Soil plant animal relations in nutrient cycling: the case of dairy farming system De Marke

Transcription

1 European Journal of Agronomy 13 (2000) Soil plant animal relations in nutrient cycling: the case of dairy farming system De Marke H. van Keulen *, H.F.M. Aarts, B. Habekotté, H.G. van der Meer, J.H.J. Spiertz Plant Research International, P.O. Box 16, 6700 AA Wageningen, The Netherlands Received 19 February 1999; received in revised form 12 July 1999; accepted 23 October 1999 Abstract Forage and ruminant production in Western Europe have increased significantly since World War II. However, in the last decade the livestock production sector has come under increasing pressure as the European Union introduced the milk quota system, effectively curbing total national and individual farm production volume, and national governments increasingly took measures to reduce the losses of nutrients from these systems to the environment. In the late 1980 s in the Netherlands a project was initiated with the objective to design, test and further develop a farming system that can serve as a starting point for the development of dairy farms on dry sandy soils with average milk production (system De Marke ) using the method of prototyping. First, a number of farming systems have been identified that, in theory, meet the formulated objectives. From this theoretically acceptable set one of the technically and economically most attractive and, from a research point of view, most interesting was implemented at the experimental farm De Marke in The functioning of this system is monitored in quantitative terms by measuring flows of dry matter and nutrients. The results sofar suggest that on dry sandy soils strict environmental standards for losses of nitrogen and accumulation of phosphorus can be attained within a short time while maintaining the current milk quota. Grass, maize and fodder beets could be grown in such a way that the norms for nitrate losses could be met. Improved utilization of animal manure and lower manuring levels allowed a reduction in the use of fertilizer-n of 74% compared to current practice. The input of P in feed and the output in milk and meat could be more or less balanced: P-surplus was only 18% of that on the current farm. To obtain reliable results the research will be continued for a longer period of time because the system has to stabilize, the soil nutrient and organic matter stores and soil fertility react slowly to changes in management, and weather conditions are always variable Elsevier Science B.V. All rights reserved. Keywords: Sustainable farming systems; Nutrient balances; Nutrient management; Environmental impact; Prototyping This paper is an elaborated version of the presentation to the 5th ESA-Congress by Spiertz et al. (1997). * Corresponding author: Tel.: ; fax: address: h.vankeulen@plant.wag.ur.nl (H. van Keulen) /00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S (00)

2 246 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) Introduction Forage and ruminant production in Western Europe have increased significantly since World War II under, on the one hand, the influence of the Common Agricultural Policy of the European Union that guaranteed high and stable product prices (De Wit et al., 1987; De Wit, 1988) and on the other, relatively low prices of inputs, particularly inorganic fertilizers and concentrates (Ketelaars and van der Meer, 1998). Inorganic fertilizers are, in general, a more reliable and easier-to-manage source of nutrients than either legumes or animal slurries, while concentrates serve as a ready means to supplement the unpredictable quantity and quality of roughage supply and allow intake levels necessary to sustain high levels of production per animal. Dairy farming is the major sector of Dutch agriculture, using about 64% of the cultivated area and serving as the main source of income for about 35% of the farmers (CBS, 1998). In the sixties and seventies these dairy farming systems strongly specialised and intensified by increased inputs of chemical fertilisers and purchased feeds. Although introduction of the milk quota system in 1984 reduced milk production by 15% the average is still high, i.e. between 1987 and 1997 about kg/ha on the specialised dairy farms. Especially, systems in the sandy regions in the East and South of the Netherlands are intensive, mainly because of the high prices of agricultural land. In the last decade the livestock production sector has come under increasing pressure as the European Union introduced the milk quota system, effectively curbing total national and individual farm production volume, and national governments increasingly took measures to reduce the losses of nutrients from these systems to the environment (Van Boheemen, 1987; Schröder, 1992; van der Meer et al., 1997). Intensification has led to a serious imbalance between inputs of nutrients in purchased fertilisers, concentrates and roughage and atmospheric deposition, and outputs in milk and meat, as illustrated in Table 1 for nitrogen at the national level in the Netherlands. At farm level for the period outputs were on average only 14% of the input for nitrogen (N), 32% for phosphorus (P) and 17% for potassium (K). So far most attention has been paid to pollution of water and soil resources by leaching and run-off of N, P and organic matter and accumulation of phosphorus. The N surplus contributes to environmental pollution through ammonia volatilisation, run off, leaching of nitrate and production of N 2 O during denitrification (Aarts et al., 1992). Ammonia volatilization has received particular attention in areas with high animal densities, where it contributes to high levels of atmospheric deposition. This high N-load causes eutrophication of terrestrial ecosystems, increased susceptibility of trees to stress, soil acidification and excessive nitrate leaching from affected ecosystems (van der Meer et al., 1997). Table 1 Nitrogen balance of the Dutch dairy farming sector (10 6 kg) a Year Imports Exports Surplus NUE b Fertilizer Concentrates Milk/Meat a Source: van der Meer et al. (1997). b NUE, nitrogen use efficiency: total output/total input.

