Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

CHARACTERIZATION OF SMALL- TO

MIDDLE-SCALE DAIRIES’ MANURE.


STORING AND LANDFARMING MODELS
IN CANTABRIA (NORTHERN SPAIN)

P. SANTORUM*, R. GARCÍA* AND B. FERNÁNDEZ**


* Centro de Investigación y Formación Agrarias, Consejería de Ganadería,
Agricultura y Pesca de Cantabria, C/ Héroes 2 Mayo 27, 39600 Muriedas,
Cantabria, Spain.
** Servicio de Agricultura y Diversificación Rural, Consejería de Ganadería,
Agricultura y Pesca de Cantabria, C/ Gutierrez Solana, s/n, 39011 Santander,
Cantabria, Spain.

SUMMARY: Modern and traditional manure managements coexist nowadays in dairies of the
region. The objectives of this work were to collect current information about the agricultural
practices in dairies and to define the existing models of small- to middle-scale dairies with
regard to the manure storage, management and application on land. Cluster analyses were
achieved in order to distinguish dairies in relation to these characteristics. Concordance between
three types of cluster analyses performed with SPSS software confirmed the existence of four
clusters. The factors defining these clusters were the size of the dairy, represented by the milk
quota in kg produced per year, the manure store type and the application system to the land.

1. INTRODUCTION

Cantabria is the third producer of milk in Spanish regions and agriculture-related companies as
well as dairy food producers generate the third part of region’s incomes. Being the stockbreeding
the occupation of the vast majority of the farmers and assuming that more than two thousand
farmers are devoted to milk production, efforts in studying dairies’ wastes must be implemented.
The significance of agriculture in the whole economy of the region is parallel to the relative
importance of controlling agricultural wastes for environmental protection.
During the last thirty years the initially high number of milk producers had to face the
renovation of their facilities to fulfill the updated legislations. Firstly, milk quality was the
objective of the new changes in facilities. Now that facilities carry out that objective, special
attention must be given to the waste storage and management. As European legislation demands,
water must be protected from nitrates of agricultural origin. Unfortunately, suitable milk
production facilities are not always complemented with properly designed waste stores or waste
management plans. In fact, given that the high average age of the farmers and the climate- and
geography-originated limitations of the mountains’ producers, modern and traditional manure

Proceedings Sardinia 2007, Eleventh International Waste Management and Landfill Symposium
S. Margherita di Pula, Cagliari, Italy; 1 - 5 October 2007
¤ 2007 by CISA, Environmental Sanitary Engineering Centre, Italy
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

managements still coexist in dairies of the region. Nowadays, escalation in dairies’ magnitude is
pushing the sector restructuring, so in this moment the high diversity of manure management
will be presumably reduced in a short time.
The collection of current information about the agricultural practices in dairies was the primary
objective of this work, as well as the definition of the existing models of small- to middle-scale
dairies regarding the manure storage, manure management and manure application on land.

2. EXPERIMENTAL STUDY

2.1 Poll design


The polls carried out in this region in 1995 (Stockbreeding development) and in 1998 (Tejero, I.)
were reviewed to design this questionnaire. Organization of the questions followed the scheme
broken down:
Ɣ General information about the farm: address, legal status, owners, family assistance,
labourers, external services contracted and expansion plan.
Ɣ Agriculture: no. of plots, grassland/forage or owned/rented area, slope, silage and machinery.
Ɣ Stockbreeding: milk quota since 2001, sheds, no. of heads, breeds, pens, feeders, drinking
places, grazing periods, no. of heads bought or put down, shed floor, bed type and cleaning.
Ɣ Milk production: milking parlour, kg/day cow, tank capacity, forage and fodder: kg/day cow.
Ɣ Water supply: origin, consumption, water tanks and pumping.
Ɣ Manure storage: size, cover, material, filling and emptying rate, distance to fertilized plots, to
urban/swimming/water supply areas, mixers, solid waste separators and expansion plan.
Ɣ Fertilizers and organic amendments: spreading method, manure spreader type, quantity, crop
and area fertilized, and period of fertilization.
Ɣ Waste disposal: origin, volume and disposal of wastewaters and solid wastes.

