Search

Predicting how forest carbon cycling will change in response to climate change and management depends on the collective knowledge from measurements across environmental gradients, ecosystem manipulations of global change factors, and mathematical models. Formally integrating these sources of knowledge through data assimilation, or model–data fusion, allows the use of past observations to constrain model parameters and estimate prediction uncertainty. Data assimilation (DA) focused on the regional scale has the opportunity to integrate data from both environmental gradients and experimental studies to constrain model parameters. Here, we introduce a hierarchical Bayesian DA approach (Data Assimilation to Predict Productivity for Ecosystems and Regions, DAPPER) that uses observations of carbon stocks, carbon fluxes, water fluxes, and vegetation dynamics from loblolly pine plantation ecosystems across the southeastern US to constrain parameters in a modified version of the Physiological Principles Predicting Growth (3-PG) forest growth model. The observations included major experiments that manipulated atmospheric carbon dioxide (CO2) concentration, water, and nutrients, along with nonexperimental surveys that spanned environmental gradients across an 8.6 × 105 km2 region. We optimized regionally representative posterior distributions for model parameters, which dependably predicted data from plots withheld from the data assimilation. While the mean bias in predictions of nutrient fertilization experiments, irrigation experiments, and CO2 enrichment experiments was low, future work needs to focus modifications to model structures that decrease the bias in predictions of drought experiments. Predictions of how growth responded to elevated CO2 strongly depended on whether ecosystem experiments were assimilated and whether the assimilated field plots in the CO2 study were allowed to have different mortality parameters than the other field plots in the region. We present predictions of stem biomass productivity under elevated CO2, decreased precipitation, and increased nutrient availability that include estimates of uncertainty for the southeastern US. Overall, we (1) demonstrated how three decades of research in southeastern US planted pine forests can be used to develop DA techniques that use multiple locations, multiple data streams, and multiple ecosystem experiment types to optimize parameters and (2) developed a tool for the development of future predictions of forest productivity for natural resource managers that leverage a rich dataset of integrated ecosystem observations across a region.

As a primary flux in the global water cycle, evapotranspiration (ET) connects hydrologic and biological processes and is directly affected by water and land management, land use change and climate variability. Satellite remote sensing provides an effective means for diagnosing ET patterns over heterogeneous landscapes; however, limitations on the spatial and temporal resolution of satellite data, combined with the effects of cloud contamination, constrain the amount of detail that a single satellite can provide. In this study, we describe an application of a multi-sensor ET data fusion system over a mixed forested/agricultural landscape in North Carolina, USA, during the growing season of 2013. The fusion system ingests ET estimates from the Two-Source Energy Balance Model (TSEB) applied to thermal infrared remote sensing retrievals of land surface temperature from multiple satellite platforms: hourly geostationary satellite data at 4 km resolution, daily 1 km imagery from the Moderate Resolution Imaging Spectroradiometer (MODIS) and biweekly Landsat thermal data sharpened to 30 m. These multiple ET data streams are combined using the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) to estimate daily ET at 30 m resolution to investigate seasonal water use behavior at the level of individual forest stands and land cover patches. A new method, also exploiting the STARFM algorithm, is used to fill gaps in the Landsat ET retrievals due to cloud cover and/or the scan-line corrector (SLC) failure on Landsat 7. The retrieved daily ET time series agree well with observations at two AmeriFlux eddy covariance flux tower sites in a managed pine plantation within the modeling domain: US-NC2 located in a mid-rotation (20-year-old) loblolly pine stand and US-NC3 located in a recently clear-cut and replanted field site. Root mean square errors (RMSEs) for NC2 and NC3 were 0.99 and 1.02 mm day−1, respectively, with mean absolute errors of approximately 29 % at the daily time step, 12 % at the monthly time step and 0.7 % over the full study period at the two flux tower sites. Analyses of water use patterns over the plantation indicate increasing seasonal ET with stand age for young to mid-rotation stands up to 20 years, but little dependence on age for older stands. An accounting of consumptive water use by major land cover classes representative of the modeling domain is presented, as well as relative partitioning of ET between evaporation (E) and transpiration (T) components obtained with the TSEB. The study provides new insights about the effects of management and land use change on water yield over forested landscapes.

Surface mining for coal has disturbed large land areas in the Appalachian Mountains. Better information on mined lands' ecosystem recovery status is necessary for effective environmental management in mining-impacted regions. Because record quality varies between state mining agencies and much mining occurred prior to widespread use of geospatial technologies, accurate maps of mining extents, durations, and land cover effects are often not available. Landsat data are well suited to mapping and characterizing land cover and forest recovery on former coal surface mines. Past mine reclamation techniques have often failed to restore premining forest vegetation but natural processes may enable native forests to re-establish on mined areas with time. However, the invasive species autumn olive (Elaeagnus umbellate) is proliferating widely on former coal surface mines, often inhibiting reestablishment of native forests. Autumn olive outcompetes native vegetation because it fixes atmospheric nitrogen and benefits from a longer growing season than native deciduous trees. This longer growing season, along with Landsat 8's high signal to noise ratio, has enabled species-level classification of autumn olive using multitemporal Landsat 8 data at accuracy levels usually only obtainable using higher spatial or spectral resolution sensors. We have used classification and regression tree (CART®) and support vector machine (SVM) to classify five counties in the coal mining region of Virginia for presence and absence of autumn olive. The best model found was a CART® model with 36 nodes which had an overall accuracy of 84% and kappa of 0.68. Autumn olive had conditional kappa of 0.65 and a producers and users accuracy of 86% and 83% respectively. The best SVM model used a second order polynomial kernel and had an overall accuracy of 77%, an overall kappa of 0.54 and a producers and users accuracy of 60% and 90% respectively.

Interest in data fusion, for remote-sensing applications, continues to grow due to the increasing importance of obtaining data in high resolution both spatially and temporally. Applications that will benefit from data fusion include ecosystem disturbance and recovery assessment, ecological forecasting, and others. This paper introduces a novel spatiotemporal fusion approach, the wavelet-based Spatiotemporal Adaptive Data Fusion Model (WSAD-FM). This new technique is motivated by the popular STARFM tool, which utilizes lower-resolution MODIS imagery to supplement Landsat scenes using a linear model. The novelty of WSAD-FM is twofold. First, unlike STARFM, this technique does not predict an entire new image in one linear step, but instead decomposes input images into separate "approximation" and "detail" parts. The different portions are fed into a prediction model that limits the effects of linear interpolation among images. Low-spatial-frequency components are predicted by a weighted mixture of MODIS images and low-spatial-frequency components of Landsat images that are neighbors in the temporal domain. Meanwhile, high-spatialfrequency components are predicted by a weighted average of high-spatial-frequency components of Landsat images alone. The second novelty is that the method has demonstrated good performance using only one input Landsat image and a pair of MODIS images. The technique has been tested using several Landsat and MODIS images for a study area from Central North Carolina (WRS-2 path/row 16/35 in Landsat and H/V11/5 in MODIS), acquired in 2001. NDVI images that were calculated from the study area were used as input to the algorithm. The technique was tested experimentally by predicting existing Landsat images, and we obtained R2 values in the range 0.70 to 0.92 for estimated Landsat images in the red band, and 0.62 to 0.89 for estimated NDVI images.