This vignette demonstrates comprehensive agricultural analysis capabilities including USDA CDL integration, crop monitoring, stress detection, and yield assessment workflows.
Use enhanced NDVI calculation for multi-temporal crop monitoring:
# Enhanced NDVI with quality filtering
ndvi_enhanced <- calculate_ndvi_enhanced(
red_data = red_band,
nir_data = nir_band,
quality_filter = TRUE,
temporal_smoothing = FALSE, # Single date
verbose = TRUE
)
# Visualize enhanced NDVI
create_ndvi_map(
ndvi_data = ndvi_enhanced,
title = "Enhanced NDVI with Quality Filtering",
ndvi_classes = "none"
)
print("Enhanced NDVI Analysis:")
ndvi_values <- terra::values(ndvi_enhanced, mat = FALSE)
valid_ndvi <- ndvi_values[!is.na(ndvi_values)]
print(paste("Mean NDVI:", round(mean(valid_ndvi), 3)))
print(paste("NDVI range:", paste(round(range(valid_ndvi), 3), collapse = " - ")))
Calculate multiple vegetation indices for comprehensive crop assessment:
# Calculate multiple indices relevant to agriculture
agricultural_indices <- calculate_multiple_indices(
red = red_band,
nir = nir_band,
indices = c("NDVI", "SAVI", "MSAVI", "DVI", "RVI"),
output_stack = TRUE,
verbose = TRUE
)
print("Agricultural Indices Calculated:")
print(names(agricultural_indices))
# Visualize key indices
if (terra::nlyr(agricultural_indices) >= 2) {
# NDVI visualization
create_spatial_map(
spatial_data = agricultural_indices[[1]],
title = "NDVI for Agricultural Assessment",
color_scheme = "ndvi"
)
# SAVI visualization (soil-adjusted)
if ("SAVI" %in% names(agricultural_indices)) {
plot_raster_fast(
agricultural_indices[["SAVI"]],
title = "SAVI (Soil Adjusted Vegetation Index)",
color_scheme = "viridis"
)
}
}
Assess yield potential using vegetation indices:
# Extract yield analysis if available
if (!is.null(crop_analysis$analysis_results$yield_analysis)) {
yield_results <- crop_analysis$analysis_results$yield_analysis
print("Yield Potential Analysis:")
if (!"error" %in% names(yield_results)) {
print(paste("Composite yield index:", round(yield_results$composite_yield_index, 3)))
print(paste("Yield potential class:", yield_results$yield_potential_class))
print(paste("Classification confidence:", round(yield_results$classification_confidence, 3)))
print("Index contributions to yield assessment:")
for (idx in names(yield_results$index_contributions)) {
contrib <- yield_results$index_contributions[[idx]]
print(paste(" ", idx, "- Normalized:", round(contrib$mean_normalized, 3),
"Raw mean:", round(contrib$raw_mean, 3)))
}
}
}
# Create yield potential map visualization
yield_potential_raster <- agricultural_indices[[1]] # Use NDVI as proxy
terra::values(yield_potential_raster) <- ifelse(terra::values(yield_potential_raster) > 0.6, 3,
ifelse(terra::values(yield_potential_raster) > 0.4, 2, 1))
names(yield_potential_raster) <- "Yield_Potential"
plot_raster_fast(
yield_potential_raster,
title = "Yield Potential Classification",
color_scheme = "terrain"
)
Calculate key performance indicators for agricultural monitoring:
# Calculate comprehensive vegetation statistics
if (inherits(agricultural_indices, "SpatRaster")) {
veg_stats <- list()
for (i in 1:terra::nlyr(agricultural_indices)) {
index_name <- names(agricultural_indices)[i]
values <- terra::values(agricultural_indices[[i]], mat = FALSE)
valid_values <- values[!is.na(values)]
if (length(valid_values) > 0) {
veg_stats[[index_name]] <- list(
mean = mean(valid_values),
median = median(valid_values),
sd = sd(valid_values),