2.2 Farm selection


Data about the collective of dairy farms was obtained from administrative database, provided by
the Animal Production Service (2004). The milk quota as well as the identity of the ownwers,
address, phone and location related to each farm corresponded to the data brought up to date on
3rd February 2004. Farms were arranged by milk quota assigned, and grouped in five types (see
Table 1).
In order to get a representative portion of each of these groups, 4% of the farms in each type
were randomly chosen to be polled. So, random values from uniform distribution were created in
SPSS software (release 12.0.1), being up to the 4% of the farms present in each type (see Table
1). A list of reserve farms by farm type was prepared, which were randomly chosen as for the
initially proposed farms, and at least 1.5% of the farms present in each type were selected.

2.3 Data collection and farm location areas


Two pollsters visited the initially proposed farms from 15 March to 15 April 2004, asking about
the facilities used, the management of the farm and the wastes produced in dairy farms in 2003.
When all the farms had been visited, so the list of reserve farms was given to the pollsters. From
the proposed reserve farms, it was necessary to poll 10 out of 59 for two weeks (see Table 1).
Half of the farms needed to be replaced were located in mountainous areas and administered a
milk quota lower than 100,000 kg per year, and these farmers were difficult to find due to the
spring grazing in high pastures and the dispersed location of livestock cabins in these areas.
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

Table 1. Polls obtained by farm type. No. of farms polled included the initially proposed and
polled ones plus the reserve farms polled. 1: milk production in kg per year.
Farm type1 Total no. Farms Reserve Reserve
of farms polled farms farms polled
< 50,000 920 37 15 5
50,000-100,000 854 34 15 3
100,000-200,000 988 40 15 1
200,000-500,000 656 26 10 1
> 500,000 102 4 4 0
Total no. of cases 3,520 141 59 10

All the lacking farms were replaced with reserve farms, taking into account farm type and
location.
Farms polled were located in five traditionally defined agricultural areas, given that no poll
was done in Tudanca-Cabuérniga, which only hosted 0.4% of the milk production in the region.
The farms visited were located in the traditionally defined agricultural areas: Coast, Pas-Iguña
and Asón (which represented the 96.5% of the milk production in Cantabria) as well as Liébana
and Reinosa. Geographical distribution of the polled farms was defined regarding to traditionally
defined agricultural areas as well as altitude criteria. The defined areas for this study were six
and included the following number of polls, out of 141: Western coast (21), Central coast (25),
Eastern coast (17), coast mid-valleys (31), mountain (21) and inland mid-valleys (26).

2.4 Statistical analyses

2.4.1 Variable selection


Each dairy farming was classified by milk quota (see farm types in Table 1) and by location (see
geographical areas defined in previous section). The corresponding parameters for each part of
the poll were compared by farm types and areas using SPSS software (release 12.0.1).
The analysis of variance was performed for continuous variables distributed in a normal
function (Shapiro-Wilk test, significance higher than 0.200) and when homogeinity of variances
assumption was accepted regarding to Levene’s test (significance lower than 0.050). Unequal
means were pairwise compared by the Tukey-Kramer’s test of honest significance difference for
homogeneous variances and by the Dunnett’s T3 test for heterogeneous variances. In case data
distribution was not normal and variances were heterogeneous, Welch’s robust test of equality of
means was applied (Reed, J.F. and D.B. Stark, 1988). Non-parametric tests were carried out
against categoric variables (chi-squared tests). Statistical significance was set at P<0.050.
Parameters which showed differences relating manure management to farm types or areas
were chosen for further analyses.

2.4.2 Analysis of conglomerates


Only independent factors were used for analysis of the selected variables. Normal distribution
was ascertained as well as homogeinity of variances for continuous variables and multinomial
distribution was checked for categorical variables. When the variable was a linear combination
of some of the variables, it was discarded for clustering, so selected ones were also lack of
multicollinearity. The distribution of standardized residuals for each group was plotted and
visually assessed for random distribution, so that variables did not present autocorrelation.
Variables which fulfilled these minimum requirements were selected.
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

Table 2. Continuous variables related to manure management, in which differences in means


were considered statistically significant, between farm types or geographical areas
(P<0.05). 1: when they are • 1. 2: spring, summer or winter spreadings from tanks.
Poll section Farm types Geographical areas
No. of tanks No. of tanks1
Capacity of tanks Tank depth
Manure Tank depth, length and width Tank volume by spreader volume
storage Tank volume/spreader volume No. of season spreadings1,2
No. of spring tank spreadings
Capacity of tanker Capacity of own tanker
Organic No. of tankers spread1 Manual spreading or irrigation area
amendments Tanker spreading area
Manual spreading or irrigation area

