- Fayetteville, AR
- ResearchGate
- Github
- Google Scholar
- ORCID
-
Web of Science
Research Visualizations
Methods represented:
Project Framework
Project Workflow & Predictive Modeling Framework — Spatial GPP, Growing Season Length, and CH₄
Four data sources — eddy covariance ground truth, gridded environmental data, satellite reflectance, and rice land cover — integrated across two phases: Phase 1 site-scale CH₄/GPP estimation and validation; Phase 2 state-scale predictive modeling using Sentinel, Landsat, and MODIS vegetation indices alongside soil and agronomic datasets.
Conceptual Framework for Ecosystem Flux Scaling
Upscaling from site-level, high-frequency Eddy Covariance measurements of CH₄ and NEE to regional, state-scale GPP and growing season length maps. Predictors span climatological, plant status (EVI, NDVI), water status (LSWI), and agronomic variables.
Vegetation Phenological Stages
Maps key phenological terminology — SOS, GUD, POS, EOS — onto the plant growth curve from emergence through senescence. Applies to single cropping systems.
Phenology Modeling — Planting & Harvest Date Prediction
Planting & Harvest Date Distribution
Violin-boxplot composite showing the spread, median, and IQR of planting dates (mean DOY 120, N=978) and harvest dates (mean DOY 257, N=673) across Arkansas rice fields. Derived from USDA survey records.
Lag from kNDVI Peak to Planting & Harvest Dates
Histograms of the lag (days) between the kNDVI maximum and observed planting date (mean 88.6 d, SD 16.7) and harvest date (mean 51.5 d, SD 14.4), demonstrating the predictive signal embedded in satellite vegetation indices.
Lag between SOS and Planting Date vs. Growing Season Length
Three-panel scatterplot comparing growing season length (HD–PD) against the lag between planting date and three SOS definitions (derivative, threshold, UD methods). Pearson r = 0.43–0.60, indicating that earlier-detected green-up relative to planting is associated with longer growing seasons.
Climate Drivers of Planting & Harvest Date Model Error
Six-panel figure (leave-2-years-out cross-validation) showing how interannual climate anomalies (min air temp, min soil temp, VPD) relate to model RMSE for PD (panel A, r = –0.14 to –0.42) and HD (panel B, r = –0.56 to –0.72, p < 0.05). Warmer and drier years yield more accurate harvest date predictions.
Gross Primary Productivity
Relationship between Yield and GPP
Spatial relationship between rice yield and VPM-modeled GPP across Arkansas production areas.
Time Series of GPP — Three Methods
Temporal comparison of GPP estimates derived from three independent approaches, showing consistency and divergence across the growing season.
Interannual Trends of Yield & GPP
Multi-year trajectory of rice yield alongside GPP, highlighting co-variation patterns and anomalous seasons.
GPP Comparison — Rice Production Regions of Arkansas
Distribution of GPP across distinct rice-growing regions, revealing spatial variability in productivity potential.
Factors Affecting GPP by Cultivar
How cultivar variety and biophysical factors modulate GPP, compared against eddy covariance ground truth.
Temporal Dynamics of GPP — Multiple Models
Side-by-side temporal profiles of GPP predicted by different model sources, enabling cross-model comparison across the growing season.
LSWI Influence on GPP across Clay Content Levels
Three-panel figure: (1) histogram of clay content in Arkansas rice fields, (2) LSWI–GPP relationship across clay strata, and (3) growing frequency vs. LSWI residuals across clay content levels.
Light Use Efficiency & fAPAR
Light Use Efficiency vs. Eight Biophysical Factors
Eight-panel scatterplot matrix (colored by days after planting) showing LUE responses to VPD (r = –0.32), relative humidity (r = 0.54), DBSI (r = –0.73), AWEInsh (r = –0.65), IAVI (r = 0.82), air temperature (r = 0.54), evapotranspiration (r = 0.42), and cumulative GDD (r = 0.67). GAM smooths overlaid. IAVI and VPD emerge as dominant controls.
fAPAR Distributions — EVI, NDVI, and LAI Parameterizations
Three-panel histogram comparing fAPAR computed from EVI (panel A), NDVI (panel B), and LAI (panel C). Highlights the proportion of observations where fAPAR > 1 (gold, physically implausible) and LUE > 1 (pink), informing parameterization choices for the VPM.
fAPAR–LAI Relationship with GPP and Phenological Timing
Scatter plot of fAPAR against Leaf Area Index (m²/m²) fitted with the Beer-Lambert exponential model (fAPAR = 1 – e–K×LAI). Points colored by days after planting (yellow→teal) and sized by EC-measured GPP, showing canopy light interception dynamics across the growing season.
Light Use Efficiency Across Time
Seasonal and interannual pattern of LUE — a key driver of GPP in the VPM framework — relative to phenological stages.
Vegetation Indices & Remote Sensing
Relationship between GPP and Vegetation Indices
Dual-axis visualization linking GPP dynamics to satellite-derived vegetation indices (NDVI, EVI, kNDVI), with regression equations for each relationship.
Harmonic Fit of kNDVI — Field 30
Harmonic regression applied to the kNDVI time series of a single rice field, extracting phenological transition dates from a smooth seasonal curve.
Phenology & Statistical Distributions
Planting & Harvest Date — Violin, Lag, and Histogram
Multi-panel composite integrating violin plots, lag distributions, and histograms to characterize the full temporal spread of planting and harvest dates across the Arkansas rice dataset.
Pairwise Correlation Plot
Triangular correlation matrix among planting/harvest dates, meteorological indicators, and vegetation indices. Strong positive correlations (yellow) between SOS, EOS, and thermal variables (GDD, soil temperature); soil organic carbon shows limited co-variation. Built with ggplot2 in R.