David J. Mulla

Estimation of soil properties and wheat yields on complex eroded hills using geostatistics and thematic mapper images

Abstract Better assessment of large scale patterns in soil fertility, erosion, and crop productivity is needed to address environmental concerns in the steeply rolling dryland farming region of eastern Washington. Spatial variability of organic carbon, soil phosphorus, and wheat yields measured on several 655 m long transects in complex eroded hills of the Palouse region of eastern Washington was studied using classical statistics and geostatistics. In addition, Landsat Thematic Mapper images of bare soil at the study site were used to estimate organic matter content in blocks having an area of 28.5m × 28.5 m. Classical statistical analysis showed moderate coefficients of variation in soil properties (25–50%) and wheat yields (30%). Spherical semivariograms of soil properties and wheat yields showed strong spatial dependence, with a range of influence on the order of from 70 m to 145 m. The semivariogram for Landsat estimated organic matter was almost identical to that for organic matter measured in the surface transects. The cross-semivariogram between measured organic matter and wheat yields also exhibited strong spatial dependence, with a range of influence of about 120m. Organic matter, soil phosphorus, and wheat yield were interpolated at unsampled locations using block kriging and block cokriging techniques. Excellent agreement was obtained between kriged wheat yields and wheat yields cokriged from Thematic Mapper estimates of organic matter. Satisfactory agreement was obtained between kriged soil phosphorus and cokriged soil phosphorus based upon remotely sensed organic matter. Results of this study provided strong evidence for nonrandom spatial patterns in soil properties and crop yields. These spatial patterns are associated with changes in surface organic matter content across the landscape resulting from extensive erosion and exposure of subsoil. Geostatistical techniques provide a powerful tool for interpolating ground measurements of soil properties and crop yield over large areas in combination with estimates of organic matter obtained from Thematic Mapper images.

A comparison of winter wheat yield and quality under uniform versus spatially variable fertilizer management

Abstract The Palouse region of eastern Washington, USA is characterized by steep, highly erodible hills with complex patterns in soil fertility and crop productivity. Because of these patterns, there is increasing interest in applying variable, rather than uniform rates of fertilizer across the landscape. This study was conducted to determine the yield and quality of wheat at several different positions on landscapes managed with uniform vs. variable rates of nitrogen and phosphorus fertilizer. A moderately to severely eroded wheat farm in summer fallow management near St. John, Washington was intensively sampled at intervals of 15 m along several transects 655 m long and separated by 122 m. Each transect crosses a wide range of landscape position, and a variety of soil fertility and crop productivity regimes. Maximum slope steepness along the transects reached 45%, although flat portions also existed. Along the south side of each transect, fertilizer was applied at a uniform rate according to the growers typical practice (73 kg N ha −1 and 6 kg P ha −1 ). Along the north side, the transects were separated into three management zones that received different rates of nitrogen and phosphorus fertilizer applications to match the existing soil fertility and crop productivity patterns. The three variably fertilized management zones received 22 kg N ha −1 and 6 kg P ha −1 (Zone 1), 90 kg N ha −1 and 6 kg P ha −1 (Zone 2), and 22 kg N ha −1 and 18 kg P ha −1 (Zone 3). Yield and quality of wheat were measured in both the uniform and variable management strips at intervals of 15 m. In spite of applying 51 kg N ha −1 less in management Zones 1 and 3 than in the adjacent uniformly fertilized strip, there were no significant differences in yield of wheat in any management zone between the variably and the uniformly fertilized strips. Grain yield was significantly different between the three fertilizer management zones, with measured yields of 4097 kg ha −1 (Zone 1), 4501 kg ha −1 (Zone 2), and 3339 kg ha −1 (Zone 3). These differences are largely due to differences between each zone in available water content and nitrate nitrogen in the profile. Protein content of wheat was significantly lower in variably fertilized strips than in uniformly treated strips, leading to significant improvements in grain quality. The parts of the landscape with the lowest available water content in the profile had the highest grain protein content due to water stress-nitrogen interactions.