The continuous data were typified to overcome differences in scales and categorical ones were
properly recodified. Lacking data were checked to confirm they were randomly distributed.
Multivariate analysis of conglomerates was performed by utilizing three clustering
techniques: non-hierarchical, hierarchical and two-level approach, using SPSS software (release
12.0.1). Clustering analysis was performed to look for clusters in the data, in such a way that
objects belonging to the same cluster resembled each other, whereas objects in different clusters
were dissimilar. Non-hierarchical clustering was assessed by k-means method, calculating
Euclidean distances (Pérez, C., 2004). Algorithm was iterated for a high number of clusters, so
that atypical cases were detected and then iteration and classification were repeated for a lower
number of groups. The defined clusters were used as onset of a step-wise selection of the
variables. Finally, to check how much the sampling of sub-populations affected to the definition
of clusters, randomly selected farms (10%) were analyzed by iteration and classification and the
final cluster centers were used as initial centers for further classification of the whole data set, as
software recommended.
In the second place, an agglomerative hierarchical clustering analysis was performed to look
for groups in the data, using a hierarchical algorithm. The similarities of the polled farms were
evaluated calculating the squared Euclidean distance and using average linkage method (between
groups). To check if the processing of ties affected the groups defined, farm positions in the
database were changed around before clustering was repeated.
Finally, two-level analyses of conglomerates were achieved with selected variables and
distances were weighed up using log-likelihood function, which can be interpreted as dispersion
within clusters. Similar to agglomerative hierarchical clustering, those clusters with the smallest
distance were merged in each step. Each cluster step took sub-clusters obtained from the
previous step as input and then grouped them into the desired number of clusters (Rezanková,
H., D. Húsek & al, 2004). This procedure automatically selected the optimum number of
clusters, as the value of highest BIC (Bayesian Information Criterion, Schwartz, G., 1978) while
the relative BIC ratio was highest and not the unit value (Bacher, J., K. Wenzig & al, 2004;
Pérez, C., 2004).

3. RESULTS AND DISCUSSION

The size of the farms included in the analysis of conglomerates corresponded to small- to
middle-scale dairies, ranging from 1,000 to 500,000 kg of milk produced per year. The factors
used for analysis of the selected variables were farm size and geographical area.
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

Table 3. Categorical variables related to manure management, in which categories were not
equally distributed by farm type or geographical area (chi-squared test, P<0.05).
Poll section Farm types Geographical areas
Manure Covered/uncovered tanks Covered/uncovered tanks
storage Tank material Tank material
Agitator installed in tanks Method of manure agitation
Solid waste separator installed Agreement with manure sampling
Scrapper installed Complete or partial emptying of tank
Organic Spreading method Spreading method
amendments Spreader type Spreader type
No tanker spreading on grasslands Tanker spreading on maize fields

Differences between management practices in farm types or geographical areas considered


statistically significant are quoted in this section. Continuous variables in which differences in
means were considered statistically significant are shown in Table 2. When categories of the
variables were not equally distributed in farm types or areas, the categorical variables were
included in Table 3. The results also showed differences by farm size and by geographical area
in distances: from store to urban area, swimming water, water courses, and spread plot (P<0.05).
As few differences were observed in farm management between the two smallest types of
facilities (66% of the variables in the whole poll had equal values), that pair of subgroups were
joined together. In the same way, farm management was found equal for the two biggest types of
farms (69% of the variables in the whole poll had significantly similar values), so two new
groups were formed by size, which presented different farm managament:
x Small-scale farms: producing less than 100,000 kg of milk per year (52% of the 141 farms).
x Middle-scale farms: producing 100,000-500,000 kg of milk per year (48% of the farms).
When variables cited were tested for the minimum requirements for clustering, only eleven
fulfilled them. Non-hierarchical clustering allowed step-wise selection of three variables: farm
size, land spreading method and type of manure stores. Seeing that spreader type plays a key role
in drawing clusters, each dairy farm (n=137) was defined as one of the following categories:
x Only their own manure tanker (liquid manure spreaders) was used for land spreading,
including 47% of the small- to middle-scale dairies.
x Manure was distributed only with hired manure tanker, coming to 22% of the farms.
x Other methods were employed for land spreading (31%): manual spreading, pumping
followed by hose irrigation and solid manure spreaders. This group included cases which used
other methods in combination with own or hired spreaders (23% of the farms).
In the same way, the type of waste stores was labeled and the wide range of manure stores used
in each farm (n=137) led us to differentiate the waste stores as:
x Covered stores: farms presenting at least one covered store (41%), as it was the store more
recently built and moreover, it represented the major part of the available capacity, which also
received the manure of the milk producing cows. Three farms were also storing in piles (2%)
and other three farms combined covered with uncovered tanks (2%).
x Uncovered stores: farms that had constructed all their waste stores uncovered (47%) or farms
that preserved all their stores uncovered and also used piles, given that the main part of the
manure volume was stored in uncovered tanks (3%).
x Piles: farms that accumulated all their manure in piles (8%).
x Daily land spreading: one farm did not store the manure, as it was daily spread on land (1%).
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

