Nutrient management issues specific to intensive grassland areas in North West Europe

In this paper we refer specifically to the permanent grassland areas that are used for intensive dairy farming, primarily within the UK, Ireland, the Netherlands, and parts of Germany, France and Denmark. The mild, moist temperate climatic conditions of these regions support a prolonged grass growing season (Lafferty et al., Reference Lafferty, Commins and Walsh1999). Soils tend to be fertile, temperatures are moderate and there is sufficient rainfall. They include grazed pastures and grassland harvested for hay and silage or a combination of both. These grasslands are subject to high external inputs of fertiliser and disturbance (e.g. biomass removal, reseeding and drainage) and are typically dominated by perennial ryegrass (Lolium perenne) and clover (Trifolium repens). Nutrient inputs are in the form of inorganic fertilisers (primarily nitrogen, phosphorus, potassium, magnesium and sulphur), and organic inputs in the form of animal manures and slurries. Typically these nutrients are applied as blanket applications across all fields. On many farms, soil sampling can be infrequent and would generally be inadequate and failing to reveal the true extent of the spatial and temporal range of plant nutrients present. Blanket applications of fertilisers and manures have created hotspots of nutrients (mainly nitrogen and phosphorus), contributing to environmental issues such as greenhouse gas (GHG) emissions and the eutrophication of freshwater bodies.

At the same time, productive areas exist where nutrient inputs are inadequate to meet crop demand, leading to deficiencies and affecting grass yields. Nitrogen, potassium and sulphur deficiencies can be common, and a decline in the application of agricultural lime has created many soils that are well below the optimum pH for permanent grassland. As a consequence, variation in grass yield of up to 4 tha^-1 within individual fields (McCormick, Reference McCormick2005) has been found to occur, and can be linked to differences in N mineralisation potential and N availability across fields (Bailey et al., Reference Bailey, Wang, Jordan and Higgins2001; Jordan et al., Reference Jordan, Shi, Bailey and Higgins2003). Grass yields at second cut are more likely to display nitrogen (N) deficiencies, with as much as 60% of cut silage swards being shown to be deficient in N at various stages of the growing season. Precision agriculture technology, by delivering real-time information on nutrient requirements as machinery moves across fields, can help to optimise yields by applying the correct amounts of nutrients where they are needed most, as well as tackle deficiencies and excesses.

In grazed pastures the spatial and temporal variation of nutrients within fields is even more complex and difficult to manage because of heterogenous animal excreta depositions such as around feeding troughs (e.g. West et al., Reference West, Mallarino, Wedin and Marx1989; Bramley and White, Reference Bramley and White1991; Cuttle, Reference Cuttle1992).

Sensors and Wireless Sensors

A large range of soil and crop sensors are currently on the market, including gamma-radiometric soil sensors, soil moisture devices, meteorological devices such as thermometers and hygrometers; sensors to quantify the physiological status of crops (for example nitrogen sensors), electromagnetic induction instrumentation, direct electrical methods and ground penetrating radar. Factors such as soil water content (Pedrera-Parrilla et al., Reference Pedrera-Parrilla, Brevik, Giráldez and Vanderlinden2016), dissolved salt content, clay content, clay mineralogy and soil temperature (Rhoades et al., Reference Rhoades, Manteghi, Shouse and Alves1989) have all been found to relate directly to electromagnetic induction data. Spectrometers collecting reflectance spectra are available as hand-held devices or can be fitted to farm machinery. A nitrogen sensor for example, attached to a tractor and fertiliser spreader, can calculate the amount of N taken up by the crop and translate this to an application rate. At the red and near infrared wavelengths, vegetation indices such as normalised difference vegetation indices (NDVI) can be calculated and used for pasture management activities (Flynn et al., Reference Flynn, Dougherty and Wendroth2008) by providing an indication of the greenness of the vegetation and its nitrogen content (Ramoelo et al., Reference Ramoelo, Skidmore, Cho, Schlerf, Mathieu and Heitonig2012), as well as soil cover and leaf area index.