Combining chlorophyll meter readings and high spatial resolution remote sensing images for in-season site-specific nitrogen management of corn

The chlorophyll meter (CM) has been commonly used for in-season nitrogen (N) management of corn (Zea mays L.). Nevertheless, it has limited potential for site-specific N management in large fields due to difficulties in using it to generate N status maps. The objective of this study was to determine how well CM readings can be estimated using aerial hyper-spectral and simulated multi-spectral remote sensing images at different corn growth stages. Two field experiments were conducted in Minnesota, USA during 2005 involving different N application rates and timings on a corn-soybean [Glycine max (L.) Merr.] rotation field and a corn-corn rotation field. Four flights were made during the growing season using the AISA Eagle Hyper-spectral Imager and CM readings were collected at four or five different growth stages. The results indicated that single multi-spectral and hyper-spectral band or vegetation index could explain 64–86% and 73–88% of the variability in CM readings, respectively, except at growth stage V9 in the corn-soybean rotation field where no band or vegetation index could explain more than 37% of the variability in CM readings. Multiple regression analysis demonstrated that the combination of 2–4 broad-bands or 3–8 narrow-bands could explain 41–92% or 61–94% of the variability in CM readings across the two fields and different corn growth stages investigated. It was concluded that the combination of CM readings with high spatial resolution hyper-spectral or multi-spectral remote sensing images can overcome the limitations of using them individually, thus offering a practical solution to N deficiency detection and possibly in-season site-specific N management in large continuous corn fields or at later stages in corn-soybean rotation fields.

Identifying important factors influencing corn yield and grain quality variability using artificial neural networks

Soil, landscape and hybrid factors are known to influence yield and quality of corn (Zea mays L.). This study employed artificial neural network (ANN) analysis to evaluate the relative importance of selected soil, landscape and seed hybrid factors on yield and grain quality in two Illinois, USA fields. About 7 to 13 important factors were identified that could explain from 61% to 99% of the observed yield or quality variability in the study site-years. Hybrid was found to be the most important factor overall for quality in both fields, and for yield as well in Field 1. The relative importance of soil and landscape factors for corn yield and quality and their relationships differed by hybrid and field. Cation exchange capacity (CEC) and relative elevation were consistently identified as among the top four most important soil and landscape factors for both corn yield and quality in both fields in 2000. Aspect and Zn were among the top five most important factors in Fields 1 and 2, respectively. Compound topographic index (CTI), profile curvature and tangential curvature were, in general, not important in the study site-years. The response curves generated by the ANN models were more informative than simple correlation coefficients or coefficients in multiple regression equations. We conclude that hybrid was more important than soil and landscape factors for consideration in precision crop management, especially when grain quality was a management objective.

Sensor planning for a symbiotic UAV and UGV system for precision agriculture

We study the problem of coordinating an Unmanned Aerial Vehicle (UAV) and an Unmanned Ground Vehicle (UGV) for a precision agriculture application. In this application, the ground and aerial measurements are used for estimating nitrogen (N) levels on-demand across a farm. Our goal is to estimate the N map over a field and classify each point based on N deficiency levels. These estimates in turn guide fertilizer application. Applying the right amount of fertilizer at the right time can drastically reduce fertilizer usage. Towards building such a system, this paper makes the following contributions: First, we present a method to identify points whose probability of being misclassified is above a threshold. Second, we study the problem of maximizing the number of such points visited by an UAV subject to its energy budget. The novelty of our formulation is the capability of the UGV to mule the UAV to deployment points. This allows the system to conserve the short battery life of a typical UAV. Third, we introduce a new path planning problem in which the UGV must take a measurement within a disk centered at each point visited by the UAV. The goal is to minimize the total time spent in traveling and measuring. For both problems, we present constant-factor approximation algorithms. Finally, we demonstrate the utility of our system with simulations which use manually collected soil measurements from the field.

Water quality modeling of fertilizer management impacts on nitrate losses in tile drains at the field scale

Nitrate losses from subsurface tile drained row cropland in the Upper Midwest U.S. contribute to hypoxia in the Gulf of Mexico. Strategies are needed to reduce nitrate losses to the Mississippi River. This paper evaluates the effect of fertilizer rate and timing on nitrate losses in two (East and West) commercial row crop fields located in south-central Minnesota. The Agricultural Drainage and Pesticide Transport (ADAPT) model was calibrated and validated for monthly subsurface tile drain flow and nitrate losses for a period of 1999-2003. Good agreement was found between observed and predicted tile drain flow and nitrate losses during the calibration period, with Nash-Sutcliffe modeling efficiencies of 0.75 and 0.56, respectively. Better agreements were observed for the validation period. The calibrated model was then used to evaluate the effects of rate and timing of fertilizer application on nitrate losses with a 50-yr climatic record (1954-2003). Significant reductions in nitrate losses were predicted by reducing fertilizer application rates and changing timing. A 13% reduction in nitrate losses was predicted when fall fertilizer application rate was reduced from 180 to 123 kg/ha. A further 9% reduction in nitrate losses can be achieved when switching from fall to spring application. Larger reductions in nitrate losses would require changes in fertilizer rate and timing, as well as other practices such as changing tile drain spacings and/or depths, fall cover cropping, or conversion of crop land to pasture.

