Is pH increasing in the noncalcareous topsoils of France under agricultural management? A statistical framework to overcome the limitations of a soil test database

• Published in: Soil use and management Vol. 33; no. 3; pp. 460 - 470
• Main Authors: Saby, N. P. A., Swiderski, C., Lemercier, B., Walter, C., Louis, B. P., Eveillard, P., Arrouays, D., Goss, Michael
• Format: Journal Article
• Language: English
• Published: Bedfordshire Wiley Subscription Services, Inc 01.09.2017; Wiley
• Subjects: Acidic soils; Acidification; Agricultural management; Arable land; Calcareous soils; Change detection; Computer Science; Environmental Sciences; Fertility; Frameworks; Hydrogen ions; Life Sciences; Localization; Long-term changes; pH effects; Position (location); Quality assessment; Resampling; Soil; Soil acidification; Soil analysis; Soil chemistry; Soil fertility; Soil monitoring; Soil pH; Soil properties; Soil quality; soil test database; Soil testing; Statistical analysis; Statistical methods; Statistics
• ISSN: 0266-0032; 1475-2743
• DOI: 10.1111/sum.12369

Soil pH is one of the most common and important measurements used to assess soil quality and manage soil fertility. Soil acidification is a slow process that can have large consequences. Therefore, it is important to detect soil changes early. Using a French soil test database, we show that the soil pH increased in 36% of arable soils monitored from 1996 to 2010, which suggests that soils became less acidic across the country. Conversely, decreases in pH were rarely detected. To perform this assessment, we used a statistical framework to analyse 488 717 soil test results from the French national soil test database. We excluded calcareous soils due to their relatively stable pH. Two periods were analysed: 1996 to 2000 (period 1) and 2006 to 2010 (period 2). Then, we performed nonparametric Mann–Whitney U tests following a resampling step to account for the noncontrolled selection of spatial and temporal localization in the observations. Our statistical framework smoothed the effects of some limitations of the database related to the variable number of samples collected during both periods and uncertainties in sampling locations. Overall, this study demonstrates the potential of using this type of database and a relevant statistical framework to detect long‐term changes in soil properties.