3 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) Table 2 Maximum levy-free (see text for explanation) nutrient surpluses in the Netherlands (kg/ha, deposition and fixation by clover not included) N Grassland Arable land P Grassland Arable land Surplus P mainly accumulates in the soil, in addition to some surface flow, but continued accumulation will lead to saturation and leaching. Hence, the quality of the Dutch environment is threatened and one of the world s most vital and scarce minerals is wasted (Tisdale et al., 1985). Since the middle of the eighties the Dutch government is making efforts to reduce the negative impact of farming on the environment by formulating norms for nutrient surpluses. For nitrogen this implies that groundwater may contain only 50 mg nitrate per litre on light sandy soils it is now often more than 200 mg/l (Boumans and van Drecht, 1995) and that ammonia emission (from faeces and urine) should be limited to 30% of the average level in 1980, i.e. an annual loss of 30 kg NH 3 -N/ha at most. Field data suggest a required P supply in excess of crop uptake of kg/ha per year to maintain an optimal soil fertility level from an agricultural point of view. Based on average precipitation surplus and Dutch groundwater quality standards (maximum of 0.40 mg/l P-total), a loss of about 1 kg P 2 O 5 /ha per year would be acceptable, compared to about 30 in the mid-eighties (Oenema and van Dijk, 1994). Hence, to meet the standards for the year 2008 the P surplus has to be reduced by 72%. To realize those norms, first the use of animal manure was restricted on the basis of P load. As a result animal manures were better distributed over the country, but surpluses of N and P decreased by only 14% (Brouwer et al., 1997) and, thus, remained unacceptably high. Subsequently in 1997 maximum permitted surpluses of N and P for farming systems were defined (Table 2). These threshold values are temporary (political) compromises between environmental demands and agricultural possibilities (Dekker and van Leeuwen, 1998). Surpluses above the levels presented in Table 2 will be levied. However, uncertainty exists among dairy farmers on the impact of a reduced surplus of P (through lower inputs of fertilizer or concentrates) on soil fertility and as a consequence on crop production. Moreover, reducing the P surplus may lead to high costs because it limits the application of slurry, which then has to be exported or additional land has to be bought. This legislation stimulates more efficient nutrient management as farmers try to avoid fines while maintaining production. These measures have resulted in reduction in the surpluses on commercial farms (Aarts et al., 1999), but further reductions are necessary. One of the goals of agricultural research is to develop guidelines for farmers to design and improve their farms, taking into account their specific circumstances such as soil type, and both short- and long-term objectives. In addition, the government needs information on the practical feasibility of measures aimed at reducing the environmental impact of dairy farming. This paper describes the results of farming systems research focusing on dairy farming on sandy soils and long-term objectives with respect to nutrient surpluses, and, hence on improving nutrient management. 2. Objective dairy farming system De Marke A dairy farm can meet the environmental objectives of the government by extensification, i.e. by producing less milk per hectare (Weissbach and Ernst, 1994). For a given milk quota at farm level more land will then be needed, which, especially in the sandy areas, is scarce and expensive. In the Netherlands land taken out of production is mainly used for nature development, recreation, housing, industry and extension of infrastructure. Extensification (i.e. reducing milk production per unit area) without area expansion leads to re-

