Phosphorus (P) is an essential nutrient for crops production including corn. However, the excessive P application, tends to P accumulation at the soil surface under crops systems. This may contribute to increase water and groundwater pollution by surface runoff. To prevent this, an agri-environmental P index, (P/Al)_M3, was developed in Eastern Canada and USA. This index aims to estimate soil P saturation for accurate P fertilizer recommendations, while integrating agronomical aspects and environmental risks. Understanding spatial variability of P will improve the economic and rational use of P fertilizers, promote the profitability and sustainability of agricultural companies, while reducing P losses. To achieve this using the kriging interpolation approach, a high sampling density (more than 100 sampling points) is required. However, this is not economically practical for the agricultural producers. Other interpolation methods, such as the spline method, could be used to set up maps with more precision.

The objective of our study was to compare two methods of interpolation (kriging vs. spline) on the precision and the reliability of the P and (P/Al)_M3 maps obtained with a decrease of soil sampling density in two commercial corn fields. The kriging was the reference method compared to the spline method that used five difference sampling densities. A spatial correlation approach of the five sampling densities map with the kriging map was performed.

Two commercial fields (around 10 ha), located in the Province of Quebec (Canada), under corn-soybean rotation were used. The five sampling densities, referred as 100%, 80%, 60%, 40%, and 20% of the georeferenced sampling points were used to interpolate map with the spline methods.  Each density was then compared with the krigged maps done with 100% of the sampling points.

A composite soil sample of four cores was taken within a radius of 1m around each sampling point at the two soil depths (0–5 cm and 5–20 cm) using a 0.05 m-diameter Dutch auger. Thus, a total of 768 soil samples was collected following these five different soil sampling strategies. Samples were analyzed for available P and Aluminum (Al) using the Mehlich-3 method (M3) and the (P/Al)_M3 index was calculated. Data were analyzed using descriptive statistics, geostatistics, and geographic information system (GIS) tools. Descriptive statistics (means, coefficients of variation) were determined using the SAS software. Geostatistical analysis (parameters calibrations, model validations) were performed using the GS+ version 9 software. Spatial variability maps of (P/Al)_M3 index were generated using 1m X 1m block kriging and spline interpolations. Both interpolation methods were performed using GS+ and ArcGIS tools. Performance of these two interpolation methods was evaluated using three geostatistical criterias: standard error (SE), mean standard error (MSE), and root mean square error (RMSE). Future P prescription maps will be also proposed for a sustainable management of P in different agro-ecosystems studied.

464 words.