Table 4. No. of cases assigned to the conglomerates formed by means of non-hierarchical,


hierarchical and two-level clustering techniques from 141 polls.
Conglomerate Non-hierarchical Hierarchical TwoStep
clustering clustering clustering
1 26 26 29
2 36 36 31
3 43 39 36
4 26 26 31
Total no. of cases 131 127 127
Atypical cases 10 14 14

These profiles were used to develop final clusters of relatedness. A conglomerate of cases may
be defined as the group of farms placed in a certain manure management. The groups defined
were the same for the three types of clustering analysis and the change of position or row of each
farm in the database was not affecting the clusters created. The number of cases classified in
each conglomerate is shown in Table 4 for the three types of clustering analysis.
When non-hierarchical analysis of conglomerates was achieved, the final cluster centers of
each group presented the characteristics shown in Table 5. The categories of manure stores were
grouped in covered stores in one hand and uncovered stores, piles and daily land spreading on
the other hand. The cluster 3 included all the farms in the poll study which spread manure by
other methods, apart from tankers. Although the major part of the farms in that cluster stored
manure in piles or uncovered tanks (63% in the conglomerate), some used covered tanks (35% in
the conglomerate) and one farm proceeded with daily land spreading (2% in the conglomerate).
Only five farms presented properties that could not be inserted in any of the cluster defined,
as they managed a milk quota of 50,000 to 100,000 kg per year, hired the spreader but owned
covered stores. Other five farms were not included in this type of clustering, given that one of
them had missing data and four farms managed more than 500,000 kg of milk per year.
Final hierarchical clustering grouped farms in 126 stages, giving a conglomeration coefficient
of 4.286. The farms initially discarded were the same as in non-hierarchical analysis. Four more
units more were found atypical given that they presented characteristics of modern facilities
(they managed more than 100,000 kg of milk per year, used their own spreader and had covered
stores), but still spread part of the manure by other methods (manual spreading, pumping or solid
spreader). The profile of the final conglomerates generated by the hierarchical method is the
same as in Table 5, except that the lack of the four atypical farms retired from the third cluster,
changed the ratio of farms in each category of manure management. So, in the cluster 3, small-
scale farms rised to 92% and the farms storing in piles, uncovered stores or by daily spreading
slightly increased, while the farms using covered stores were reduced to 28%. It is clear that the
conglomerate 3 was better defined after discarding the four atypical farms, when this type of
clustering was applied, defining a traditional manure management.
When two-level clustering was achieved calculating the number of conglomerates
automatically, seven clusters were created. The comparison of these seven clusters with the ones
designed by non-hierarchical or hierarchical clustering, showed that the new clusters were a
combination of the formerly created ones, and their definition was not considered logical
because they represented a mixed complex of the farms visited. When four clusters were
requested in two-level approach, the profile designed was very similar to the ones described by
non-hierarchical or hierarchical clustering methods, although seven farms were assigned to a
different cluster.
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

Table 5. Final cluster centers of conglomerates defined by non-hierarchical analysis. In brackets,


percentage of cases included in the conglomerate that fulfilled the condition cited.
Cluster Milk production1 Land spreading2 Manure store
1 Small (100%) Hired tanker (61%) Pile (0%) or uncovered (100%)
2 Middle (92%) Own tanker (94%) Covered (100%)
3 Small (84%) Other (100%) Pile (23%) or uncovered (40%)
4 Middle (100%) Own tanker (77%) Pile (4%) or uncovered (96%)