4 248 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) duced milk production and, hence, a strong decrease in income. Moreover, it is doubtful whether the growing world population can be provided with enough high-quality food in the future with a reduction in the area of cultivated land and more extensive land use. In the coming decades doubling of food production is expected to be necessary. Hence, the project De Marke aims at reducing environmental problems by focusing on the question to what extent technological options to improve nutrient management on dairy farms offer opportunities to maintain milk production per hectare. Characteristic for the dairy farm is the combination of plant and animal production. Animal manure and roughage form the links between the animal and the plant components. Nutrients not exported in milk and animals or lost, pass through a cycle: part of the nutrients from fodder is excreted by the animals in faeces and urine. This manure is being used in crop cultivation, its nutrients are taken up by the crop and then again consumed by the cattle. Nutrients exported in milk and animals and lost in the cycle are compensated for by the input of nutrients in fertilizers and purchased roughage and concentrates. Nutrient flows can be influenced by management, aiming at reducing losses. For instance, at most farms cattle take up much more nitrogen than is necessary physiologically. Hence, a reduction in the nitrogen content in the ration will hardly affect milk production, but will lead to less nitrogen in the urine and, consequently, lower losses. It is important that, wherever possible, measures in the different parts of the cycle be mutually synergystic, and geared to the specific possibilities and limitations of the farm. Cost minimization is an important criterion. The objective of this project is to design, test and further develop a farming system that can serve as a starting point for the development of dairy farms on dry sandy soils with average milk production (about kg/ha). The farming system should meet the strict environmental norms derived from policy papers as indicated in Section 1. To avoid desiccation the use of groundwater should be limited. Hence, irrigation is applied, but restrictively. Water stress may have negative effects on utilization of already applied fertilizers (environmental considerations) and may require purchase of larger quantities of fodder (economic and environmental considerations). Manure production is restricted to the amount that can reasonably be applied at the farm. 3. Research methodology 3.1. Prototyping In this research project the method of prototyping is used (Vereijken, 1992, 1997). Objectives of the farming system in terms of environmental criteria have been identified. This was followed by establishment of input output relationships for the plant and animal components of the farm, applying existing and especially developed and adapted calculation programmes. This refers, among others, to the relationship between crop yield and the required amounts of fertilizers and water, and that between daily milk production per animal and the required amounts of energy, protein and phosphorus (Aarts et al., 1992; Biewinga et al., 1992). On the basis of these relationships a number of farming systems have been designed that, in theory, meet the formulated objectives. From the theoretically acceptable set of systems one of the technically and economically most attractive and, from a research point of view, most interesting was implemented at the experimental farm De Marke in Subsequently, it was further developed with the aim to more fully realize the objectives. The functioning of this system also called De Marke is monitored as completely as possible in quantitative terms by measuring, wherever possible, flows of dry matter and nutrients. Moreover, detailed observations are carried out at a number of fixed sites, to increase insight in the processes taking place in the soil, mainly with respect to nitrogen, phosphorus, water and organic matter. The selected research method has some important advantages. The scientific approach to production systems concentrates at the farm level, the level at which management operates. The inte-

5 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) grated effect of processes is illustrated and the transfer of the knowledge acquired is facilitated as the functioning of the system can be directly observed. By building a bridge between theory and experimental testing the danger is avoided that, as in desk-studies, too much attention is paid to elements that appear less relevant in practice or the other way around that insufficient attention is paid to issues that prove important later. Prototyping also has some disadvantages. The experimental system has been selected, rather subjectively, from a number of options and reliable comparison with other systems is impossible, since only one system can be implemented and even that system is continuously developing. Therefore, the results can not be tested statistically. These disadvantages can (at least) partly be overcome by: (i) using an appropriate monitoring programme, (ii) additional (inter)disciplinary research aimed at determination of causal relationships that explain the behaviour of the system and can be used to investigate the consequences of alternatives, and (iii) combine the experimental research at the prototype farm with modelling to explore alternative possibilities (Van de Ven and van Keulen, 1996). Prototyping is also expensive: Implementation of the experimental system at De Marke required an investment of 10 million guilders and the annual budget amounts to about 1 million guilders Experimental methods For grass cut for conservation, silage maize and beets, dry matter yield was determined on the weighing bridge. It has been assumed that during conservation and feeding an additional 7% is lost. Intake of pasture grass has been estimated by visually estimating standing dry matter at the beginning and the end of the grazing period and assuming a daily growth rate of 50 kg dry matter/ ha. The quantities of N and P in faeces produced in the stable have been determined as volume times contents. The quantities in faeces and urine excreted during grazing have been estimated from the input in fodder by subtracting the output in milk and meat and the production in the stable. This calculation procedure may lead to relatively large inaccuracies in the partitioning of nutrients between stable and pasture. 4. Experimental farm De Marke To test the selected farming system 55 ha of land were bought in De Achterhoek, the eastern sandy part of The Netherlands. Hence, the experimental farm is almost twice the size of the average farm in the sandy area. This was considered necessary for research reasons: to obtain sufficiently reliable results more animals are needed than on the average farm. The land comprising the experimental farm De Marke, reclaimed from heather around the turn of last century, consists of an upper layer of cm with an average organic matter content of 4.9%, overlying practically humusless sand. The groundwater table is at most places too deep for water to reach the root zone. The water holding capacity is, therefore, less than 50 mm on 60% of the cultivated land and for the remainder varies between 50 and 100 mm. Hence, the land belongs to the driest 10% of the sandy soils in the country. The plots of De Marke, acquired from various landowners, had a different history with respect to the application of animal manure, resulting in considerable differences in soil fertility, especially for phosphorus (Schoumans, 1995). The total land area comprises the home plot, close to the farm buildings, that can be irrigated, if necessary, and is preferentially used for grazing, and the field plot more distant, with no irrigation possibilities and mainly used for silage making. The farming system of De Marke operates through very careful management of crops, livestock and manure Crops De Marke consists of 31 ha of grassland and 24 ha of maize (until ha and 6 ha of fodder beets). About one third of the grassland is permanent, the remainder in rotation with maize.

