Latitudinal Gradients
Strong latitudinal gradients can be seen in both the field data and the OBIS data (Fig. 2). For the field S2/m data, the relationship with latitude was stronger (ρ = 0.765, p <0.001) than it was in the OBIS S2/m data (ρ = 0.682,p < 0.001). When using SQS, no significant pattern was found in the field data; Simpson’s D returned a stronger gradient in the OBIS data (Table 1). When only cells centred on field sites were included, a much weaker pattern was observed using S2/m (ρ = 0.556, p < 0.05), a stronger pattern was observed using SQS, and no pattern was returned using Simpson’s D. Including observational data in the OBIS dataset consistently reduced the latitudinal signal, as measured by ρ, across all combinations of analytical methods and diversity metrics (Fig. 2, Table 1). Fisher’s alpha returned a significant signal in all datasets, but was much weaker in the field data (Table 1).
The spatial autoregressions show that field diversity is strongly predicted by a set of abiotic variables (R2 = 86.5% for S2/m: Table 2). Diversity is not as well predicted by abiotic variables when using OBIS datasets. Submodels for each dataset and metric show that temperature is the best consistent overall predictor of richness in each case (Table 2, Table S1-2), with nitrate and phosphate content also important for subsets of OBIS data.