Table 3.

Regression equations obtained by the MFP method in each domain (D) by using exploratory and 3-variable models.

D	Adj-R2	Variable	Functional form	Coefficient	Std Error	pvalue
T	0.66	Intercept		− 23.28	3.416	p < 0.001
		grNO3	log(grNO3/100)0.5	1.499	0.191	p < 0.001
		SSS	SSS/100	49.81	9.141	p < 0.001
		avrN2	(avrN2/10–4)2	0.032	0.012	p < 0.01
			(avrN2/10–4)3	− 0.002	0.001	p < 0.01
		sNO3	log(sNO3 + 0.1)	− 1.136	0.175	p < 0.001
			(sNO3 + 0.1)	0.783	0.134	p < 0.001
	0.27	Intercept		1.307	1.285
		grNO3	log(grNO3/100)	0.590	0.118	p < 0.001
M	0.63	Intercept		5.210	2.300	p < 0.05
		SST	SST/10	− 3.420	1.188	p < 0.05
		sCHL-a	sCHL-a	1.090	0.538	p < 0.1
G	0.56	Intercept		− 2.772	0.624	p < 0.001
		grNO3	grNO3/100	1.940	0.305	p < 0.001
		MLD	MLD/10	0.433	0.181	p < 0.05
	0.49	Intercept		− 1.702	0.468	p < 0.001
		grNO3	grNO3/100	1.641	0.299	p < 0.001
A	0.75	Intercept		5.410	0.842	p < 0.001
		SST	SST + 1.1	− 2.751	0.514	p < 0.001
GM	0.63	Intercept		2.148	0.759	p < 0.01
		grNO3	(grNO3/100)0.5	1.738	0.300	p < 0.001
		sNO3	sNO3 + 0.1	0.246	0.087	p < 0.01
		avrN2	(avrN2/10–4)	− 0.255	0.062	p < 0.001
		SST	SST/10	− 2.370	0.328	p < 0.001
		dmaxN2	(dmaxN2/100)2	1.482	0.491	p < 0.01
			(dmaxN2/100)2*log(dmax N2/100)	− 2.120	0.647	p < 0.01
		maxN2	log(maxN2/0.001)	0.909	0.272	p < 0.01
			log(maxN2/0.001)2	0.372	0.091	p < 0.001
	0.50	Intercept		2.851	0.576	p < 0.001
		SST	SST/10	− 2.471	0.243	p < 0.001
		grNO3	log(grNO3/100)0.5	1.636	0.274	p < 0.001

T indicates tropical and subtropical regions, M Mediterranean Sea, G Galician coastal upwelling, A Antarctic Peninsula, and GM global model built with the complete dataset except those collected in the Antarctic Peninsula. Exploratory models used all the predictors, and 3-variable models only used sea surface temperature (SST), surface chlorophyll-a (sCHL-a) or vertical nitrate gradient (grNO₃). Adj-R² is adjusted squared coefficient of determination, and Std Error standard error. The number of stations used to build each model is indicated in Table 1.