Future agriculture with minimized phosphorus losses to waters: Research needs and direction

The main challenge

Improvements in agriculture in the last 50 years have dramatically increased grain and protein production in a very cost-effective manner (Fig. 1). For example, a tripling of global agricultural output in the last 50 years enabled food to become more abundant and cheaper, with agricultural prices falling about 1 % a year between 1900 and 2010, despite an increase in the world’s population from 1.7 to nearly 7.0 billion (Fuglie et al. 2012; Ball et al. 2014). Specialization and fragmentation of arable and animal production systems, however, have brought new pressures to bear on agricultural management within catchments. Historically, catchments generally had a sustainable nutrient balance. More recently, however, large amounts of nutrients removed, either as inputs (fertilizer and feed products) or outputs (meat and produce) on a national, regional, or global scale, which brings new pressures, challenges, and therefore, calls for new solutions (Cordell et al. 2009). Global complexities and interdependencies in P balancing are exemplified by the fact that increased grain and animal production in Brazil is making inroads into traditional U.S. markets and U.S. producers supply a large percent of the meat consumed in Japan, as water quality constraints in Japan limit cost-effective production there.

Several reviews and status perspectives of global cycling of P have acknowledged that the biogeochemical P cycle is broken at global, national, regional, and farm scales (Elser and Bennett 2011; Sharpley and Jarvie 2012; Jarvie et al. 2013b; Ulrich et al. 2013; Haygarth et al. 2014). The broken cycle of P from mining, fertilizer production, and land application of P in livestock manure, human waste, and food waste is the underlying problem limiting sustainable P use. Contributing to a solution is the need to realign the inputs of P, reuse P from manures and residuals (Bonvin et al. 2015; Stutter 2015), recover P from waste, redefine systems, and reduce P losses (the 5R’s of P; see also Schoumans et al. 2015; Withers et al. 2015). This is an update of “4R” nutrient management stewardship (right form, right time, right place, right amount; International Fertilizer Association 2009; International Plant Nutrition Institute 2014), specifically to address the broken P cycle, which is the root cause of P-related use impairments of water from accelerated eutrophication. Thus, increasing the efficiency of P use within agricultural systems that include intensively concentrated and spatially separated livestock and arable operations is critical for closing the broken P cycle (Fig. 1).

Increased P-use efficiency

At a cropping system level, management strategies available to improve overall P-use efficiency of cropping systems include the use of diverse crop rotations, the presence of cover or catch crops, and crop breeding for higher internal P-use efficiency, which will enhance crop P acquisition strategies. However, plant and microbial strategies need to be further improved, particularly the uptake of residual soil P (from applied fertilizers and manures) and subsoil P uptake (Richardson et al. 2011). At a fertilizer P-use level, the development and use of slow release fertilizers (including testing of products from pyrolysis of P-rich materials), mixing mineral and organic P fertilizers, fertigation, and the advancement of biofertilizers (i.e., inoculation of seed with effective P-solubilizing microorganisms), are important research questions. Applying P in the right form, right time, right place, and right amount is also vital, and one important question is how to enhance the availability of solubilized residual soil P without increasing P loss by leaching or runoff.

At a livestock system level, delegates identified the need to reduce feed P imports, encourage lower P animal diets that will not impact production goals, and encourage the addition of phytase to feed to make dietary phytate available to livestock. The high cost of recovering P from manure, in terms of the expense of equipment, infrastructure, and recurring chemicals remains a severe limitation to P recovery for most individual farm operations. Thus, there is an immediate need for the development of cost-efficient methods for P recovery from manure and organic by-products and sludges (e.g., separation-drying-pelleting, chemical extraction) but most are still too expensive relative to the value of the fertilizer product. Some technologies need up-scaling to improve cost-efficiency; each technology must avoid introducing other contaminants in treated by-products; others have legislative and attitudinal barriers to overcome.

Increased P-use efficiency comes at a cost, however. A label for sustainable production system, green labels, and environmental stewardship metrics could support higher food prices. However, the devil is in the details of program operation, to ensure that the metrics are defined, quantified, certified, and routinely verified.