6 250 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) Most of the maize is used as whole crop silage and about 10 ha for ground ear silage (homegrown concentrate). Annual N application (available N in the slurry +fertiliser-n) does not exceed 250 kg/ha on grass and 100 kg/ha on maize. Actual rates of application take into account residual effects of preceding crops, particularly ploughing-in a grass sward or a cover crop. Maize is always followed by a cover crop to absorb residual inorganic N. In recent years no fertiliser P has been applied. Irrigation is applied only to enable grazing during part of the day (because of reasons of animal welfare) or to prevent early dying of crops due to drought Li estock Milk production per cow is high at 8150 kg per lactation, with almost 4.5% fat and 3.5% protein. Young stock is kept only for replacement. This explains the low N and P output in animals sold compared to the current farm. Much attention is paid to avoid protein and P surpluses in the ration of the cows. Grazing is restricted, particularly in late summer and autumn to avoid large N losses from urine patches. From April to September cows are stabled during part of the afternoon and at night, and supplemented with maize silage Animal manure De Marke has a free-stall barn, with a closed coated floor, sloping towards a central longitudinal urine drain. Alleys are equipped with a scraper to remove faeces and urine to a completely enclosed storage silo. All slurry produced is applied to the farm; on grassland by injection with open slits between early March and 15 August, and on maize land by injection just before sowing. The main characteristics of the experimental farming system are summarized in Tables 3 and 4. A comparison is being made between the expected values of these characteristics (prognosis) the Table 3 N and P balances of the average specialized dairy farm in the middle of the eighties, the prognoses at the start of De Marke and the values realized in the years 1993/1994 and 1994/1995 (kg/ha) Characteristic Average 1983/1986 De Marke prognoses 1993/ /1995 N P N P N P N P Input Fertilizer Feed Deposition N-fixation clover Various Sum Output Milk Meat Feed Mutation Number of animals Manure storage Feed stock Sum Input output a a Accumulation in soil and losses to air and groundwater.

7 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) Table 4 Characteristics of the cattle component of the average specialized dairy farm in the middle of the eighties, the prognoses at the start of De Marke and the characteristics realized in the years 1993/1994 and 1994/1995 Characteristic Average 1983/1986 De Marke 1993/ /95 prognoses Realized milk production (kg/ha) FPCM a (kg/ha) Milking cows/ha FPCM/cow (kg) Young stock/cow Purchased concentrate (kg DM/ha) Purchase of roughage (kg DM/ha) Feed intake cattle (kg DM/ha) Feed intake cattle (kg DM/kg FPCM) Nutrients (kg/ha) N P N P N P N P Input Feed Output Milk Meat Mutation cattle Input output (=Faeces+urine) a Fat and protein corrected milk. results of the model calculations at the basis of the farming system design and the results realized in the (financial, 1 May to 1 May) years 1993/1994 and 1994/1995, the first two years that the system functioned completely. The average farm from the middle of the eighties is taken as reference (henceforth referred to as the current farm ) as a discontinuity occurred at that time, because of the introduction of milk quota and of environmental legislation. 5. Results 5.1. The farm system The N and P balances of the farm are presented in Table 3. The prognosis was that the surplus at farm level would decrease from 487 kg N and 32 kg P/ha (surpluses at current farms) to 122 kg N and 0 kg P, respectively. The N surplus in the year 1993/1994 (140 kg N/ha) was close to the expected value, but it was clearly higher (198 kg N/ha) in the subsequent year, due to a higher fertilizer application rate and the reduction in stocks of slurry. The P surplus exceeded the expected value in both years, but if the changes in stocks of slurry are disregarded it is slightly negative in 1994/1995, i.e. slightly more P was exported from the farm in milk and cattle than imported in fertilizers and feed. As the import of P in feed appeared higher than expected application of fertilizer P was discontinued from the spring of 1994 onwards. Because of autonomous developments (especially legislation on low-emission application techniques and increasing milk production/cow) N surplus at the current farm has decreased by about 80 kg/ha (17%) to about 400 kg/ha (Van Eck, 1995), while the P surplus was hardly affected (30 kg/ha; Oenema and van Dijk, 1994).