In recent years, wireless sensor networks (WSNs) have emerged, and have been seen as a potentially valuable resource for PA because of their ability to facilitate real-time decision making, and act as a cost-effective communication technology (Garcia-Sanchez, Reference Garcia-Sanchez, Garcia-Sanchez and Garcia-Haro2011). Wireless communication includes technologies such as ZigBee, Bluetooth, Wibree and WiFi. A WSN connects the physical and computational world by monitoring environmental phenomena through ubiquitous ‘smart’ devices called sensor nodes or motes. Motes are battery-powered, small in size and usually connected through wireless links and a communication gateway. Data are forwarded from the motes to a base station which processes the data. The WSNs are valuable where there are a number of interacting factors affecting growth e.g. temperature, humidity, solar radiation, and particular terrain attributes such as pH, water content, topography (Akyildiz et al., 2002).

If successful, site-specific fertiliser applications, adjusted in-season, can increase fertiliser N use efficiency by up to 368% compared with common farmer practices (Diacono et al., Reference Diacono, Rubino and Montemurro2013; Miao et al., Reference Miao, Stewart and Zhang2011). If sensors are combined with GPS guidance systems, and GIS fertiliser prescription maps, another 10–20% savings can be made, depending on the inherent variability within the paddock (Hedley, Reference Hedley2014). As sensors rapidly proliferate in a wide variety of applications and industries, the cost of sensors has been continuously declining. Sensors have the ability to obtain large amounts of data in a short period of time. The massive volume of data generated by sensors, however, has the potential to overwhelm while providing limited value unless the structure and processes are in place to take advantage of their full potential. Correct interpretation of sensor data continues to be a problem. Many sensors rely on battery lifetime, which is finite, and other issues have arisen such as computing power, memory size and data security concerns (Aqeel-ur-Rehman et al., Reference Aqeel-ur-Rehman, Islam and Shaikh2014).

Variable-Rate Technology, Auto-Steer and Controlled Traffic Farming

Since 1999 there have been massive improvements in the accuracy and miniaturization of Global Navigation Satellite System (GNSS) technology. The GNSS technology facilitates the linkage of data to specific geographical coordinates (geo-referencing) and this can be combined using auto-guidance in machinery. Global positioning systems (GPSs) are now commonly used on farms for auto-steer systems and the production of geo-referenced information e.g. yield mapping. GNSSs attached to soil and crop sensors will enable soil and crop monitoring at fine spatial scales.

The introduction of GPS alone (in autosteer) onto farm machinery can increase efficiencies by 5–10% through a reduction in overlaps and gaps in fertiliser spreading (Craighead and Yule, Reference Craighead and Yule2014). Auto-steer and controlled traffic farming systems guide machinery along repeatable tracks in fields with accuracy, thereby reducing errors and fatigue, and increasing the timeliness of operations. Controlled traffic farming has proved to be successful within arable farming situations, with a reduction in machinery and input costs of up to 75% in some cases, whilst also increasing crop yields (Tulberg et al., Reference Tullber, Yule and McGarry2007; Bowman, Reference Bowman2008). Controlled traffic farming could enhance the precision of fertiliser applications within grassland systems, thus reducing the incidence of nutrient hotspots developing, and nutrient losses through leaching and gaseous emissions. By confining vehicles to a minimal area of traffic lines, unnecessary crop damage can be avoided and soil compaction reduced.

Variable-rate fertiliser application methods strive to apply fertiliser according to the specific requirements of the soil or crop, as they vary across a field. One way of doing this is to pre-determine fertiliser requirements at designated GPS points by manual soil sampling and transforming the data into an application rate map. The increased range of applicator and technology now available makes it possible to operate in real-time, avoiding the need for pre-determined rate maps.

Along with inorganic fertilisers, animal slurry is also an essential nutrient source for intensive grassland. Schellberg and Lock (Reference Schellberg and Lock2009) presented a technique which allows site-specific application of slurry to grassland and arable land based on pre-defined application maps. The system consisted of a valve controlling rate of flow by an on-board PC. During operation, rate of flow was measured and scaled against set point values given in an application map together with the geographic position of the site.

