Nitrogen:  NH4+-N, NO3--N, NO2--N, and TN. TN sums the nitrogen in amm oniu m (NH4+), oxidized forms (Nox ) NO2- + NO3-), and aggregate organic forms. Initially, the wastewater had a high concentrat ion of TN mainly composed of NH4+-N with a very low conte nt of NO2--N and NO3--N. Amm oniu m ranged widely from 10 to 71 mg L-1 in the influent. After 48 h treat ment, the amm onium concentrat ions in the efflue nt were str ongly redu ced, by app roximately 96 ( 5% and were no longer measurable after 72 h processing by the plants. Figure 3B shows that the decrea se was slower in the control: after 72 h, 33% NH4+ was still prese nt in the wastewater.

NO2- concentrat ion values re mained very low (Figure 3C). The highest conte nt was observed after 48 h of processing (3.30 ( 0.25 mg L-1).

Parallel to the NH + decrea se, we observed the app eara nce of NO3- (Figure 3D). The conte nt incre ased from 0 to 30 mg L-1 in 72 h in the prese nce of plants. The formation of NO3-L-1 (data not shown). Dissolved oxygen conce ntrat ions were app roximately 6.5 ( 1 mg L-1 after 48 h and provide oxygen necessary to the conversion of amm oniu m to nitrite and to nitrate . In com parison, other studies have indicated a low efficie ncy of amm oniu m conversion to nitrite in wetland systems owing to limited oxygen tra nsfer capability (3).

In nitrogen re moval, the p H is a significant parameter. Princic et al. have shown that the optimal p H range for NH4+ conversion to nitrite is betwee n 5.8 and 8.5 (31) and for the nitrification between 6.5 and 8.5. In wetlands, p H values were betwee n 6 and 7 (5, 32). In our treat ment system, the p H was constant and the values were betwee n 7 and 8 (data not shown); hence, the nitrification was active. Moreover, at these p H values, Sanchez-Mo nedero et al. reporte d the nitrogen may be volatilized and be given  off  (33).

The tota l nitrogen decrea sed more quickly in the prese nce of Chrysanthem u m cinerariaefoliu m than in the control chann el (Figure 3A). The root s are used like a supp ort for the formation of the biofilm which converte d amm oniu m into nitrate , so the tota l nitrogen did not decrea se significantly. Like for TP, C0 for TN was not negligible versus C1 (Table 2) and was higher for str onger initial loa ds. The kTN variations could not have bee n conn ecte d to either the influe nt loa d nor the du ration of the experiment. Three  major  nitrogen re moval mechanisms have been identified: microbial deni- trification (34), volatilization, and plant nitrogen up take (35, 36).

TN re moval incre ased du ring the 6 months (Figure 4) and peaked at 85 ( 2% in July. As the plant developed, root volume incre ased, thu s enh ancing the filtering capacity of the root s and the needs for nitrogenous components for the plant (29). The values of nitrogen re moval rates obtained from May reached the Europea n standards per mitting release into eutro phically sensitive areas.

After 48 h of plant culture, 30 L of wastewater are redu ce by 9 L. This ph enomenon is generally reporte d in many macrophyte systems (36, 37). To character ize the evapo- tra nspiration, Ayaz and Saygin advised  that  the SO42- conce ntrat ion be followed in treate d wastewater (37). In our NFT system, the sulfate conce ntrat ions incre ases after 48 h of treat ment of 30% with Chrysanthem u m cinerariaefoliu m and re mains identical in the control chann el (data not shown). The conce ntrat ion of  TN and  TP  in  the  efflue nt  was app roximately 30% lower than the influe nt conce ntrat ion after 48 h of recirculation. The re moved mass of these elements was higher than that reflecte d by measured concentrations beca use of evapotranspiration. Plants growing

FIGURE5. PCAfor theplant system inApril 2000.Correlationcircle for parameters considered on first two  axes.

TABLE3. Total Plant BiomassandPyrethrin Productions in YoungFlowers Derived from Plant Cultivated with Nutrient Solutionandwith Wastewatera