8 252 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) The cattle component The milk quota of De Marke was slightly lower than that of the current farm, but reductions in the second half of the eighties have resulted in comparable quota during the reporting period. Model calculations (Table 4) suggest that the feed requirements for a given milk quota decrease when milk production per cow increases, as per kg of milk less energy is needed for maintenance of the animals and less young stock is needed for replacement (at a similar life span). Hence, the aim at De Marke was to realize a considerably higher milk production per cow than the current farm, which was realized and has resulted in a much lower stocking rate. The higher milk production has not resulted in the expected savings in feed. The energy intake (kvem (kvem is the abbreviation of the Dutch (kilo) digestible energy for milk production (Tamminga et al., 1994))) of the dairy cattle appeared substantially higher than expected on the basis of the feeding standards, without the animals showing signs of fattening (Meijer, 1994). The models were originally developed for a production range up to about 6500 kg/cow (personal communication, A. Meijer, Experimental Station for Cattle, Sheep and Horse Husbandry (PR)), and it appears, therefore, that extrapolation to the realized levels at De Marke is not without problems. The number of young stock per milking cow was higher than anticipated and, on average, practically similar to that on the current farm. The main reason is that in the reported period the herd was in its building-up phase, requiring maximum possibilities for selection. In practice, in the sandy areas, the number of young stock per milking cow increased to 1 (Leneman et al., 1999), while at De Marke it has decreased to 0.7. Because of the high energy requirements of the dairy cattle and the high number of young stock, dry matter intake per unit of milk production during the first year was 11% higher than predicted, but considerably below that at the current farm. In the second year it was only 4% higher than predicted. In formulating the ration, including grazing management, the aim is to minimize its protein and phosphorus contents. At De Marke animals are taken out of pasture about 1 month earlier than at the current farm and daily grazing time is restricted. In summer the cows are stabled during the afternoon and the night and are supplemented with maize. In the first year the cows were stabled at night only, but the alternate periods of availability of protein-rich (daytime grazing) and protein-poor (maize at night) feed appeared too long and led to digestive problems in the highly productive cattle. The dairy cows are moved to a new plot every 2 days, after which the pasture is further grazed by young stock. Restricting grazing time results in less urine and manure patches, and in a larger part of the faeces and urine being produced in the stable. Nutrients are thus utilized much more efficiently, as less N is lost from pasture, and application of slurry is synchronized to crop demand. As manure is spread on the land only between early March and mid-august, the required storage capacity at De Marke is substantially higher than at the current farm, that did not face any restrictions on manure application in the middle of the eighties. However, the time during which manure can be spread on sandy soils has been limited by legislation to the period 1 February 15 September, hence the storage capacity at the current farm has also increased. The difference between input and output in Table 4 represents excretion by the animals. At the current farm it exceeded 400 kg N/ha, indicating that only 16% of the N consumed by the animals was converted into milk and meat. The anticipated reduction at De Marke (50%) was based on increased utilization efficiency at animal level to 25% through a low-protein ration, high milk production per cow and a reduction in the number of young stock. The realized utilization efficiency of feed N was 22% due to the higher protein content in the ration protein intake exceeded the value expected on the basis of the CVB-standards (Meijer, 1994) by more than 18% and a higher number of young stock. P-utilization efficiency was also lower than expected, but the difference was smaller than for N.

9 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) The manure component Manure production at pasture was in agreement with the prognosis (Table 5) and nutrients in faeces and urine produced in excess of what was anticipated appear to have ended up in the stable. By restricting grazing time to about one third of current (assuming day-and-night grazing) excretion of nutrients at pasture was on average only 22 (N) and 17% (P) of the total, compared to 50% at the current farm. In the eighties, on a substantial part of the farms, dairy cattle were stabled during the night so that, on average, a smaller part of the nutrients in faeces and urine than given in Table 5, will have been excreted at pasture. Presently, on more than half of the farms on sandy soils dairy cattle are stabled at night. The difference between input and output of nitrogen in the manure represents volatilization of ammonia from the stable, at pasture, during storage and following application. Ammonia emission had to be estimated as it was not measured at De Marke during the period treated here, so that the results of the model, in relative terms, are reproduced. In absolute terms the losses were 3 4 kg/ha higher as more nitrogen was excreted by the animals. As at the current farm N-excretion is much higher and manure was surface-applied, N-loss as ammonia was 105 kg/ha. Due to the compulsory injection of manure, ammonia emission at current farms will have decreased at this moment by about 40 kg N/ha (Lekkerkerk et al., 1995). Of the total loss of more than 20 kg N/ha from faeces and urine 4 kg was lost during grazing (volatilization 7.5% of the N-excretion, Vertregt and Rutgers (1988)), 4 kg following application at pasture (5% of the ammonia-n), 0 kg after injection on arable land (1.25% of the ammonia-n) and 14 kg from the stable (7% of the N-excretion), which was, therefore, the main source of ammonia. (Ketelaars et al., 1995). At the moment, as more measured data for De Marke are available, the impression is that, even under careful management, volatilization losses are somewhat higher than assumed here (Van der Schans et al., 1999) The soil/crop component The area of the different forage crops is determined on the basis of a compromise between the production capacity of the crops under the conditions of De Marke, the value of their products in the ration and the possibilities to use animal manure efficiently. The proportion of grass in the rotation at De Marke is lower than at the current farm and the proportion of maize consequently higher (Table 6). The main reason is the aim to reduce output of N in urine and faeces, demanding energy-rich feed Table 5 Nutrient flows of the manure component for the average specialized dairy farm in the middle of the eighties, the prognoses at the start of De Marke and the values realized in the years 1993/1994 and 1994/1995 (kg/ha) Characteristic Average 1983/1986 De Marke 1993/ /1995 prognoses N P N P N P N P Input Production faeces and urine at pasture Production faeces and urine in the stable Output Faeces+urine at pasture (after volatilization) Applied slurry (after volatilization) Change in manure stock Input output (=Ammonia from manure)