Climate change

The increasing concerns of limited crop production, brought about by climate change, have direct and immediate consequences for food security. However, this is not explicitly a P issue, but more a water availability and use efficiency issue, which also has important implications for plant P uptake and P-use efficiency. More reliable predictors of the relationships among water use, plant growth, and P uptake are needed to assess the potential impacts of changes in rainfall distribution on P-use efficiency. However, there is a need to increase water-use efficiency and management strategies in Southern Europe through a combination of water storage, new irrigation techniques, and controlled drainage systems, and to optimize P management accordingly. Additionally, desalinization becomes even more important, together with changes in crop types and plant species and varieties, to enhance water and P-use efficiency where possible.

Changes in rainfall will likely affect runoff patterns, river flows, and the mobilization and transport of P. Increased soil and water temperatures will affect chemical reaction kinetics and microbial activities, which control the cycling and release of P along the land–water continuum (Whitehead et al. 2009). Also, shifts to more frequent extreme rainfall events may potentially impact P transfers and loss. In the Lake Erie Basin, 33 % more rain in intense events in the spring since 2009 than in the preceding 10 years (10 % of rain fell March, April, May) have led to the increased P runoff and contributed to the increased extent and intensity of algal blooms in the lake (Joosse and Baker 2011; Sharpley et al. 2012b; Smith et al. 2014). Options to promote resilience to climate change include improved water management strategies that limit losses of P and recycle water on arable land at the same time, methods to increase soil infiltration capacity, and methods to strengthen soil aggregates against dissolution (e.g., structured liming, application of organic matter, and adopting land practices that enhance organic matter build-up).

Policy interventions

Conference delegates identified the main challenges to P management as P imbalances at global and local scales, with P surplus in intensive livestock production regions, and the general disconnect of livestock production from areas of arable production where feed concentrates are produced. It was further concluded that low P-use efficiency by crops can contribute to the accumulation of residual P in soils. It was clear that demands for relatively inexpensive food and global markets are the main drivers of P transfers, disconnects, and losses which have broken the P cycle. Although it is clear that cheap food is needed in some parts of the world, the question was raised whether cheap food is what we really need in all cases. It was also pointed out that cheap food would not be cheap, if societal and environmental costs of unsustainable productions systems were accounted for and passed on to the consumer. Whilst policy interventions exist, they vary greatly from country to country, such that there is no clear, coherent roadmap for future changes in production systems and P cycles.

There is also an increasing call for a more serious debate about the structure of agriculture; should we centralize animal breeding and production even more to make P recovery from manure viable, or should we strive for decentralization of animal production to facilitate manure P application at rates that are aligned with crop needs? In a similar vein of system restructuring, the reduction in meat consumption, e.g., by introducing tax on meat has long been discussed as a means of decreasing the demand for concentrating livestock production and associated P surpluses with respect to local crop needs. In Sweden, for instance, reducing meat consumption has been debated in relation to both climate change and P runoff. In the Swedish general elections of September 2014, the debate transitioned to the possible adoption of a meat tax as a policy intervention to reduce meat consumption. In contrast, a recent study emphasized the need to facilitate informed public choices and presented a consumer meat guide where four environmental impact indicators were used in a life cycle approach (Röös et al. 2014). Several research needs associated with making environmental information comprehensive to the public were identified.

The main challenge

When P is land applied in the form of either mineral or organic fertilizer, surface and subsurface hydrological pathways transport P from the land surface, where it is a valuable resource used to achieve and maintain optimum plant yields, to receiving waters where it can be a major contributor of water-use impairment (Kleinman et al. 2009). Hydrological pathways along which P is transported are often complex and tortuous, involving multiple contributions from surface and subsurface transfers (Haygarth and Jarvis 1998a; Heathwaite and Dils 2000) (Fig. 1). It is difficult to make a reliable a priori assessment of whether transport and partitioning processes known from the literature are relevant for a particular catchment (Holländer et al. 2009). Educated guesses can be provided based on available basic information (e.g., topography, climate, soils) if this information suggests sufficient similarity with known cases. Such predictions can be improved by increasing the local knowledge based on different sources (e.g., farmers, geophysical information). However, not all factors and processes may be known from the literature and past experience may have serious biases.

The complexity of subsurface drainage and difficulties in deconvoluting flow pathways were identified as a key gap in our understanding of P transport and fate, particularly in groundwater-dominated watersheds (e.g., Jarvie et al. 2014), and during mixing of groundwater and surface waters during the hyporheic zone transport (e.g., Lapworth et al. 2011). Moreover, it was recognized that hydrological pathways along which P is transmitted do not simply transport P conservatively, but act as a series of reactive conduits, mediating P flux transformations through retention and recycling of P, on a range of timescales from years to centuries (Jarvie et al. 2012) (Fig. 1).

