1Department of Soil Science, University of Saskatchewan, Saskatoon, SK,
2Research and Development, Saskatchewan Wheat Pool, Saskatoon, SK
In recent years, a number of technological advances including the development of Global Positioning Systems (GPS), variable rate delivery carts for fertilizer and seed, and yield monitors have resulted in a growing interest in using these technologies for precision farming in Saskatchewan. Unfortunately, however, the technological advances have proceeded more rapidly than the agronomic advances, which has limited the potential for application of precision farming techniques. Clearly, the ability to vary inputs precisely is of little value if there is no agronomically meaningful information available to assist producers in making decisions regarding how, when and where to vary these inputs. Finally, irrespective of whether or not any precision farming technique is agronomically meaningful, to be a viable alternative management practice, implementing precision farming techniques must be economically viable.
A collaborative study between the Department of Soil Science and the Saskatchewan Wheat Pool (and funded by ADF) was initiated in 1996 to investigate the agronomic and economic feasibility of variable rate fertilizer application for wheat and canola. Results from this study provide the basis for the following discussion.
On the rolling plains of the western Canadian prairies, crop productivity can be highly variable within a field. The most pronounced variability usually occurs on hummocky or rolling terrain where crop yields are often relatively low on upper slope positions (i.e., knolls) as compared to lower slope positions (i.e., depressions). The differences in productivity are well recognized by producers and researchers alike and are generally attributed to loss and redistribution of topsoil (Gregorich and Anderson 1985; Pennock and de Jong 1990; McConkey et al. 1997) and moisture (Verity and Anderson 1990) from knolls to depressions. The challenge for implementing precision farming is to determine the specific factors that are responsible for productivity differences within a field and develop appropriate management strategies.
With few exceptions, virtually all of the agricultural land in Saskatchewan was influenced by the last great glacier that began to retreat some 14,000 - 17,000 years ago. In some areas, the retreating glacier resulted in the formation of large glacial lakes. These landscapes typically are relatively flat and have heavy textured soils (i.e., clays) with few, if any stones. For example, the Regina heavy clays and much of the relatively flat area in the Rosetown region developed on old glacial lake bottoms. These deposits are known as "lacustrine" deposits. In addition, there are many areas of the province that are dominated by "glacial till" deposits. These deposits consist of the unsorted mineral material that had been ground up within the glacier and was left behind as the glacier melted. Landscapes developing on these glacial till deposits typically have a series of knolls and depressions (including sloughs) in any given field. Soils developed on glacial till deposits are easily identified by the scattered stones on the soil surface. As much as 38% of the agricultural land in Saskatchewan developed on glacial till.
In Saskatchewan, soil formation began after the great ice sheets retreated to the north. Most of the soil forming processes which have influenced the landscapes since that time are controlled by the amount of water present in any given point in the landscape. The pattern of water movement in a landscape is, in turn, controlled by slope (Pennock et al. 1994).
Because the soil-forming processes are closely linked to water movement, a clear pattern of soils occurs in many landscapes (Fig. 1). Water is shed from knolls or upper slope positions and drier conditions occur, limiting the amount of organic matter production that in upper slopes of both undisturbed and agricultural fields. The result has been the development of relatively shallow soils that are easily eroded by wind, water and tillage. As a result, the knolls often have a grayish appearance, indicating the presence of lime in the remaining topsoil.
Soils in midslope positions show a consistent change in their properties with distance downslope. These soils typically have lime-free topsoil, which increases in thickness downslope to the edge of the lower slope positions. Soils with a characteristically thick layer of topsoil are referred to a "Chernozems" by soil scientists.