10 254 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) Table 6 Characteristics of the crop component of the average specialized dairy farm in the middle of the eighties, the prognoses at the start of De Marke and the values realized in the years 1993/1994 and 1994/1995 Characteristic Average 1983/1986 De Marke prognoses 1993/ /1995 Area (% of total) Grassland Maize Fodder beets Artificial fertilizer (kg/ha) N P K N-fixation clover (kg/ha) Grass Net yield (kg DM/ha) a Maize Beets (leaf included) Average of farm a Net yield is the value after subtraction of grazing, harvesting, conservation and feeding losses, and thus equals animal intake. with a low N content, to compensate for the high nitrogen contents in the grass products in the ration. Supplementary feeding with maize serves as an energy buffer and, moreover, limits the risk of grass tetany during the grazing season. Maize silage has a higher energy content than grass silage, reducing the need for concentrate supplement in the winter period. Moreover, maize and fodder beets are much more efficient in terms of water and nutrient requirements per unit harvestable dry matter than grass (Aarts and Grashoff, 1993). At De Marke moisture availability is generally the most limiting factor for crop production and although the aim is to minimize water use for irrigation, because groundwater is a scarce resource, and because it requires energy and labour, the area of grassland exceeds that of arable land. Important reasons are the higher yields of N and P (reducing import of these elements in purchased feeds), the possibilities for grazing and the greater opportunities to use animal manure. Moreover, grassland stimulates soil organic matter build-up and, therefore, indirectly soil moisture supply. In recent years the proportion of maize in the rotation at current farms has also increased (in 1992/93 it comprised about 25% of the area at dairy farms in the sandy areas (Leneman et al., 1999)) and the situation at De Marke is not an exception anymore. Fodder beets are hardly grown in actual practice, but at De Marke they were intended to replace part of the concentrates. Moreover, the yield of dry matter and energy per hectare is as a rule very high, and its nitrogen uptake capacity exceeds that of maize, so that beets as first crop after the grassland period was considered safer. As large amounts of beets in the ration for highly productive dairy cattle appeared to lead to digestive problems, after the first year the proportion of beets in the ration in winter was reduced, and part of the beets was ensiled with maize, so that they could be incorporated in the summer ration. However, the beets then have to be harvested early, hence at a lower dry matter yield, and cultivation of a catch crop is necessary (to take up the nitrogen mineralized after harvest). Despite incorporation of the beets in the summer ration the original area of beets (6 ha) appeared too large for judicious feeding practice, and was reduced to 4 ha in 1994, and eventually discarded after At De Marke part of the maize is harvested as ground ear silage (GES), very high in energy and mainly used as concentrate replacer for the most

11 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) productive animals. Maize stover is also harvested and used as an absorbing bottom layer for the rather wet grass silage in autumn, which is then covered with beet leaves. This silage appeared very suitable for feeding young stock and dry cows. In current practice GES is also increasingly used, but the stover is not (yet) harvested, and less autumn grass is ensiled, as grazing generally continues for 1 month longer. The total grassland area is divided into 22 ha for ley farming and 9 ha for permanent pasture. Continuous arable farming leads to lower soil organic matter content than ley farming, eventually resulting in increased drought sensitivity. Yield of maize under ley farming is higher than under continuous cropping and the risk of buildup of weed populations (by selection or building up of resistance) is lower. Moreover, during the grass period herbicides, used on the arable crops, can be decomposed. Grass/clover mixtures were always sown immediately after harvest of the maize, hence clover entered the winter period in the seedling stage and appeared very sensitive to frost, leading to low clover densities after most winters. Consequently, nitrogen fixation by clover in grassland, calculated as 4.5% of clover dry matter yield (Biewinga et al., 1992), was substantially lower than expected. Sowing in spring may help to avoid this problem. The grass sward, after 3 years, is broken up in early spring. Subsequently, fodder beets are grown because they can use the nitrogen mineralized from the decomposing grass very efficiently followed by two (home plot) or four (field plot) years of maize. Additional research at De Marke has shown that cultivation of maize after 3 years of grass is also possible when no artificial fertilizer is applied: mineral nitrogen in the soil did not reach alarmingly high levels, and the quality of the groundwater was no different from that under other plots. Fertilizer application is plot-specific and based on the uptake capacity of the crops, taking into account soil moisture supplying capacity and the place of the crop in the rotation. For phosphorus the principle of equilibrium fertilization is applied, i.e. no more is applied than taken up by the crop. However, during the grassland period fertilizer application exceeds crop uptake, during the arable period it is lower. Under ley farming a larger part of the N-requirement of grassland can be covered by slurry because on grassland P in slurry is more restrictive than N and on leys higher quantities of P are allowed. On permanently arable land N in the manure is, as a rule, restrictive for the quantity of slurry that can be applied and adding fertilizer P may be necessary. In the ley system phosphorus requirements of the crops can almost completely be covered by animal manure. On average, during the 2 years, only 74 kg N/ha from fertilizer were required, i.e. about one quarter of that on the current farm. In 1993 and 1994 permanent grassland was manured on average with 47 tons of cow slurry and 136 kg fertilizer-n per ha (in total 237 kg of mineral N). Leys received 75 tons of slurry and 123 kg fertilizer-n per ha (in total 289 kg of mineral N). Maize and beets were fertilized with 27 and 34 tons of slurry/ha, respectively (65 and 85 kg of mineral N). On both grassland and arable land slurry was applied by injection, on grass with open slits. Between the rows of maize Italian ryegrass was sown in June, to take up the nitrogen mineralized during the ripening phase and after harvest of the maize. In the spring of 1995 the grass was grazed by young stock, in 1994 it was ploughed in. Grassland was irrigated when necessary to continue grazing in dry periods or to avoid re-sowing. Beets and maize were irrigated only when there was a risk of premature death. The field plot of the farm 29% of the total area with a relatively high proportion of arable land is never irrigated. During dry years as in 1994 on average about 50 mm/ha total farm area was applied, about 90% on grassland and 10% on maize land. The growing season 1993 was characterized by a warm and dry spring and a wet summer and on average only 9 mm of irrigation water was applied on grassland. The spring of 1994 was cold and wet and was followed by a very dry summer. The differences in yield among crops appeared in accordance with the expectations. In both years realized net dry matter yield of grassland was close to the prognosis for an average year. Yield of maize in 1993 was slightly below the expected