Short-term P retention along transport pathways may help protect downstream receiving waters from the acute effects of high P loads, providing an important ecosystem service, particularly in headwater streams (Hoellein et al. 2012). However, longer-term re-release of stored P can provide a chronic source of ‘legacy’ P. A fundamental research challenge is, therefore, to gain a better understanding of the magnitude and timescales over which P is retained and remobilized along transport pathways, how this impacts downstream receiving waters, and how this contributes to storage and release of ‘legacy’ P within the landscape (Sharpley et al. 2013; Lehtoranta et al. 2015).

Research needs

The main challenge

The U.S. Environmental Protection Agency (USEPA 2010a) considers non-point sources of sediment, nutrients, and pesticides as one of the leading causes of water quality impairments, as do most countries in Europe. By definition, non-point source contaminants are much harder to identify and thus, harder to manage than point sources. This is confounded by the fact that landscape hydrology is highly variable, both spatially and temporally.

Non-point source models represent mathematical descriptions, ranging from simple (risk assessment indices) to more complex scientific understanding about chemical, physical, and biological processes that influence both point and non-point source contaminant loads within a catchment (Fig. 1). In their most comprehensive form, models can integrate information over a catchment scale and suggest where beneficial management practices (BMPs) are most likely to decrease catchment-scale nutrient losses. Thus, use of non-point source models provides a method of simulating the risk of P loss or estimating actual losses including the relative effects of change in climate, land use, and land management practices on sediment and nutrient loadings from large complex catchments. As a result, models can quantify change to gage progress. Numerical ranking provides strong appeal to policy makers and managers; however, this appeal can sometimes bring false confidence and misconceptions (Boesch et al. 2001).

Monitoring

Monitoring is critical to addressing the main objectives of non-point sources management strategies, and present unique challenges to reliably represent site-specific variations in time and space (Verheyen et al. 2015). Monitoring programs are designed to identify nutrient losses and their sources areas, quantify the effects of mitigation measures, and document conservation program effectiveness. However, there is a cumulative uncertainty associated with water quality monitoring. This uncertainty is derived from, for example, stream flow measurement, water sample collection frequency, sample preservation and storage, and analysis (Toor et al. 2008).

Water quality data must further be related to information on catchment characteristics (e.g., soil properties, drainage conditions, contribution from point sources) and on agricultural activities such as crops grown, fertilization regimes, and soil cultivation practices. Access to such data is crucial for the interpretation of water quality data. Thus, the inherent landscape and management characteristics of monitored catchments must be stated, so that they can be related to surrounding agricultural areas where less information on agricultural management and nutrient loads are available. This would improve the applicability of monitoring results for larger agricultural areas (Kyllmar et al. 2014).

Modeling

It is of critical importance that model developers clearly define what the model is useful for and what it is not designed to do. Likewise, users must decide what they want to accomplish with a model. For example, one must consider the scale (field, catchment, or basin), time (flow event, annual, or multi-year), and level of accuracy (0.1 or 10 kg ha⁻¹ year⁻¹) that needs to be simulated, as well as the amount and quality of data available. Model uncertainty arises due to an imperfect representation of the physics, chemistry, and biology of the real world; numerical approximations of the governing equations; inaccurate model parameter estimates; and uncertainties in model input data (Harmel et al. 2006). Sources of model uncertainty are often interrelated and further complicating matters is the observation that multiple models and parameter sets may provide similar predictions of a given dataset (Beven 2006). Nevertheless, because models play an important role in scientific studies and public policy decision-making, every effort must be made to obtain reasonable estimates of model uncertainties (National Research Council 2008). When insufficient knowledge is available on the error distributions of model inputs and parameters, reasonable estimates can often be based on values reported in the literature or on professional judgment of the modelers (Haan et al. 1998; Harmel et al. 2010; Bolster and Vadas 2013).

Standardized methods to quantify this uncertainty involve forcing the model to “fit” historically measured data, if available, with predetermined limits of performance (Harmel et al. 2010). This will assist modelers in quantifying the “quality” of monitoring data for calibration and verification and will assist in determining model accuracy, and evaluating model performance. Whenever possible, the uncertainty should be represented in the model output (e.g., as a mean plus a standard deviation) or as confidence limits on the output of a time series of concentrations or flows. Knowledge of the cause and effects of uncertainty, as well as the measurement of uncertainty, is more important than the best-fit model output in making “real-world” management decisions. Thus, it is incumbent on the modeler to explicitly express the assumptions made in representing the system which is being modeled. This will allow the user to assess how these assumptions affect the model outcome and may ultimately affect decisions based upon modeled results.