While the milk quota gained importance, (defining each cluster 100%), the method of land
spreading lost slight importance as attribute in cluster 3 (other method, 92%) and 4 (own tanker,
64%), although gained slight importance when own spreader was selected for designing the
cluster 2 (100%). The manure store, although being an important variable, also presented smaller
percentages of uncovered stores which characterized the cluster 4 (87%).
The outcome of the study was manure management cluster membership of dairy farms which
were polled (see Figure 1). The three types of conglomerate analysis gave the same classification
for 95% of the farms, so coherence of the four groups defined was ascertained. From top to
bottom the formed hierarchical clusters were identified as traditional (cluster 3), intermediate
with hired spreader (cluster 1), intermediate with own spreader (cluster 4) and modern facilities
and habits (cluster 2).
In the three types of clustering, the traditional group had a greater internal variance than the
others did. In the mountainous area or in inland mid-valleys, young cattle was managed in
livestock cabins in high areas, where manure was piled, while the rest of the herd was maintaned
in sheds with uncovered stores (4 farms) or covered stores (1 farm). The cluster 3 included the
farms which used both covered and uncovered stores (3 cases) and the case which land spread
daily. The intermediate and modern groups were very homogeneous, regarding to their high
representativity and the low variation between the different types of clustering performed.

Table 6. Continuous (mean±SEM) or categorical variables (% of farms), by hierarchical clusters.


1
: kg/year. 2: tank volume/tanker volume. 3: stocking rate >2 head/ha. 4: grazing • 4
h/day, at least in spring and summer. 5: complete emptying of the tank when
spreading. 6: in the tank. Values with different superscripts are significantly different.
Continuous First cluster Second cluster Third cluster Fourth cluster
Milk1 50,111±4,747a 228,760±20,236b 52,571±6,525a 191,381±18,160b
Milk1/ha 5,942±856a 13,187±2,163b 5,282±872a 12,903±2,009b
Heads 17±2a 49±4b 19±1a 44±5b
1 a b c
Milk /head 7,089±246 8,214±307 4,804±362 8,118±205b
Tank (m3) 104±21a 401±52b 126±20a 327±61b
2 a b a
Tank/tanker 37±7 124±19 49±10 138±45b
Categorical First cluster Second cluster Third cluster Fourth cluster
Stocking3 15 32 21 32
4
Grazing 19 24 37 21
Pen 20 40 10 30
5
Complete 20 13 41 26
Agitator6 7 41 19 33
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