12 256 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) value. The very low yield of beets in 1994 was due to drought, an attack of rhizoctonia and early harvest, as part of the beets were ensiled with maize. Although the yields of grass and maize were on average lower than those on current farms, average farm yield is almost equal, because of a higher proportion of fodder beets and maize in the rotation. The feeding value of the products was, on average, slightly higher than that of the average of samples from practice. The amount of N in slurry after correction for ammonia volatilization was higher than at the current farm because of the strong reduction in the proportion of manure excreted at pasture and to the emission-reducing measures for slurry. In both years slurry was the main source of nutrients at De Marke (Table 7). In 1993/1994 total nutrient supply to the soil was similar to the prediction (355 kg N and 37.2 kg P/ha); in 1994/1995 it was substantially higher because more slurry and fertilizer were applied to grassland, as a different calculation method for fertilizer requirements was used. Moreover, the correction for lower yields due to drought was insufficient. Therefore, the store of slurry decreased in 1994/1995. The contribution of clover to the N-supply of the farm was lower than expected as explained earlier. N and P contents of the harvestable crop (gross yield) were in both years almost equal to the predictions, but considerably lower than those in practice. The relatively much higher nitrogen production at the current farm 398 kg N/ha versus, on average, 276 at De Marke reflects the combined effect of a higher proportion of grassland and a much higher fertilizer level on grassland. On average, during the 2 years, 135 kg N and 4.7 kg P per ha of the nutrients applied were not recovered in the harvested products. These quantities must thus have accumulated in the soil, leached, lost in surface runoff (Sharpley and Withers, 1994), or denitrified (only N). Modifications in the soil store can only be identified in the long run because of its large size: the soils at De Marke contain on average more than 7000 kg of N and more than 3100 kg of P per ha The feed component The proportion of nutrients from both harvestable crop and purchased feed ingested by the animals was in accordance with the predictions and, therefore, higher than at the current farm. Hence, nutrients in feed are used more efficiently, as a result of lower grazing losses due to short grazing periods per plot, stabling at night and a Table 7 Nutrient flows of the soil/crop component of the average specialized dairy farm in the middle of the eighties, the prognoses at the start of De Marke and the values realized in the years 1993/94 and 1994/95 (kg/ha) Characteristic Average De Marke 1993/ / /1986 prognoses N P N P N P N P Input Faeces and urine at pasture (after volatilization) Slurry application (after volatilization) Fertilizer Deposition N-fixation clover Harvest loss Output Harvestable crop Input-output a a Denitrification, nitrate leaching and accumulation.