Remote Sensing

Space borne remote sensing-based methods gather point- or area-data at spatial resolutions of less than 2 cm to over 10 m, and are useful for monitoring areas of land, and crops at various stages of crop development. Grassland vegetation characteristics such as standing biomass, canopy characteristics, biomass production (Maselli et al., 2006), sward height and floristic composition (Feilhauer et al., Reference Feilhauer, Thonfeld, Faude, He, Rocchini and Schmidtlein2013) have all been mapped using different sensor specifications depending on the scale of application. Data from Landsat TM/MSS, SPOT, AVHRR, MODIS and RapidEye® sensors are most commonly used for land cover classification purposes and land cover change mapping. For managed grasslands and paddocks, sensors with high spatial and temporal resolution are more effective, for example GeoEye, RapidEye and QuickBird. These provide imagery with a higher spatial resolution and revisit time, thereby permitting the detection of spatio-temporal patterns at scales closer to that of field surveys. Integrating multispectral and multitemporal remote sensing data with local knowledge and simulation models has been demonstrated successfully as a valuable approach to identifying and monitoring a wide variety of agriculturally related characteristics. With all remote sensing, it is essential to have high quality ground-truthed data for cross validation. Ali et al. (Reference Ali, Cawkwell, Dwyer, Barrett and Green2016) reviewed the application of satellite remote sensing for grassland and its transition from grassland mapping to grassland or pasture monitoring and management. The red edge, where there is a rapid increase in reflectance from the red to NIR reflectance (around 680–730 nm) has been shown to have a strong correlation with the grass chlorophyll content (Pinar and Curran, Reference Pinar and Curran1996). RapidEye® is the first high-resolution multispectral satellite system that operationally provides a red edge channel. Schuster et al. (Reference Schuster, Förster and Kleinschmit2012) reported that this enables higher classification accuracy for managed grassland types.

Ideally, a time series of imagery acquired through a growing season is most beneficial in terms of providing maximum information on yields and management. This is particularly important in grassland areas because of the huge temporal differences in plant phenology and farm management (e.g. grazing and mowing activities). Acquiring a multi-temporal image series is more effective than using single data acquisitions (Kasischke et al., Reference Kasischke, Mekack and Dobson1997; Pierce et al., Reference Pierce, Bergen, Dobson and Ulaby1998; Ranson and Sun, Reference Ranson and Sun1994). Obtaining multiple images can, however, be difficult in North West Europe where the dominance of clouds can often prohibit the generation of suitable images at crucial times during the growing season. The insufficient revisit time of satellites can be a constraint to the detection of the dynamics of growth patterns and defoliation regime on pastures. Climatic events and natural hazards can also evolve rapidly and unpredictably during the growing season and so present problems for data capture. This is one of the major constraints of remote sensing.

Synthetic Aperture Radar (SAR)

There has been a growing interest in the potential offered by microwave spaceborne instruments which measure the strength of the back-scattered signal from the surface under almost all weather and light conditions, allowing frequent repeat measurements throughout the growing season (Ali et al., Reference Ali, Cawkwell, Dwyer, Barrett and Green2016). SyntheticaAperture radar (SAR) instruments offer a number of different acquisition modes, with different polarizations, incidence angles and orbital directions (ascending or descending). Barrett et al. (Reference Barrett, Nitze, Green and Cawkwell2014) used the Essential climate variable soil moisture product (Liu et al., Reference Liu, Dorigo, Parinussa, De Jeu, Wagner, McCabe, Evans and van Dijk2012) to provide an estimate of the soil moisture conditions on SAR acquisition dates. The SAR data are almost independent of weather and illumination conditions, and this makes their use for operational purposes especially appealing. Synthetic aperture radar is therefore an important alternative or complementary data source for areas with persistent cloud cover.

Digital Image Analysis

Digital image analysis is able to distinguish individual plant species based on characteristic shape, colour and texture. It has the potential to support the identification of plant species required for example, for site-specific weed control in grassland swards (Gebhardt et al., Reference Gebhardt, Schellberg, Lock and Kühbauch2007). Digital image analysis includes close range imagery using digital photographs, and any kind of remotely sensed and geographically referenced pixel-based information. Distinguishing between species and identifying features and characteristics in grasslands can be complex because of the diversity of optical plant properties within a mixed sward, such as leaf colour and shape, overlapping of leaves and tillers, shades on leaves and soil, as well as the non-uniform soil background (Gebhardt et al., Reference Gebhardt, Schellberg, Lock and Kühbauch2007). Spectral data is also very sensitive to the stage of crop maturity at the date of image acquisition.

