| José M. Paruelo, William K. Lauenroth, and Pablo A. Roset |
| Authors are assistant professor, IFEVA-Departamento de Ecología, Facultad de Agronomía. Universidad de Buenos Aires. Av. San Martín 4453 (1417) Buenos Aires. Argentina, Professor, Department of Rangeland Ecosystems Science and NREL. Colorado State University. Fort Collins, Colo. 80523 USA, and graduate student IFEVA-Departamento de Ecología, Facultad de Agronomía. Universidad de Buenos Aires. Av. San Martín 4453 (1417) Buenos Aires. Argentina. |
Abstract |
| We present a non-destructive, photographic method to estimate biomass in semiarid grasslands. Though the method needs to be calibrated, it allows for a dramatic increase in the number of samples compared with the clipping method. The method is based on a relationship between the percentage of "green pixels" in a digital image and green biomass. We identified "green pixels" as those satisfying the following condition: G/B > 1 and G/R > 1, where G, B and R are the intensities of a particular pixel in the green, blue, and red bands respectively. The percentage of green pixels of the image and green grass biomass showed a correlation of 0.87 (n = 36, p < 0.001) when data were pooled from 3 sample dates. The relationship was slightly curvilinear and a log transformation of green biomass yielded a better correlation (r = 0.91, n = 36, p < 0.001). The percentage of green pixels showed a lower correlation with total green biomass than with grass biomass (r = 0.59 for the linear model and 0.73 for the log transformed model). The relationship between the percentage of green pixels and either green grass or total green biomass changed during the growing season. Both the slope and the Y-intercept of the model differed significantly among dates. Correlation coefficients for different dates ranged between 0.76 and 0.95. |
| Key Words: ANPP estimation, non-destructive methods, semiarid grassland |