg ha‐1 soybean yield goal and a moderate P model) showed that soil + historic sampling approach increased profitability

Evaluating Modified Atmospheric Correction Methods for Landsat Imagery : Image-Based and Model-Based Calibration Methods

3.74 ha‐1 when compared to the uniform P treatment.

Estimating Annual Root Decomposition in Grassland Systems

Abstract To increase the accuracy of remotely sensed data for agricultural forecasting, pixel values must be corrected for atmospheric effects and converted to spectral reflectance. The objective of this research was to compare two atmospheric correction methods of Landsat imagery under a range of atmospheric conditions. Ground‐based dark‐object subtraction (GDOS) is an image‐based calibration method that used in situ ground data that the dark‐object subtraction (DOS) method did not use, whereas atmospheric calibration (AC) is a model‐based calibration method that required a standard atmospheric profile refined with the use of in situ atmospheric data. GDOS and AC methods improved the reflectance values and had relationships with measured bands, which were approximately 1 to 1 in all bands. However, the GDOS generally had lower root‐mean‐square errors (RMSE) than AC. Data from this study suggest that at the present time the GDOS method may be more accurate than the AC method.

Corn and Soybean Mapping in the United States Using MODIS Time-Series Data Sets

ABSTRACT Calculated belowground buried root bag decomposition rates may be impacted by soil disturbance and that mesh bags can exclude potential degraders. This paper explicitly compares the sequential soil sampling method to the buried root bag approach to determine if biomass degradation estimates over a season differ. The research was conducted at two eastern South Dakota grassland sites (loamy and thin upland ecological sites) in 2011 and 2012 in an area where the grassland vegetation was killed to prevent new root growth. In the sequential core technique, a composite sample consisting of three 4-cm diameter soil cores from the 0- to 15- and 15- to 30-cm depth were collected monthly from May to October, whereas five residue bags were placed 7 cm below the soil surface in spring and removed at the last soil sampling date. The sequential core (61% ± 7.2) and residue bag (58% ± 7.2) techniques had similar root decomposition amounts; however, the sequential core technique had a lower labor requirement than the residue bag technique when the increased sampling requirement was considered.