Linking remote sensing and insect defoliation biology – A cross-system comparison

Abstract

The success of space-based forest defoliation monitoring beyond thematic classification hinges on the ability to link units of continuous foliar change with associated units of spectral or vegetation index change across different systems. To explore this, we used multi-temporal Landsat sensor data with respect to three defoliation metrics (shoot-based estimates (%), frass biomass (kg·ha−1), and proportion of foliar loss) across three different forest systems in the United States. While shoot-based defoliation was measured in the field, the later two defoliation metrics were estimated based on the insect phenology model, BioSIM. The systems include spruce budworm (SBW, Choristoneura fumiferana C). in heterogenous mixed forests, Minnesota, jack pine budworm (JPBW, Choristoneura pinus F). in homogenous pine forests, Wisconsin, and spongy moth (SGM, Lymantria dispar dispar L). in mixed broad-leaved forests, Maryland. Our generalized annual defoliation model for budworm was poor (R2adj = 0.35, RMSE = 23.70%) due to difficulty in detecting SBW defoliation in Minnesota. Thus, individual annual JPBW shoot-based defoliation models (R2adj = 0.27 to 0.75 and RMSE = 19.16% to 11.32%) outperformed the SBW models (R2adj = 0.08 to 0.27, RMSE = 23.18% to 34.13%). Cross-system differences between rates of change in ground-based metrics of defoliation with respect to observed change in spectral index values (dVI) were significant for shoot-based estimates of percent annual defoliation for SGM vs. JPBW (p-values <1.00 × 10−4) and SGM vs. SBW (p-values < 4.60 × 10−3). However, such comparison between SGM and JPBW were not significant when either frass biomass (kg·ha−1, p-value = 0.46) or proportion of foliage loss was used (p-value = 0.67). For a given unit of defoliation, the magnitude of vegetation index change, dVI, was significantly different between SGM and JPBW for each defoliation metric used (p-values <1.00 × 10−4), but not between SGM and SBW (p-value = 0.47) or between JPBW and SBW (p-value = 0.06) when shoot-based annual defoliation metrics were used. This study is first to report that, for a given quantity of defoliation, substantial differences in both the magnitude and direction of vegetation index change exists for different defoliator systems, which currently precludes the development of a viable, unified defoliation model. However, outcomes of this research provide an important step in our understanding of these defoliator systems that may enable specific scaling of continuous defoliation at landscape levels and across defoliator systems in the future.