Despite such cautionary realities, the role of models will be more and more important over the next decade in making catchment management and policy decisions to identify critical source areas and target BMPs and evaluate effects of climate change. Thus, model evaluation and uncertainty is essential and should be clearly documented in any model development (Beven 2006; Reichert and Mieleitner 2009; Andersson et al. 2014). It is, therefore, critical that any use of non-point source models must be associated with data collection and monitoring to further improve process representation in models and to test model estimates (Oreskes et al. 1994; Jakeman et al. 2006).

Communication

The tendency described earlier for decision makers to “believe” models because of their presumed deterministic nature and “exact” form of output must be tempered by responsible use of the models, such that model computations or “estimates” are not over-sold or given more weight than they deserve (Boesch et al. 2001; Pappenberger and Beven 2006). Above all, model users should determine that model computations are “reasonable” with respect to providing output that is realistic and based on input parameters that are within accepted ranges. Modelers should use all available measurements and multiple levels of comparison to evaluate if model estimates are realistic.

The main challenge

As a consequence of the spatial separation of arable and animal production systems, fertilizer P is imported to areas of grain production (Fig. 1). Phosphorus in harvested grain and other plant material is then transported to areas of animal production, where inefficient animal utilization of P in feed (<30 % utilized) results in P excreted as manure. This has led to a large-scale, one-way transfer of P from grain-to animal-producing areas that crosses catchment and even national boundaries and has dramatically broadened the emphasis of catchment management strategies. In many cases, environmental risk assessment has defined critical areas where manure is required to be applied in amounts aligned with crop P need, taking legacy P into account (van Bochove et al. 2006; Buczko and Kuchenbuch 2007; Heckrath et al. 2008; Sharpley et al. 2012a, b).

As the intensity of animal production within a catchment increase, farm P surplus (input minus output) becomes greater, soil P levels increase from land application of manure, and the overall risk of P and N loss increase (Pote et al. 1996; Haygarth et al. 1998; Withers et al. 2003; Djodjic et al. 2005; Svanbäck et al. 2013).

Research needs

Greater P recycling in general and a greater coordination of recycling at global, regional, local, and even farm levels are needed. For instance, in areas with large P accumulations in the soil, we need to focus on how to efficiently use this P, which in some cases might be by reducing P inputs to less than that removed in crop and forage export for several decades, until soil P concentrations are reduced. Exploitation of P resources in soil by rhizosphere management, i.e., by manipulating rhizosphere (the thin layer of soil surrounding roots) chemistry and biology to increase P mobilization and acquisition and reducing the reliance on chemical fertilizer P may offer opportunities to improve P utilization in low input systems (Richardson et al. 2011). Such opportunities could also include optimization of P inputs from chemical fertilizers, manures, wastes, or by-products that embrace the “4R” nutrient management approach, along with the use of different plant genotypes, and rhizosphere management strategies to stimulate P mobilization. Generally, the value of P in manures and in urban and industrial by-products needs full recognition and has to be appropriately accounted for in nutrient management schemes (Stutter 2015), which may require innovative integration of financial incentives or stricter regulations. Closing the P and other nutrient cycles is a fundamental question that needs to be answered locally as well as globally in order to increase P-use efficiency and meet stricter water quality standards and nutrient criteria, and produce cheap food in sustainable production systems.

The main challenge

There are many BMPs that can be implemented over a wide range of scales to minimize the loss of P from agriculture to surface and ground waters (Fig. 1). Marginal abatement cost curves and cost-benefit ratios can be useful to identify best (most cost-efficient) measures in a given situation, as can a landscape analysis based on elevation maps and soil mapping data (e.g., Hahn et al. 2014). Available BMPs are commonly grouped into measures that seek to reduce P accumulation in soils (i.e., control at source), decrease P mobilization from the source areas (e.g., by soil management or soil amendments) and those that are applied along the transport route between the source and watercourses (Uusitalo et al. 2015) (e.g., P filters; Fig. 1).

