Remote sensing of gypsy moth defoliation to assess variations in stream nitrogen concentrations

• Published in: Ecological applications Vol. 14; no. 2; pp. 504 - 516
• Main Authors: Townsend, P.A, Eshleman, K.N, Welcker, C
• Format: Journal Article
• Language: English
• Published: Ecological Society of America 01-04-2004
• Subjects: Appalachian Mountains; change vector analysis; Creeks; Defoliation; forest disturbance; Forest ecosystems; Forested watersheds; Freshwater; gypsy moth; Land cover; Landsat; Lymantriidae; Moths; nitrate; Nitrogen; Remote sensing; Streams; water quality; watershed; Watersheds; USA; USA, Pennsylvania; USA, Maryland; ANW, USA, Maryland
• ISSN: 1051-0761; 1939-5582
• DOI: 10.1890/02-5356

Disturbance is an important mechanism controlling dissolved nitrogen (N) leakage to receiving streams, rivers, lakes, and estuaries from forested watersheds. In oak forests of the eastern United States, defoliation by the gypsy moth caterpillar represents a recurring disturbance regime that can impact water quality. Following an upsurge in gypsy moth activity in the mid-Appalachian region in 2000-2001, we examined the relationship between defoliation and N export within the Fifteenmile Creek watershed in Maryland and Pennsylvania using intensive stream surveys and remotely sensed imagery. Concentrations of dissolved N species were determined six times during seasonal baseflow and once during stormflow conditions at the outlet of 35 randomly selected sub-watersheds between April 2001 and June 2002. Summer Landsat images were used to characterize forest disturbance within each of the sub-watersheds in 1999, 2000, and 2001. Quantitative differences in imagery between the defoliation years of 2000 and 2001 (and 1999, a non-defoliation year) derived from change vector analysis provided indices of forest change. Dissolved N concentrations in the 35 reaches were dominated by nitrate-N (average 61%, ranging 19-89% of total dissolved nitrogen), although dissolved organic nitrogen was present at measurable levels (0.03-0.16 mg/L). Percent non-forested land cover accounted for only 8-33% of the spatial variation in nitrate-N concentrations, while multivariate models that also included change vector statistics from satellite imagery (magnitude and angular direction of image change) accounted for 72-80% of the spatial variation in nitrate-N concentrations. Measurements of defoliation during both the current and (when appropriate) previous year were statistically significant and indicated the importance of defoliation for increasing N inputs to the system at unseasonably high rates as a consequence of the deposition of caterpillar frass. Jackknifed residuals were within 25% of the actual measurements, indicating that remote sensing models could be used to predict variations in N concentrations for ungauged watersheds in the region. Our results provide more evidence that the remote sensing approach is a valuable component in predicting spatially varying relationships between ecosystem status and water quality conditions at the regional scale, thereby allowing assessments of watershed disturbances as part of land management.