5:15 PM - 7:15 PM
[ACG39-P05] Quantifying landscape heterogeneity around flux towers using CubeSat images with high spatiotemporal resolutions
Keywords:Remote Sensing, Landscape Heterogeneity, Flux Tower
Eddy covariance flux towers Provides probably the best estimates of ecosystem-level Gross Primary Production (GPP) at the ecosystem level. However, due to the uneven global distribution of flux towers, flux tower measurements are insufficient for estimating GPP at a global scale, necessitating the integration of satellite remote sensing data for global GPP monitoring. Nevertheless, the matching between satellite observation resolution and flux tower footprints, combined with landscape heterogeneity around flux towers, introduces uncertainties into GPP estimation. Since footprint range and shape vary dynamically due to factors such as wind direction, wind speed, and surface roughness, observations from sites with significant landscape heterogeneity often fail to fully represent the mean ecological characteristics of the grid cells they are located in. Landscape heterogeneity is prevalent across FLUXNET, leading to an overestimation of mean biome productivity. However, few studies have attempted to assess flux tower heterogeneity using high-resolution satellite imagery. The emergence of satellite constellations has provided new opportunities for quantifying landscape heterogeneity around flux towers with high spatiotemporal resolution observations.
This study utilizes high spatial and temporal resolution CubeSat observations (3 m, daily) to quantify landscape heterogeneity around flux towers in Japan. Since GPP cannot be directly measured, recent studies have demonstrated that NIRvP exhibits strong consistency with GPP across scales, making it a reliable proxy for photosynthesis. However, as footprint data from flux towers are not publicly available, this study simulates the daily variation of footprint extent. By comparing NIRvP (near-infrared reflectance of vegetation multiplied by incoming sunlight) within a 3 × 3 window centered on the flux tower and within the simulated footprint, we quantified landscape heterogeneity around flux towers using a full year 2020 from Japan Flux sites.
Results indicated that high-resolution NIRvP observations can better capture fine-scale heterogeneity, as the 3 m NIRvP map effectively represents landscape distribution, making it a valuable tool for quantifying heterogeneity around flux towers. Moreover, landscape heterogeneity is commonly observed across various vegetation types. In the Japan region, although many flux sites are necessary for accurately upscaling flux data, the number of sites with high homogeneity remains limited. Among the 33 flux sites analyzed, only 7 sites (21%) exhibited a mean absolute percentage error (MAPE) below 10%, indicating that the majority of sites are affected by significant landscape heterogeneity. Highlighting the importance of incorporating high-resolution satellite observations into GPP estimation models.
This study utilizes high spatial and temporal resolution CubeSat observations (3 m, daily) to quantify landscape heterogeneity around flux towers in Japan. Since GPP cannot be directly measured, recent studies have demonstrated that NIRvP exhibits strong consistency with GPP across scales, making it a reliable proxy for photosynthesis. However, as footprint data from flux towers are not publicly available, this study simulates the daily variation of footprint extent. By comparing NIRvP (near-infrared reflectance of vegetation multiplied by incoming sunlight) within a 3 × 3 window centered on the flux tower and within the simulated footprint, we quantified landscape heterogeneity around flux towers using a full year 2020 from Japan Flux sites.
Results indicated that high-resolution NIRvP observations can better capture fine-scale heterogeneity, as the 3 m NIRvP map effectively represents landscape distribution, making it a valuable tool for quantifying heterogeneity around flux towers. Moreover, landscape heterogeneity is commonly observed across various vegetation types. In the Japan region, although many flux sites are necessary for accurately upscaling flux data, the number of sites with high homogeneity remains limited. Among the 33 flux sites analyzed, only 7 sites (21%) exhibited a mean absolute percentage error (MAPE) below 10%, indicating that the majority of sites are affected by significant landscape heterogeneity. Highlighting the importance of incorporating high-resolution satellite observations into GPP estimation models.