0 5 10 15 20 25
Case ß ØØØØØØØØØÚØØØØØØØØØÚØØØØØØØØØÚØØØØØØØØØÚØØØØØØØØØà
140 ØÞ
141 Øà
005 Øà
137 Øà
139 Øà
133 Øà
134 Øà
131 Øà
132 Øà
116 Øà
130 Øà
088 Øà
094 Øà
084 Øà
085 Øà
077 Øà
078 Øà
070 Øà
071 Øà
068 Øà
069 Øà
063 Øà
065 Øà
050 ØÚØØØØØØØØØÞ6
055 Øà ß ØÞ
013 ØÝ ÙÙ
034 Ø8ØØØØØØØØØÝ Ù
076 ØÝ Ù
083 ØÞ Ù
087 Øà Ù 7
011 Øà ß ØØØØØØØØØØØØØØØØØØØØØØØØØØØØØØØØØØØÞ
064 Øà Ù Ù
066 Øà Ù Ù
061 Øà Ù Ù
062 Øà Ù Ù
052 ØÚØØØØØØØØØÞ Ù Ù
057 Øà ß ØÝ Ù
051 ØÝ Ù6 Ù
081 ØØØØØØØØØØØÝ Ù
115 ØÞ Ù
135 Øà Ù
018 Øà Ù
105 Øà Ù
107 Øà Ù
045 Øà Ù
049 Øà 25
Ù
032 ØÚØØØØØØØØØÞ Ù
033 Øà Ù Ù
028 ØÝ Ù Ù
125 ØÞ Ù Ù
128 Øà Ù Ù
003 Øà Ù Ù
122 Øà Ù Ù
124 Øà Ù6 Ù
110 Øà ß ØØØØØØØØØØØØØÞ Ù
118 Øà Ù Ù Ù
096 Øà Ù Ù Ù
108 Øà Ù Ù Ù
035 Øà Ù Ù Ù
037 Øà Ù Ù Ù
024 Øà Ù Ù Ù
026 Øà Ù Ù Ù
016 ØÚØØØØØØØØØÝ Ù Ù
022 Øà Ù Ù
010 ØÝ Ù Ù
092 ØÞ Ù Ù
111 Øà Ù Ù
017 Øà Ù Ù
021 ØÚØØØØØØØØØÞ Ù Ù
042 Øà Ù Ù Ù
019 ØÝ Ù 12
Ù Ù
120 ØÞ Ù ß ØØØØØØØØØØØØØØØØØØØØØØØÝ
127 Øà Ù Ù
004 Øà Ù Ù
109 Øà Ù Ù
117 Øà Ù Ù
102 Øà Ù Ù
106 Øà Ù6 Ù
100 Øà ß ØØØÞ Ù
101 Øà Ù Ù Ù
095 Øà Ù Ù Ù
099 Øà Ù Ù Ù
048 Øà Ù Ù Ù
086 Øà Ù Ù Ù
038 Øà Ù Ù Ù
039 Øà Ù Ù Ù
027 Øà Ù Ù Ù
031 Øà Ù Ù Ù
014 ØÚØØØØØØØØØÝ Ù 8 Ù
020 Øà ß ØØØØØØØØØÝ
007 ØÝ Ù
025 ØÞ Ù
138 ØÚØØØØØØØØØØØÞ Ù
009 ØÝ ÙÙ
074 Ø8ØØØØØØØØØÞ Ù Ù
136 ØÝ ÙÙÙ
126 ØÞ ÙÙÙ
129 Øà ÙÙÙ
001 Øà ÙÙÙ
119 Øà ÙÙÙ
123 Øà Ù ß ØÝ
113 Øà ÙÙ 7
114 Øà ÙÙ
104 Øà ÙÙ
112 Øà ÙÙ
093 Øà ÙÙ
103 Øà ÙÙ
090 Øà ÙÙ
091 Øà ß ØÝ
082 Øà Ù6
089 Øà Ù
079 Øà Ù
080 Øà Ù
060 Øà Ù
067 Øà Ù
054 Øà Ù
056 Øà Ù
047 Øà Ù
053 Øà Ù
043 Øà Ù
044 Øà Ù
040 Øà Ù
041 Øà Ù
023 ØÚØØØØØØØØØÝ
036 Øà
002 Øà
012 ØÝ

Figure 1. Hierarchical cluster analysis. Dendrogram with rescaled distance of the cluster
combines.
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

Descriptive statistics of the manure managements were significantly different for hierarchical
and two-level approach method, suggesting that both are suitable for defining manure
managament (P<0.05). As the level of significance was slightly lower in the hierarchical method,
the variables were displayed for this type of clustering in Table 6. These analyses indicated that
the characteristics of the traditional and the intermediate with hired tanker clusters were similar,
but dissimilar to the group formed by the intermediate with own tanker and the modern clusters
(P<0.05). Only when milk production per head and year was compared, traditional and
intermediate with hired tanker were considered different.
When regarding factors delimiting production systems (Álvarez & Pérez, 2004), we
concluded that traditional cluster fitted in extensive production, intermediate with hired tanker
was between extensive and semi-intensive productions, while the rest corresponded to semi-
intensive productions.
The representation of the breeding systems was determined in each cluster by calculating
stocking rates, grazing rates and tank volume divided by the number of heads. The stocking rates
were equal for the manure managements defined by clusters (P<0.05). The sequence of complete
emptying of the tanks in the four manure managements was “traditional > intermediate with own
spreader > intermediate with hired spreader > modern farms”, but no differences were found
(P<0.05), as for the rest of the categorical variables.
The absence of grazing in 63-81% of the farms polled was also an important factor defining
the storage systems, although the volume of tank devoted for each head was not significantly
different between the four groups (mean±SEM: 7,31±0,58).
Given that the four groups defined by analyses of conglomerates represent the majority of the
polled farms, the study of dairy wastes will be easily ascertained sampling the four types.
Delineating manure management will help us in foreseeing the characteristics of manure
produced in each of these models and the usefulness of their wastes as soil amendment and
landfarming agent.
The size of traditional farms and the mean age of the owners are involving the progressive
dissapearance of this type of dairies or the change to extensive livestock breeding for meat
production. These impressions reinforce the need of the study of these traditional manure
managements, to check the environmental or health implications of their solid land spreading.
As the current tendency involves escalation of milk production, the rest of the farm types
would evolve to the concentration of the herd and so, understimated capacity of the stores is
expected in milk producing facilities. The environmental problems related to this
underestimation could be the rise in the number of spreading times and the decrease in the
grassland area in which it is spread, with the corresponding health risks which could be
associated. The increasing need of disposal ways for manure to relieve these problems should
drive research to the development of better manure treatments and reuses. Furthermore, the
organic amendment on agricultural plots could be complemented with disposal for landfarming
as external services. This should be one of the best disposal ways of reusing manure, as long as
the surpluses of farms are disposed and suitable manure type is chosen to improve landfarming.