cv = sd(valid_values) / mean(valid_values),
min = min(valid_values),
max = max(valid_values),
q25 = quantile(valid_values, 0.25),
q75 = quantile(valid_values, 0.75)
)
}
}
print("Crop Performance Metrics:")
for (index in names(veg_stats)) {
stats <- veg_stats[[index]]
print(paste(index, "- Mean:", round(stats$mean, 3),
"CV:", round(stats$cv, 3),
"Range:", paste(round(c(stats$min, stats$max), 3), collapse = "-")))
}
}
Analyze individual fields or management zones:
# Create sample field boundaries
field_1 <- sf::st_polygon(list(matrix(c(
-83.5, 40.0, -83.3, 40.0, -83.3, 40.2, -83.5, 40.2, -83.5, 40.0
), ncol = 2, byrow = TRUE)))
field_2 <- sf::st_polygon(list(matrix(c(
-83.3, 40.0, -83.1, 40.0, -83.1, 40.2, -83.3, 40.2, -83.3, 40.0
), ncol = 2, byrow = TRUE)))
fields_sf <- sf::st_sf(
field_id = c("Field_1", "Field_2"),
crop_type = c("Corn", "Soybeans"),
geometry = sf::st_sfc(field_1, field_2, crs = 4326)
)
# Extract vegetation statistics by field using spatial join
field_analysis <- universal_spatial_join(
source_data = agricultural_indices,
target_data = fields_sf,
method = "zonal",
summary_function = "mean",
verbose = TRUE
)
print("Field-Level Analysis Results:")
if (inherits(field_analysis, "sf")) {
print(sf::st_drop_geometry(field_analysis))
}
Support precision agriculture through spatial variability analysis:
# Calculate coefficient of variation for management zones
if (inherits(agricultural_indices, "SpatRaster") && terra::nlyr(agricultural_indices) > 0) {
# Use NDVI for variability analysis
ndvi_layer <- agricultural_indices[[1]]
# Calculate local variability using focal statistics
local_cv <- terra::focal(ndvi_layer, w = matrix(1, 3, 3),
fun = function(x) sd(x, na.rm = TRUE) / mean(x, na.rm = TRUE))
names(local_cv) <- "Local_CV"
# Classify variability zones
variability_zones <- terra::classify(local_cv,
matrix(c(0, 0.1, 1, # Low variability
0.1, 0.2, 2, # Medium variability
0.2, 1, 3), # High variability
ncol = 3, byrow = TRUE))
names(variability_zones) <- "Variability_Zones"
# Visualize variability
plot_raster_fast(
local_cv,
title = "Spatial Variability (Coefficient of Variation)",
color_scheme = "plasma"
)
plot_raster_fast(
variability_zones,
title = "Management Zones (1=Low, 2=Medium, 3=High Variability)",
color_scheme = "categorical"
)
}
Use vegetation indices to detect crop stress and disease:
# Create stress detection workflow
stress_indices <- c("NDVI", "SAVI", "DVI")
# Calculate stress thresholds based on crop type
corn_thresholds <- list(
healthy = list(NDVI = c(0.6, 1.0), SAVI = c(0.4, 0.8)),
stressed = list(NDVI = c(0.3, 0.6), SAVI = c(0.2, 0.4)),
severely_stressed = list(NDVI = c(0.0, 0.3), SAVI = c(0.0, 0.2))
)
# Apply stress classification
if (inherits(agricultural_indices, "SpatRaster") && "NDVI" %in% names(agricultural_indices)) {
ndvi_values <- terra::values(agricultural_indices[["NDVI"]], mat = FALSE)
# Classify stress levels
stress_classification <- ifelse(ndvi_values >= 0.6, "Healthy",
ifelse(ndvi_values >= 0.3, "Moderate Stress", "Severe Stress"))
# Create stress map
stress_raster <- agricultural_indices[["NDVI"]]
stress_numeric <- ifelse(ndvi_values >= 0.6, 3,
ifelse(ndvi_values >= 0.3, 2, 1))
terra::values(stress_raster) <- stress_numeric
names(stress_raster) <- "Stress_Level"
plot_raster_fast(
stress_raster,
title = "Crop Stress Classification (1=Severe, 2=Moderate, 3=Healthy)",
color_scheme = "terrain"
)
# Calculate stress statistics
stress_table <- table(stress_classification)