13 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) Table 8 Nutrient flows for the feed component for the average specialized dairy farm in the middle of the eighties, the prognoses at the start of De Marke and the values realized in the years 1993/1994 and 1994/1995 (kg/ha) Characteristic Average 1983/1986 De Marke prognoses 1993/ /1995 N P N P N P N P Input Harvestable crop Purchased feed Output Feed intake Feed sold Mutation feed stock Input output a a Grazing, harvesting, conservation and feeding losses. short grazing season careful silage making, and utilization of maize straw and beet leaves, that at the current farm are left in the field. Inputs of N and P with feed from outside the farm and from the feed stocks were substantially higher than predicted because of a larger herd size (Table 8). Moreover, more N and P was imported with feed than necessary according to the feeding standards. Fodder with adapted nutrient contents is rather expensive and it is cheaper to limit the input of N and P on the farm through restricted fertilizer purchases. Differences between years in the balance of inputs and outputs are partly due to inaccuracies in determining the feed stocks. 6. The fate of the nutrient surplus Through denitrification a maximum of some tens of kg N/ha were lost annually from the top soil or the grass sward (Corré, 1996). Denitrification can be important in parts of the plots where water stagnates (lack of oxygen) and there is ample supply of easily decomposable organic matter. Denitrification from deeper layers is considered unimportant at De Marke because of the deep groundwater level, the low contents of organic matter in the subsoil and the limited amount of nitrate that leaches from the top soil. It has been assumed that during both years denitrification was 25 kg N/ha. In Fig. 1 the time course of nitrate content in the upper groundwater is given, measured by RIVM (National Institute of Public Health and Enviromental Protection), averaged for the farm Nitrogen The nitrogen surplus of the farm (Table 3) is the sum of ammonia volatilization, denitrification, surface run-off, leaching and modifications in the soil store. Surface run-off does not play a role, as there is hardly any surface water at De Marke. Only at extremely high rainfall will water be removed through ditches. About 22 kg of N per ha was lost as ammonia from manure and an estimated 2 kg from crop residues. Total loss of ammonia-n thus equals 24 kg plus an unknown quantity from ensiled products. Fig. 1. The concentration of nitrate in the upper groundwater of De Marke.

14 258 H. an Keulen et al. / Europ. J. Agronomy 13 (2000) The nitrate contents strongly decreased over time and were, on the last reported date, on average slightly below 50 mg/l, the requirement for drinking water. Nitrate contents under current plots close to De Marke measured by the Drinking Water Company Oostelijk Gelderland, in the autumn of 1993, were almost four times as high. Apparently, the adapted management results in realisation of the objective clean groundwater within a few years, where a much longer period was expected (Goossensen and Meeuwissen, 1990). No clear differences were found among crops, nor an effect of irrigation. On the basis of precipitation surplus and the measured nitrate contents in the groundwater annual leaching is estimated at 50 kg N/ha. The sum of N-losses through ammonia volatilization, denitrification and leaching of nitrate equals 99 kg/ha. The average surplus at farm level is 169 kg N/ha. The difference between these two values (70 kg N/ha) must be attributed to measuring errors, losses from ensiled feed and changes in the store of soil organic-n. On average, the store of soil organic-n appeared to have increased annually by 17 kg/ha between 1990 and 1995, but obviously that estimate is not very accurate. At De Marke net mineralization showed large variations among plots and among years. Generally, the results suggest that the more humid the conditions during the growing season the higher net mineralization (Aarts, 1996a,b). This could imply that if the soil store is in equilibrium in the long run it decreases in relatively wet years and increases during dry years Phosphorus The P-content in the upper groundwater was only 0.13 mg/l when sampled in October 1994, i.e. much lower than the target value for sandy soil (0.4 mg/l; Willems and Fraters (1995)). As hardly any P leached it is most likely that the small P-surplus accumulated in the soil. This is in agreement with the observation that the total amount of P in the soil did not change in the period 1989/ /1996 (Habekotté et al., 1998). However, despite a slightly positive P-surplus, the P-status of the soil in terms of plant-available P (P-Al and P w ; Van der Paauw (1956) and Van der Paauw (1971)) decreased by some 2 4% annually. This might be the expected response to the instantaneous reduction in the input of external P (slurry and chemical fertilizer). Because of the high inputs of P prior to implementation of De Marke, the labile pool of P in the upper layer of the soil is relatively high compared to the stable pool, which results in net transfer from the former to the latter (Janssen et al., 1987; Wolf et al., 1987). This process continues until the reverse flow is in equilibrium again. As P w and P-Al are most closely related to the size of the labile pool, these soil fertility indicators may be expected to further decrease in the (near) future. The decrease in P w and P-Al, in combination with the low P-fertilizer level caused problems in a number of plots during the early growth of maize (Aarts, 1995). Soil analyses indicate that the solubility of P has decreased, especially where the soil contained high concentrations of iron and aluminum (Schoumans, 1995). Until now, the significantly lower P-status of the soil and the associated slower early growth of maize have not resulted in lower yields (Habekotté et al., 1998). 7. Discussion and conclusions The results from De Marke so far suggest that on dry sandy soils strict environmental standards for losses of nitrogen and accumulation of phosphorus can be attained within a short time, while maintaining the current milk quota. The nutrient flows realized within the farming system agree well with the values calculated on the basis of existing knowledge. Intake of N and P by cattle was, however, higher than expected as the N- and P-contents of the ration were above the feeding standards and the number of young stock higher than anticipated. Nevertheless, input of N and P in purchased feed was only 50% of that of the current farm. The higher intake of N and P in feed resulted in higher than expected excretion in faeces and urine. This did not lead to problems with respect to the application of manure on the