Lower slope soils are characterized by the greatest variations between and within fields due to the influence of the groundwater. Where the groundwater table is far below the surface, lower slopes receive - and benefit from - snowmelt and runoff water from upper slopes, resulting in the development of thick, dark, topsoil. Alternatively, the water from upper slope positions can periodically pond on the surface in the depressions. Ponded water displaces any oxygen held in the soil pores and a set of soil forming processes occur that are typical of oxygen depleted environments. Soils influenced by periodic ponding can develop a greyish or mottled appearance and are referred to as "Gleysols" by soil scientists. Where the groundwater table is closer to the surface, drainage may be limited and, depending on the parent material, movement of groundwater to the soil surface may result in the build-up of salinity. Hence, in a given field, the lower slope areas may have soils that are well developed and very productive, or the soils may be saline and relatively unproductive
Predictable soil variability may translate into predictable differences in both nutrient and moisture availability and, consequently, productivity potential. Understanding these relationships is a critical first step in developing appropriate precision farming strategies. Our experience has shown that the relationship to moisture is clear - soil conditions are driest on the knolls, increase through the midslope areas and are wettest in the lower slope positions. These differences are, for the most part, unmanageable at a practical level. (Arguably, reduced tillage that promotes good residue cover may minimize moisture variations by trapping snow in the winter and facilitating infiltration during precipitation events). In contrast, fertility variations can be managed and thus are the focus of much of the current precision farming technology.
Nitrogen levels are known to differ in fields and, at least in the broadest sense, these differences are well understood. The largest pool of N in the soil occurs in the organic matter. Breakdown, or "mineralization" of organic N depends on a number of factors, but a recent summary suggests that between 5 and 15% of the organic N can be mineralized in a given year (Campbell et al. 1994). Hence, at the simplest level, the amount and quality of organic matter in the soil controls the potentially available N, and the moisture and temperature conditions in a given year will determine the fraction of the potentially available N that is mineralized.
Clearly, a relationship exists between fertility and soil development. Soil organic matter increases from the knolls through the midslopes and will usually be at its highest in the lower slope positions. Hence the maximum amount of organic N available for mineralization also will increase from knolls to depressions. Temperature and moisture conditions for mineralization are usually optimum in the lower slope positions, and thus the rate of mineralization may be greatest in these areas. Overall, it is anticipated that the ability of the soil to supply N to the growing crop will be strongly related to slope position.
Three major challenges must be met in the development of variable rate fertilizer application strategies: 1) to correctly assess the variability of factors controlling productivity; 2) to identify this variability and map areas with similar characteristics; and 3) to develop crop response models to determine appropriate variable input rates (Sawyer 1994). In 1996, we initiated a study to measure the inherent fertility variations in typical Saskatchewan landscapes and to determine the different yield responses of wheat and canola to fertilizer rates in these landscapes.
The study site was located 40 km north of Saskatoon, SK, Canada near the community of Hepburn (SW 7-40-5-W3 or 52o25'N, 106o41'W) on a glacial till landscape (loam to clay loam). Soils at the site were dominantly Chernozemic with significant Gleysolic soils in the depressional areas. Slope gradients at the site range from 5 to 10%. Surface drainage at the site was local.
Two research experiments (i.e., wheat and canola) were established, each covering an area 250 m by 300 m (Fig. 1). The sites encompassed several cycles within the knoll and depression landscape - one cycle extending from one knoll down into a depression and then to the top of the next knoll. The sites were extensively surveyed using a Total Station and analysis of aerial photographs was used to develop maps comprising three management units: 1) upper slope; 2) midslope; and 3) lower-slopes (McCann et al., 1996). Soils from within each landscape position were sampled (15 cm increments to 60 cm) and were subjected to a series of analytical procedures to determine characteristics including mineral N, total carbon, organic carbon, pH, electrical conductivity and moisture content.
A series of replicated (6 replicates) fertility treatments were imposed across a complete landform cycle within the study site (Fig. 1). Nitrogen was applied as urea (46-0-0) at 5 rates (0, 0.5, 1.0, 1.5 and 2.0 times the recommended rate) and phosphorus was applied as monoammonium phosphate (11-55-0) at 3 rates (0, 1.0 and 2.0 times the recommended application rate for wheat and 0, 0.5 and 1.0 times the recommended rate for canola). For the purposes of this paper, only the N fertilizer treatment responses will be discussed. The recommended fertilizer rates were based on fall soil analysis (Enviro-Test Labs, Saskatoon). Fertilizer treatments were seeded using a modified Morris air seeder (2.13m width with 30 cm row spacing). Fertilizer N was side-banded whereas P was placed in the seed furrow. Treatments were harvested using a small plot harvester (10m sections were harvested within each of the management units).
Results from this study indicate that management units (i.e., upper, mid, and lower slope) developed on the basis of black and white aerial photographs were agronomically meaningful for both wheat and canola production, i.e., different management units represented areas within the field with measurably different soil properties. For example, soil tests revealed that plant available soil moisture in the spring was strongly related to the management units - the overall means increased from the upper slope through the mid slope to the lower slope positions (Fig. 2). It is interesting to note that although overall soil moisture declined over the three years of the study, the landscape pattern of moisture distribution persisted.
