Selection and use of representative sites in a large neutron moisture meter network

Abstract

The assessment of the representivity of individual sites in a large neutron moisture meter (NMM) network, and the errors incurred by selecting subsets of various sizes, and using different methods to estimate the full set value from the subsets, is described. Soil water data collected over 13 months from the NMM network (240 sites, divided into eight strata) were analysed to determine the relationship between individual sites and their stratum means. Relationships were assessed using the following four methods: (1) root mean square of the discrepancy; (2) linear regression; (3) quadratic regression; (4) power curve regression. Sites best related to their stratum means were then included in various size subsets, and estimates based on these subsets were compared with the respective full stratum means using 24 months' data. The best estimates of stratum means were obtained by using quadratic regressions and c. eight sites per subset. Using this method, the root mean square discrepancy between estimates of profile water content and observed full stratum means averaged 2.1 mm when the total profile was estimated directly and 2.5 mm when the profile water content was computed as the sum of individual estimates for four arbitrary horizons within the profile. Daily readings of one group of selected subsets indicated the potential of frequent NMM readings at key sites to indicate short-term changes in the soil water content of an area. The performance of the selected subsets has been monitored for a further 12 months as an independent check on their representivity. The root mean square discrepancy of estimates during that period was 2.7 mm. Subsequently only the selected eight sites in each stratum of the field study are being read on a routine basis.