stress_percent <- round(prop.table(stress_table) * 100, 1)
print("Crop Stress Distribution:")
for (level in names(stress_table)) {
print(paste(level, ":", stress_table[[level]], "pixels (", stress_percent[[level]], "%)"))
}
}
Create early warning systems for crop management:
# Define warning thresholds
warning_thresholds <- list(
drought_stress = 0.4, # NDVI below this indicates drought stress
disease_risk = 0.3, # NDVI below this indicates disease risk
optimal_growth = 0.7 # NDVI above this indicates optimal conditions
)
# Generate alerts
if (exists("ndvi_values")) {
alerts <- list()
# Calculate percentages in each category
drought_pixels <- sum(ndvi_values < warning_thresholds$drought_stress, na.rm = TRUE)
disease_pixels <- sum(ndvi_values < warning_thresholds$disease_risk, na.rm = TRUE)
optimal_pixels <- sum(ndvi_values > warning_thresholds$optimal_growth, na.rm = TRUE)
total_pixels <- sum(!is.na(ndvi_values))
alerts$drought_risk <- round((drought_pixels / total_pixels) * 100, 1)
alerts$disease_risk <- round((disease_pixels / total_pixels) * 100, 1)
alerts$optimal_conditions <- round((optimal_pixels / total_pixels) * 100, 1)
print("Agricultural Alert System:")
print(paste("Drought stress risk:", alerts$drought_risk, "% of field"))
print(paste("Disease risk:", alerts$disease_risk, "% of field"))
print(paste("Optimal conditions:", alerts$optimal_conditions, "% of field"))
# Generate recommendations
if (alerts$drought_risk > 20) {
print("RECOMMENDATION: Consider irrigation scheduling")
}
if (alerts$disease_risk > 10) {
print("RECOMMENDATION: Increase disease monitoring")
}
if (alerts$optimal_conditions > 70) {
print("STATUS: Crop conditions are generally favorable")
}
}
Track crop development throughout the growing season:
# Simulate time series data for growing season
growth_stages <- c("Planting", "V6", "V12", "R1", "R3", "R6")
ndvi_progression <- c(0.2, 0.4, 0.7, 0.8, 0.75, 0.6)
# Create time series visualization concept
seasonal_data <- data.frame(
Stage = growth_stages,
NDVI = ndvi_progression,
DOY = c(120, 150, 180, 200, 220, 260) # Day of year
)
print("Seasonal NDVI Progression:")
print(seasonal_data)
# Create conceptual growth curve
if (requireNamespace("ggplot2", quietly = TRUE)) {
library(ggplot2)
growth_plot <- ggplot(seasonal_data, aes(x = DOY, y = NDVI)) +
geom_line(color = "darkgreen", size = 1.2) +
geom_point(color = "red", size = 3) +
geom_text(aes(label = Stage), vjust = -0.5) +
labs(title = "Typical Corn NDVI Progression",
x = "Day of Year",
y = "NDVI Value") +
theme_minimal() +
ylim(0, 1)
print(growth_plot)
}
Use vegetation indices to optimize harvest timing:
# Define maturity indicators based on NDVI decline
maturity_thresholds <- list(
corn = list(
early_maturity = 0.7, # NDVI starts declining
harvest_ready = 0.5, # Optimal harvest window
post_harvest = 0.3 # Past optimal timing
),
soybeans = list(
early_maturity = 0.6,
harvest_ready = 0.4,
post_harvest = 0.25
)
)
# Calculate harvest readiness
if (exists("ndvi_values")) {
# Use corn thresholds for demonstration
thresholds <- maturity_thresholds$corn
harvest_assessment <- list(
early_maturity = sum(ndvi_values <= thresholds$early_maturity &
ndvi_values > thresholds$harvest_ready, na.rm = TRUE),
harvest_ready = sum(ndvi_values <= thresholds$harvest_ready &
ndvi_values > thresholds$post_harvest, na.rm = TRUE),
post_optimal = sum(ndvi_values <= thresholds$post_harvest, na.rm = TRUE),
still_growing = sum(ndvi_values > thresholds$early_maturity, na.rm = TRUE)
)
