Wetlands

, Volume 27, Issue 4, pp 1098–1111

P-sorption capacity estimation in southeastern USA wetland soils using visible/near infrared (VNIR) reflectance spectroscopy

  • Matthew J. Cohen
  • Jeremy Paris
  • Mark W. Clark
Article

DOI: 10.1672/0277-5212(2007)27[1098:PCEISU]2.0.CO;2

Cite this article as:
Cohen, M.J., Paris, J. & Clark, M.W. Wetlands (2007) 27: 1098. doi:10.1672/0277-5212(2007)27[1098:PCEISU]2.0.CO;2

Abstract

Phosphorus (P) is frequently the limiting nutrient in aquatic ecosystems, so wetland P attenuation is of particular landscape importance. Providing reliable knowledge about the capacity of wetlands to provide P sequestration is limited by knowledge of soil sorption capacities. We examined P-sorption in wetland soils using samples (n = 326) collected from 171 wetlands across three southeastern ecoregions, stratifying by land use intensity (reference vs. impacted), vegetation (forested vs. herbaceous), and hydrologic setting (riverine vs. non-riverine). Single-point isotherm values ranged from −73 to 990 mg P kg−1 (mean = 462.7 ± 295.2 mg P kg−1). Using a mixed-effects ANOVA, no significant P-sorption differences were observed for vegetation or condition, and only a moderate effect of hydrology (p = 0.01). We observed a strong ecoregion effect (Regions IX ≈ XIV > XII; p < 0.001) and a strong interaction between ecoregion and condition (p < 0.001). Site-level and within-site random effects were both significant (p < 0.001 and p = 0.04, respectively), though the latter were small. The overall model explained only 29% of observed variance, suggesting limited generality for prediction. Given increased sample density requirements for assessment of P-sorption, pedotransfer functions (PTF), which estimate hard-to-measure properties from more readily observable properties, may be useful. We developed and validated two PTF models by relating observed sorption with 1) biogeochemical P-sorption covariates (total P/C, water extractable P, oxalate extractable Fe/Al/P/Ca/Mg) and 2) visible/near-infrared (VNIR) diffuse reflectance spectra. Advantages of VNIR for predicting soil properties include low cost, high sample throughput, minimal sample preparation and reagent waste, and high analytical precision. Standard error of prediction (SEP), r2, and relative performance determinant (RPD) values were compared between PTF models for hold-out validation data. Models were of comparable utility; with SEP values of 144.1 vs. 157.2 mg P kg−1, r2 values of 0.61 vs. 0.69, and RPD values of 1.61 vs. 1.88 for the biogeochemical vs. VNIR models, respectively. Given other advantages of using VNIR spectra, it appears to be a useful tool for mapping and monitoring P sorption in wetland soils.

Key Words

near infrared reflectance spectroscopy (NIRS) numeric nutrient criteria phosphorus sorption capacity Southeast USA 

Copyright information

© Society of Wetland Scientists 2007

Authors and Affiliations

  • Matthew J. Cohen
    • 1
  • Jeremy Paris
    • 2
  • Mark W. Clark
    • 2
  1. 1.Forest Water Resources Laboratory School of Forest Resources and ConservationUniversity of FloridaGainesvilleUSA
  2. 2.Wetland Biogeochemistry Laboratory Soil and Water Science DepartmentUniversity of FloridaGainesvilleUSA

Personalised recommendations