Management of diffuse pollution in agricultural watersheds : Lessons from the Minnesota river basin

The Minnesota River (Minnesota, USA) receives large non-point source pollutant loads. Complex interactions between agricultural, state agency, environmental groups, and issues of scale make watershed management difficult. Subdividing the basins 12 major watersheds into agro-ecoregions based on soil type, geology, steepness, and climate enhances predictability of stream water quality parameters. An eight-step framework for agricultural watershed management is presented.

Spatially variable liming rates: a method for determination

The variability of soil acidity and crop liming requirements in a field in East Texas were examined. Sixty-eight soil samples were taken in a systematic manner from the field. Geostatistical techniques were used to analyze the soil acidity variability of the samples and assist in developing a liming application rate map for the field. Soil pH, soil texture, and buffer pH variations showed spatial dependence. Application of the average recommendation rate for the field would have resulted in an overapplication of lime in 9 to 12% of the field and an underapplication on 37 to 41% of the field. Varying lime application within the field so different areas receive appropriate rates would have caused a greater total lime application of 8 to 28%, depending on recommendation method, compared to the mean application rate. The data indicated the application of lime “where needed” could maximize application benefits.

Identifying critical agricultural areas with three-meter LiDAR elevation data for precision conservation

Determining which portions of agricultural landscapes are major sources of pollution within a watershed is time consuming and labor intensive. Small critical areas of the landscape contribute disproportionate amounts of sediment and phosphorus to nearby waterways. Critical areas are defined here as areas of accumulated overland runoff that are hydrologically connected to surface waters. With advancements in light detection and ranging (LiDAR) technologies, landscape topography can be represented with highly accurate terrain data. The objective of this study is to determine the effectiveness of using LiDAR–based terrain attributes to identify fine-scale critical areas in selected Minnesota watersheds and to analyze cost efficiency of this type of analysis. The LiDAR digital elevation model data were acquired for two south central Minnesota watersheds, and the terrain attributes slope, flow accumulation, and stream power index were calculated with a 3 m (9.8 ft) spatial resolution. Field surveys were conducted in these watersheds along the riparian corridor to identify side inlets and active gullies that contribute to surface water quality degradation. Terrain attributes were able to identify 80% of field-verified gullies in the study watersheds. Furthermore, an even higher percentage of gullies with a high sediment delivery potential were identified using terrain attributes. Gully size was ranked during field surveys, and 31 of the 32 largest gullies ranked in the field were successfully identified with LiDAR–based terrain attributes. In contrast, only 7 of these gullies could be identified using 30 m (98 ft) digital elevation model terrain attributes. The LiDAR approach for identifying critical source areas using terrain attributes has a large potential for cost savings relative to time-consuming field surveys. With an ever-increasing availability of LiDAR data, terrain analysis may prove very useful in the future for targeting best management practices to critical areas for reductions in nonpoint source pollution.

MIXED TRANSFORM FINITE ELEMENT METHOD FOR SOLVING THE NON-LINEAR EQUATION FOR FLOW IN VARIABLY SATURATED POROUS MEDIA

A new computational method is developed for numerical solution of the Richards equation for flow in variably saturated porous media. The new method, referred to as the mixed transform finite element method, employs the mixed formulation of the Richards equation but expressed in terms of a partitioned transform. An iterative finite element algorithm is derived using a Newton–Galerkin weak statement. Specific advantages of the new method are demonstrated with applications to a set of one— dimensional test problems. Comparisons with the modified Picard method show that the new method produces more robust solutions for a broad range of soil– moisture regimes, including flow in desiccated soils, in heterogeneous media and in layered soils with formation of perched water zones. In addition, the mixed transform finite element method is shown to converge faster than the modified Picard method in a number of cases and to accurately represent pressure head and moisture content profiles with very steep fronts.