Figure 2. Available spring soil moisture at the Hepburn canola site.
Crop productivity similarly was strongly influenced by landscape position. For example, the mean yields on upper slope units at both the wheat and canola sites were consistently less than yields achieved on lower slope units, regardless of the rate of N fertilizer addition (Fig. 3 and 4).
Figure 3. Mean yield of canola (averaged over all N treatments).
Figure 4. Mean yield of wheat (averaged over all N treatments).
Interestingly, application of fertilizer N did not overcome the impact of slope position, particularly for canola grown on upper slopes, suggesting that factors other than N fertility, such as soil moisture availability, limited yields (Fig. 5 and 6). In both 1996 (Fig. 5) and 1997 (data not shown), the seed yield response of canola to N application generally was curvilinear on both the upper- and mid-slope positions. In the lowerslope positions, however, seed yield responses to N tended to be linear, indicating that yield maximums were not achieved, even at the 2 5 the recommended rate of N application. These results suggest that producers wanting to implement variable rate fertilizer application for canola should consider increasing fertilizer inputs in the most responsive areas of the field; namely the lower slope positions.
In 1998, canola was relatively unresponsive to N application, irrespective of landscape position (Fig. 6). Lack of response to fertilizer application during the 1998 field season was a reflection of the severe drought experienced early in the growing season at this site.
At the wheat site, yield responses to fertilizer N differed between years (e.g., Fig. 7 and 8). In 1998, wheat was unresponsive to fertilizer N additions, irrespective of landscape position (Fig. 8). The unpredictable nature of the wheat N response curves at Hepburn limited the potential for developing a successful variable rate N fertilizer strategy for this location. Clearly, the success of variable rate fertilizer application depends on our ability to predict crop response to inputs.
Although responses to N fertilizer were not consistent, some similarities in wheat yield responses to landscape existed. For example, wheat grain yields typically were lower on the upper slope units as compared to lower slope units.
Figure 7. Response of wheat, grown in 1997 at Hepburn, SK,
to N fertilizer on upper, mid and lower slope units. The
recommended rate of N fertilizer in 1997 was 67 kg/ha.
Figure 8. Response of wheat, grown in 1998 at Hepburn, SK, to N fertilizer on upper, mid and lower slope units. The recommended rate of N fertilizer in 1998 was 70 kg/ha.
A simple economic analysis revealed that increased economic returns were associated with the variable rate scenario developed for canola as compared to a blanket application of the recommended rate of N fertilizer (data not shown). However, it is important to note that a significant improvement in economic returns is required to cover the expenses associated with full implementation of variable rate fertilizer application. The variability in the wheat yield responses to fertilizer N limited the potential for developing a variable rate scenario and data suggest that implementation of a variable rate scenario at Hepburn according to our "prescription" would have resulted in a net loss in returns as compared to the blanket application of the recommended rate of N fertilizer. The economic viability of variable rate fertilizer application continues to be a focus of our ongoing research and we hope to have a more complete picture in the coming months.
The approach that we used in our study was to develop management units on the basis of landscape position. Results from this study indicate that these management units (upper-, mid- and lowerslope), developed on the basis of black and white aerial photographs, were agronomically meaningful for both wheat and canola production. Moreover, these management units were strongly tied to landscape position and are predictable, and recognizable.
Clearly, this approach to variable rate fertilizer management is only applicable to fields that have recognizable knolls and depressions. Thus, this approach is particularly well suited to landscapes developed on glacial till deposits. However, a different approach to soil sampling (e.g., grid sampling) and developing fertilizer recommendations is required for fields that are relatively flat, e.g., soils developed on lacustrine deposits (remnants of glacial lake beds).
