3.  METHODS

3.1 THE FIELD WORK, CHEMICAL ANALYSES AND SPECIES SCORING.

   The samples were token in the period of 21‑8‑'84 to 7‑9‑'84. The sampling sites were chosen on the basis of the vegetation types, aiming to cover as much as variation as possible.  By  means  of  aereal  photographs,  the  position  of  each  sampling  site  was located, in order to enable comparison in the future. At each sampling site, a relevee of the  vegetation  was  made.  The  size  of  the  relevee  depended  on  the  size  of  the homogeneous part of the vegetation of the sampling site. In flarks or hollows the size of the relevee was usually about 4m², but in spring areas it could be only 1m² . The abundance of the vegetation was expressed in a percentage scale. Species present in the same kind of vegetation nearby scored 1, species within the relevee that covered 2% or less always scored a value of 2. On each site a water sample for chemical analyses was token, in a glass bottle of 200 ml. As the samples would be analysed in Amsterdam, we had to take precautions to prevent the chemical composition of the samples from being changed by biological activity. Therefor we filtered them  in  a capsule with a micropore‑filter, put under pressure by a foot‑pomp. The samples were stored in a refrigerator at 4 °C. In order to make pH and conductivity analyses in the very evening, a 150 ml plastic bottle was filled with filtered water. If there was a relatively large variation of substrate for the desmids to grow on, e.g. different moss species or parts of the bottom not covered by plants, samples were token of these different substrates. On each site, attention was paid to features like presence of iron deposits, which could be seen by a reddish colour of the bottom, presence of iron bacteries  which  form  a  bluish oil‑like  film  on  the  water  surface,  water  movement, thickness of the sampled algal layer, depth of the sampled water and depth on which the  samples  were  token.  In  the  evening  of  the  sampling  day,  the  pH  and  the conductivity of the sampled sites were measured with field meters. 40%‑Formol was added to the algal samples, to reach a final concentration of about 4%.    Three months after sampling, the chemical analyses were carried out at the Hugo de Vries laboratory in Amsterdam.

• SO₄^2-, P.PO₄^- , P‑tot, N.NO₂, N.NO₃ and N.NH₄ were determined colorimetrically.
• Cl^-  potentiographically.
• Ca²⁺  , Mg²⁺and Fe by atomic absorption.
• Na⁺  , K⁺  with a flame spectrophotometer.
• Alkality titrimetric.

   Algal samples were counted by means of a Zeiss microscope. A magnification of 125x  was  used  to  search  the  slide,  and  species  were  usually  identified  at  a magnification of 500x. Drawings were made by means of a drawing tube at a 2000x magnification, or 1000x‑ 250x if the species were relatively large. Species scoring was done until the number of  100  specimens  was  reached.  Often  one  or  more  species occurred abundantly in a sample: up to 95% of the counted specimens could concern a single species. In order to avoid a strong reduction of the number of counted species, only the first 10 specimens of a species were counted for the total of 100 specimens at which the species scoring was stopped. However, the remaining specimens were still recorded, so that, in fact, the final number of cells counted exceded the 100 cells. When the total was reached, I examined the  slide  at  a  more  quick  rate  for  not‑counted species, which were noted as "+". Consequently the  larger‑sized  species  were  more likely to be noted in this way. Finally the numbers of specimens counted for each species were expressed in percentages. Samples very poor in desmids were counted as far as possible.

3.2 CLASSIFICATION

 The resulting 194 counting lists ('desmid‑relevees'), belonging to 164 sites (table 12, 13 ) were  processed  by  various  clustering  methods.  All  analyses  were  done  with  the program  package  CLUSTAN  (Wishart  1978).  The  amount  of  species  that  could  be processed is 200. As I counted more than 450 species, I had to reduce the number. Therefore I removed all species that occurred less than 8 times in the total  of  194 samples, and also species of which the identification caused problems. To avoid too much influence of species with high scores, the percentage scale was transformed to an octave scale (Gauch, 1977). As many countings comprised more than 100 specimens, I also transformed the scale of Gauch by multiplying it with 1.25 (table 1). Otherwise for example much species which scored a value of 2 would be classified with a value of 1, the same class of species that scored '+'.

Table 1.Transformation of the species scores to an octave scale (Gauch, 1977)

species score (x)         In %					octave value
0					0
0	<	x	<	0,4	1
0,4	<	x	<	0,8	2
0,8	<	x	<	1,7	3
1,7	<	x	<	3,4	4
3,4	<	x	<	6,8	5
6,8	<	x	<	13,6	6
13,6	<	x	<	27,1	7
27,1	<	x	<	54,2	8
54,2	<	x	<	100	9

The data were processed in three ways (see table 2):

•    Cluster analysis  with  hierarchic  fusion  :  At  first  each  single  'desmid  relevee'  is a cluster. Then, after calculation the two most resembling clusters are fused into a new cluster with intermediate values, Then again, the two most resembling clusters are fused, and this process is repeated until just one cluster remains. Similarities were calculated with Ward's method.
  
  
•    Clustering with a binary scale implies that in respect of the species just the presence or absence is noted. Also Ward's method is used for calculating the similarities.
  
•    Cluster analysis with the relocation method : The relevees are put at random in one of n clusters, in which n is ones own choice. Then, for each relevee the similarity with each cluster is calculated (Ward's method) and the relevee is put in the cluster to which it has the most similarity (relocation). When this process is done for all relevees, a second cycle starts, in which the procedure is done again. These cycles are repeated until no relocation occurs.

  In order determine ecological relationships between the individual desmid species, 'octave' and 'binary' clusterings were also carried out for the desmid species. Therefore the axis of the previous clusterings were exchanged. Again Ward's method was used for calculating the similarities.

Likewise, the macrophyte vegetation relevees were clustered.

An analysis of variance was carried out on the chemical parameters (table 5).

Table 2.  Survey of the clustering methods used.