4. CONCLUSIONS

The way the manure was used as soil amendment included the following: manual spreading,
pumping, solid spreading and slurry spreading (with hired spreader or with own spreader).
Nearly 20% of the farmers combined the use of farm machinery and the manual spreading or
pumping.
All three clustering techniques suggested that the 141 dairy farms in Cantabria (Northern
Spain) can be divided into four manure managements which corresponded to modern,
Sardinia 2007, Eleventh International Waste Management and Landfill Symposium

intermediate with own spreader, intermediate with hired spreader and traditional facilities and
habits, respectively. The factors defining these clusters were the size of the dairy, represented by
the milk quota in kg produced per year, the manure store type and the application system to the
land. The modern cluster was defined by a middle-scale farm (from 100,000 to 500,000 kg)
which owned a covered tank and spread manure with its own spreader. The intermediate with
own spreader group corresponded to a middle-scale dairy (from 100,000 to 500,000 kg) which
used an uncovered store and its own spreader. The intermediate with hired spreader group
corresponded to a small-scale dairy (less than 100,000 kg of milk production per year) using
uncovered tanks and hiring the spreader for manure spreading. The traditional cluster
characteristics included: small-scale production (less than 100,000 kg), uncovered tanks or piles
to store manure and the use of manual spreading, pumps or solid spreaders. Farm and breeding
managements were also different in the clusters defined.

ACKNOWLEDGEMENTS

The Authors are grateful with Dr Antonio Colino, Departamento de Matemática Aplicada y de
Ciencias de la Computación (Universidad de Cantabria), for his valuable help in the statistical
work.

REFERENCES

Álvarez, A. and J.A. Pérez (2004) El efecto de la intensificación sobre los costes de las
explotaciones lecheras. Frisona Española 143, 100-102.
Animal Production Service (2004). Database of dairy farms with milk quota updated on 3rd
February 2004 in Cantabria (Spain).
Bacher, J., K. Wenzig, M. Vogler (2004). SPSS TwoStep cluster – A first evaluation.
www.statisticalinnovations.com/products/twostep.pdf.
Pérez, C. (2004) Técnicas de análisis multivariante de datos. Aplicaciones con SPSS. Pearson
Prentice Hall, Madrid.
Reed, J.F. 3rd and D.B. Stark (1988). Robust alternatives to traditional analysis of variance:
Welch W, James JI, James JII, Brown-Forsythe BF. Comput. Methods Programs Biomed.
26(3), 233-7.
Rezanková, H., D. Húsek, V. Snásel (2004). Clustering as a tool for data mining.
www.uivt.cas.cz/semweb/download.php?file=04-09-Husek&type=doc.
Stockbreeding development (1995). Poll on environmental attitudes in farms. Stockbreeding,
Agriculture and Fishery Regional Ministry, County Council.
Schwartz, G. (1978) Estimating the dimension of a model. The annals of statistics, 6(2), 461-
464.
Tejero, I. (1998) Poll for farmers, in final report of the project “Wastewater treatment plant for
the stockbreeding industry in Cantabria”. Participants: Environmental Engineering Group
(University of Cantabria) and OCYASA, S.L. Funding body: County Council of Cantabria.

You might also like