Farm input decisions

Carefully matching dietary P inputs to animal requirements can reduce the amount of P excreted by animals (Poulsen 2000; Valk et al. 2000). Implementing a carefully planned diet tailored to meet the specific P requirements of animals in each phase of their growth will minimize nutrient loss to the environment in feces, urine, and gases. Reducing farm inputs of P in animal feed is a very effective BMP that can contribute to bringing about a lasting decrease in P loss to the environment. In addition, BMPs that are designed to drawdown excess soil P reserves can contribute to achieving a P balance (Messiga et al. 2015). In fact, other nutrient management measures are generally aimed at decreasing the potential for P loss and are seen as short-term “band aids” and not long-term solutions.

Source management

Controlling at source is often the most cost-efficient measure, because soils that leak substantial amounts of P are logically beyond the point where further P applications give any substantial yield increases. Amending such soils with additional P is waste of resources, if not always from a farmer’s point of view (when the intent is to dispose animal wastes), from the viewpoint of society at least. Careful nutrient management planning on a field-by-field and farm basis is a major component of any remedial action plan to minimize the risk of nutrient loss from agricultural lands. As discussed above, one of the most important farm measures may be to process the manure to facilitate a more even distribution in the landscape and reduce the excess supply to some fields.

Transport management

Transport management refers to efforts to control the movement of P from soils to sensitive locations such as bodies of fresh water. Phosphorus loss from fields via surface runoff and erosion may be reduced by conservation tillage and crop residue management, terracing, contour tillage, and cover crops. Critical source area identification serves as an important basis for targeting fields and implementing edge-of-field remediation measures such as buffer strips, riparian zones, impoundments (e.g., settling basins), and wetlands. These practices tend to reduce rainfall impact on the soil surface, reduce runoff volume and velocity, and increase soil resistance to erosion. Control of P mobilization by soil management, such as conservation tillage practices, usually decreases the loss of particle-associated P but may increase those of dissolved P (Smith et al. 2015). Conservation tillage would be a method of choice on landscape positions prone to erosion, but on flat, non-eroding parts of a field, there might be more effective ways to bring about a decrease in P loss.

Chemical amendments have shown promising effects in P loss abatement (Uusitalo et al. 2015), and some measures can be real win–win solutions, e.g., structure liming (a mix of CaO and CaCO₃), which both improves the soil structure and crop yield, and reduces P losses (Ulén and Etana 2014; Svanbäck et al. 2014). However, none of these measures should be relied on as the sole or primary practice to reduce P losses in agricultural runoff.

Transport of P from agricultural catchments depends to a large extent upon the coincidence of source (soil, crop, and management) and transport factors (runoff, erosion, and proximity to water course or body). Source factors relate to catchment areas with a high potential to contribute to P export. For P, source areas can be spatially confined and limited in extent, generally reflecting soil P status and fertilizer and manure P inputs (Gburek and Sharpley 1998; Pionke et al. 2000). However, spatial uncertainties may still be considerable (Hahn et al. 2013).

Filters placed in the landscape, on field margins or in streams, are by rule more expensive options than those closer to the source of P loading. They may occupy productive land (e.g., wider buffer zones), require construction work (constructed wetlands), or need special materials (e.g., reactive permeable barrier-type P sequesters) that, even if having a low unit cost (e.g., steel slags), are economical to use in the proximity of their production sites only. Filters also need variable levels of maintenance.

The main challenge

For real and lasting changes to occur in agricultural systems, balancing production and environmental stewardship constraints, there needs to be a greater consideration of socioeconomic drivers of what, how, and why some conservation practices are adopted and others are not (Kleinman et al. 2015) (Fig. 1). This leads to a greater emphasis on consumer-driven programs and education, rather than assuming that farmers will absorb the total costs associated with implementing remedial practices. Remembering that, except for decreasing off-farm import of P and increasing on-farm P-use efficiency, BMPs are only a temporary band-aid to minimizing the off-site transport of P and receiving water impacts. There needs to be a discussion of how regional as well as national agricultural infrastructures can control P inputs to farming systems and assess large-scale nutrient balances. For example, cost-share monies for confined animal feeding operations in northeastern U.S. catchments are now linked to farmers demonstrating that nutrient inputs to the farm are reduced by feeding animals at a level consistent with standardized dietary requirements of P. This exemplifies how public policy can address the source or cause of excess P concentrations, and how public investments can provide a long-term mechanism for overcoming infrastructural barriers.

Research needs