Unmanned Aerial Vehicles (Drones)

In recent years there have been marked advances in the use of airborne and satellite platforms at lower altitudes, such as unmanned aerial vehicles, to sense across various wavelengths (visible, near infrared and thermal), using multispectral and hyperspectral cameras. Unmanned aerial vehicles (UAVs) have the ability to minimize pixel size to as little as a few millimetres and thereby overcome spatial resolution problems. They are low-cost and light-weight. UAV platforms are evolving rapidly, both technically and with regard to regulations. The miniaturization of onboard sensors has resulted in an explosion in their potential application (Huang et al., Reference Huang, Thomson, Hoffmann, Lan and Fritz2013; Zhang and Kovacs, Reference Zhang and Kovacs2012). UAVs offer design and performance advantages over conventional photo-reconnaissance aircraft and other remote sensing platforms, such as their small size, low weight, slow flight speed, extended range, extreme altitude and endurance. Their potential application in intensive grassland farming has yet to be fully explored, but their ability to generate very high resolution (2 to 10 cm) farm-level imagery within time frames of a few minutes to a few hours offers great scope and potential (Chen et al., Reference Chen, Vierling, Rowell and DeFelice2004).

The on-board sensors on UAVs can provide information on biomass, leaf area index, disease and water stress, which can be used to inform decision-making in crop management, yield forecasting and environmental protection (Zhang and Kovacs, Reference Zhang and Kovacs2012). They have been successfully employed in the derivation of vegetation indices to explain canopy condition, such as chlorophyll content and stress level, and its variability in space and time.

The ‘red edge inflexion point’ from sensors mounted on UAV platforms has been found to be correlated with nitrogen status, and so could potentially be used as an indicator of soil nitrogen supply and uptake across fields. Remote sensing applications have also been shown to indicate pH reliably (Roelefsen et al., Reference Roelofsen, van Bodegom, Kooistra, van Amerongen and White2015). LiDAR sensors can also be supported on UAV platforms. LIDAR (light dand ranging) sensors measure directly both the vertical location and horizontal distribution of plant canopy elements, ideally using wavelengths in the infrared or near infrared (Fowler, Reference Fowler2000).

Data Handling

The rapid development of ICT-based sensor technologies in recent years, along with dedicated software that enables real-time soil and crop sensing, has facilitated the great advances in PA applications that have taken place in recent years. However, the challenge is how to handle appropriately the vast amounts of data that are generated using PA technology. Output data need to be processed and standardised across systems, so that they are user-friendly and on which farmers can base management decisions. Understanding the response of sensors to grass crops, to nutrients, and their interaction with other resources such as water and soil properties still requires research. For technical solutions to be of benefit, they need to be easily integrated into current management regimes. This is a complex procedure and as such, data management and compatibility are currently some of the main limitations to widespread adoption of PA. Data sharing from non-standard software and data formatting are also big issues. Data ownership is particularly sensitive, particularly if contractors, farm advisors or technology providers become involved in farm management decisions.

The Internet of things (IoT) is the interconnection of uniquely identifiable embedded computing devices within the existing Internet infrastructure, which will in future make it possible to connect all devices that are available on the farm. Research is still required to enable communication between sensors and farm equipment. Farming ‘in the cloud’ is based on internet storage sites rather than relying on storage devices such as USB sticks and flash cards, but this all requires considerable research before it can be applied practically on farms. A combination of technology with the Internet would allow for example, the results of ‘big data’ collected by remote sensing imagery or UAV equipment to be obtained by the farmer on a real-time basis or within a few hours of acquisition to facilitate decision making (EIP-Agri, 2015). This combined with wireless sensors or mobile telephone devices could offer considerable market potential. A range of industry providers, particularly SMEs could benefit from these advances in technology.