TABLE 4.

Percent identification success when single parameters were excluded during training of RBF networks (with architecture as in Table 2) to discriminate 34 plankton species

Species	None	% Identification with excluded parametersa:
		1	2	4	5	6	7	1–7
Chlorella salina	83.0	81.5	79.5	62.8	81.8	79.8	71.5	11.8
Chlorella vulgaris	93.2	93.5	93.0	91.8	93.5	94.0	92.8	79.5
Chrysochromulina camella	91.0	90.5	90.5	65.8	88.5	90.5	70.8	41.5
Cryptomonas baltica	98.5	98.2	98.5	98.5	92.0	96.2	98.8	71.2
Cryptomonas calceiformis	99.0	99.0	98.8	98.8	97.8	98.8	98.8	87.8
Dunaliella tertiolecta	94.0	92.2	93.8	94.8	94.2	91.2	95.2	92.8
Gymnodinium simplex	83.8	83.5	83.8	70.0	79.8	70.2	76.2	32.2
Halosphaera russellii	83.5	63.0	76.2	76.5	81.2	82.2	77.5	51.2
Heterocapsa triquetra	92.0	90.5	90.8	91.0	84.8	90.5	89.8	63.5
Microcystis aeruginosa	95.5	95.2	95.2	95.2	95.0	94.8	95.8	78.2
Oscillatoria sp.	88.8	90.8	89.2	88.5	89.0	89.5	88.8	77.2
Oscillatoria redeckii	96.5	97.0	97.2	97.0	96.8	96.5	96.2	69.5
Prymnesium parvum	95.2	93.8	94.0	94.0	93.8	84.8	93.5	49.8
Pseudopedinella sp.	64.0	63.0	63.0	44.2	65.8	48.0	47.0	23.0
Selenastrum capricornutum	77.2	77.8	76.5	65.0	77.0	79.8	51.8	34.5
Skeletonema costatum	86.5	85.5	84.8	78.8	82.5	75.8	78.2	21.8
Staurastrum sp.	96.2	95.5	95.5	96.0	96.2	96.5	95.8	81.0
Tetraselmis rubens	64.0	61.0	51.0	48.0	62.8	54.5	52.0	10.5

^aOnly parameters and species for which there were marked differences from the network trained on all 11 parameters have been included. Parameter numbers are as presented in Table 2.