Of particular importance is the observation that mean yields on upperslope units were consistently less than yields achieved on lowerslope units. Moreover, the greatest limitation to yield on the upperslope position at the research site was moisture. Application of N fertilizer to these upperslope positions did not overcome the limitation imposed by moisture availability. As a consequence, yield response to increasing rates of fertilizer N on the upperslope positions, in particular, tended to be curvilinear which demonstrates that a factor other than N fertilizer availability (i.e., moisture), imposed yield limitations in these positions. Interestingly, canola, in particular, continued to respond to increasing increments of fertilizer N on lowerslope units in both 1996 and 1997 which suggests that maximum response to fertilizer application was not achieved, even when fertilizer N was applied at 2 X the recommended rate of application. Thus, producers wanting to implement variable rate fertilization should consider increasing fertilizer inputs in the most responsive areas of the field; namely the lowerslope positions.
Generally speaking, the response of canola to fertilizer application was strongly tied to spring moisture availability which, in turn, was directly related to landscape position. This observation suggests that the success of variable rate fertilizer application in any given year likely is dependent, to a large degree, on the overall levels of soil moisture and the differences in soil moisture between landscape position. Furthermore, because yield responses were relatively consistent (and thus predictable) from year to year (although the overall mean varied depending on spring soil moisture), canola appears to be an excellent crop choice for producers interested in implementing variable rate fertilizer application.
Wheat, in contrast, did not respond consistently to fertilizer N application from year to year nor was the response of wheat to added N fertilizer predictable on the basis of spring soil moisture. We believe that the unpredictable nature of the response of wheat to N application as compared to the relatively predictable canola yield response to N reflects basic differences in the physiology of these two crops. In particular, the yield potential of wheat is known to be set relatively early in the growing season and is strongly influenced by early spring soil moisture availability (Johnston and Fowler, 1992). In addition, the response of wheat to N fertilizer additions includes both a yield component and a protein component - both of which are strongly influenced by available soil moisture. Thus, we were unable to predict the impact of N fertilizer on seed yield of wheat in the various landscape positions from year to year whereas canola yield responses were more consistent and predictable. Clearly the success of a variable rate fertilizer program rests on the ability to predict crop response to added inputs. On the basis of our observations, we concluded that although canola may be well suited to variable rate fertilizer management, wheat may not be as well-suited due to the unpredictable nature of the wheat yield response in the different landscape positions.
Many factors determine overall crop productivity in a given year. The variability of nutrients other than N and P, such as S or various micronutrients can be critical for certain crops; competition from weeds may be greatest in the lower slope positions; or in a wet year, problems with water-logging or root rot in lower slope positions may lower the yields. Through a combination of research trials on non-level fields and producer trials of different fertility-weed control scenarios, a more complete understanding of the management of variable productivity conditions will emerge.
Funding from Saskatchewan Agriculture and Food (Agriculture Development Fund) is gratefully acknowledged. The Saskatchewan Wheat Pool is recognized as a major contributor to this project. The financial support from Westco Fertilizers Ltd. and the Potash and Phosphate Institute of Canada is gratefully acknowledged.
Campbell, C.A., D.A. Rennie, L.D. Bailey. 1994. Nitrogen, Phosphorus, Potassium, Sulphur. Answers to Your Questions. Agriculture and Agrifood Canada Publication, Swift Current. ERDAS, Inc. 1991. Field Guide. ERDAS, Inc., Altlanta, Georgia.
Gregorich, E.G. and D. W. Anderson. 1985. Effects of cultivation and erosion on soil of four toposequences in the Canadian Prairies. Geoderma 36: 343-354.
Johnston, A.M. and D.B. Fowler. 1992. Response of no-till winter wheat to nitrogen fertilization and drought stress. Can. J. Plant Sci. 72:1075-1089.
McCann, B.L., D.J. Pennock C. van Kessel and F.L. Walley. 1996. The Development of management Units for site-specific farming. Proceedings of the 3rd International Conference on Precision Agriculture. Minneapolis, MN. June 23-26. ASA/CSSA/SSSA.
McConkey, B.G. , D.J. Ulrich and F. B. Dyck. 1997. Slope position and subsoiling effects on soil water and spring wheat yield. Can J. Soil Sci. 77: 83-90.
Pennock, D. J. and E. de Jong. 1990. Spatial pattern of soil redistribution in boroll landscapes, southern Saskatchewan, Canada. Soil Sci. 150: 867-873.
Sawyer, J.E. 1994. Concepts of variable rate technology with considerations for fertilizer applications. J. of Prod. Agric. 7:195-201.
Verity, G.E. and D.W. Anderson. 1990. Soil erosion effects on soil quality and yield. Can. J. Soil Sci. 70: 471-484.