Evaluating methods of estimating global radiation and vapor pressure deficit using a dense network of automatic weather stations in coastal Brazil

A dense network of automatic weather stations (AWS) in eastern Brazil provided the opportunity to test established relationships between global radiation (R-s), photosynthetically active radiation (R-p) and net radiation (R-n). We also examined the variation in vapor pressure deficit (VPD) across the region, and predictions of VPD from temperature data. Predictions of R-s using the MTCLIM package [Agric. For. Meteorol. 93 (1999) 211] accounted for 75% of the variance (r(2) = 0.75) in monthly mean measured values but only 62% of the variance in daily values. A procedure described by [Int. J. Bio-meteorol. 44 (2000) 204] was less accurate. Relationships between R-n and R-s gave lower intercept values (indicative of net long-wave fluxes) than expected. Data for a year gave a value of 0.43 for the ratio of R-p to R-s; instrumental problems prevented longer-term comparisons. VPD during daylight hours (VPDday) varied significantly between the northernmost weather station (at 17degrees26'S) and the most inland, at 17degrees55'S but at slightly higher altitude (66 m compared to 160 m). The r(2) values for the linear relationships between maximum and minimum temperatures and VPDday varied across the region, ranging from 0.52 to 0.79. Using a process-based forest production model (3-PG; [For. Ecol. Manage. 95 (1997) 209]) we show that differences in VPD can lead to considerable (28%) reductions in the yield of plantation eucalyptus. (C) 2003 Elsevier B.V. All rights reserved.