Geostatistical techniques applied to mapping limnological variables and quantify the uncertainty associated with estimates

Abstract

Abstract Aim: This study aimed to map the concentrations of limnological variables in a reservoir employing semivariogram geostatistical techniques and Kriging estimates for unsampled locations, as well as the uncertainty calculation associated with the estimates. Methods: We established twenty-seven points distributed in a regular mesh for sampling. Then it was determined the concentrations of chlorophyll-a, total nitrogen and total phosphorus. Subsequently, a spatial variability analysis was performed and the semivariogram function was modeled for all variables and the variographic mathematical models were established. The main geostatistical estimation technique was the ordinary Kriging. The work was developed with the estimate of a heavy grid points for each variables that formed the basis of the interpolated maps. Results: Through the semivariogram analysis was possible to identify the random component as not significant for the estimation process of chlorophyll-a, and as significant for total nitrogen and total phosphorus. Geostatistical maps were produced from the Kriging for each variable and the respective standard deviations of the estimates calculated. These measurements allowed us to map the concentrations of limnological variables throughout the reservoir. The calculation of standard deviations provided the quality of the estimates and, consequently, the reliability of the final product. Conclusions: The use of the Kriging statistical technique to estimate heavy mesh points associated with the error dispersion (standard deviation of the estimate), made it possible to make quality and reliable maps of the estimated variables. Concentrations of limnological variables in general were higher in the lacustrine zone and decreased towards the riverine zone. The chlorophyll-a and total nitrogen correlated comparing the grid generated by Kriging. Although the use of Kriging is more laborious compared to other interpolation methods, this technique is distinguished for its ability to minimize the variance of the estimate and provide the estimated value of the degree of uncertainty.