total_pixels <- sum(!is.na(ndvi_values))
print("Harvest Timing Assessment:")
for (stage in names(harvest_assessment)) {
percentage <- round((harvest_assessment[[stage]] / total_pixels) * 100, 1)
print(paste(stage, ":", harvest_assessment[[stage]], "pixels (", percentage, "%)"))
}
}
Use water indices to assess irrigation requirements:
# Calculate water-related indices
green_band <- red_band
terra::values(green_band) <- runif(2500, 0.1, 0.2)
names(green_band) <- "Green"
# Calculate NDWI for water stress detection
ndwi <- calculate_water_index(
green = green_band,
nir = nir_band,
index_type = "NDWI",
verbose = TRUE
)
# Irrigation needs based on NDWI
ndwi_values <- terra::values(ndwi, mat = FALSE)
valid_ndwi <- ndwi_values[!is.na(ndwi_values)]
if (length(valid_ndwi) > 0) {
irrigation_zones <- ifelse(valid_ndwi < -0.3, "High Need",
ifelse(valid_ndwi < -0.1, "Moderate Need", "Adequate"))
irrigation_table <- table(irrigation_zones)
irrigation_percent <- round(prop.table(irrigation_table) * 100, 1)
print("Irrigation Needs Assessment:")
for (zone in names(irrigation_table)) {
print(paste(zone, ":", irrigation_table[[zone]], "pixels (", irrigation_percent[[zone]], "%)"))
}
# Visualize irrigation zones
irrigation_raster <- ndwi
terra::values(irrigation_raster) <- as.numeric(as.factor(irrigation_zones))
names(irrigation_raster) <- "Irrigation_Needs"
plot_raster_fast(
irrigation_raster,
title = "Irrigation Needs Assessment (1=Adequate, 2=High, 3=Moderate)",
color_scheme = "water"
)
}
Analyze crop rotation patterns using multi-temporal CDL data:
# Simulate multi-year CDL data
create_rotation_analysis <- function() {
# Create sample rotation data
rotation_data <- data.frame(
year = rep(2021:2023, each = 4),
field = rep(c("Field_A", "Field_B", "Field_C", "Field_D"), 3),
crop = c(
"Corn", "Soybeans", "Corn", "Wheat", # 2021
"Soybeans", "Corn", "Soybeans", "Corn", # 2022
"Corn", "Soybeans", "Corn", "Soybeans" # 2023
),
yield = runif(12, 80, 200)
)
return(rotation_data)
}
rotation_analysis <- create_rotation_analysis()
print("Crop Rotation Analysis:")
print(rotation_analysis)
# Analyze rotation patterns
rotation_patterns <- list()
fields <- unique(rotation_analysis$field)
for (field in fields) {
field_data <- rotation_analysis[rotation_analysis$field == field, ]
rotation_patterns[[field]] <- paste(field_data$crop, collapse = " → ")
}
print("Rotation Patterns:")
for (field in names(rotation_patterns)) {
print(paste(field, ":", rotation_patterns[[field]]))
}
Calculate sustainability indicators:
# Calculate diversity index for crop rotation
calculate_crop_diversity <- function(rotation_data) {
diversity_scores <- list()
for (field in unique(rotation_data$field)) {
field_crops <- rotation_data$crop[rotation_data$field == field]
unique_crops <- length(unique(field_crops))
total_years <- length(field_crops)
# Simple diversity score (0-1, higher = more diverse)
diversity_scores[[field]] <- unique_crops / total_years
}
return(diversity_scores)
}
diversity_scores <- calculate_crop_diversity(rotation_analysis)
print("Crop Diversity Scores (Higher = More Diverse):")
for (field in names(diversity_scores)) {
print(paste(field, ":", round(diversity_scores[[field]], 2)))
}
# Sustainability recommendations
avg_diversity <- mean(unlist(diversity_scores))
print(paste("Average field diversity:", round(avg_diversity, 2)))
if (avg_diversity < 0.5) {
print("RECOMMENDATION: Consider more diverse crop rotations")
} else if (avg_diversity > 0.8) {
print("STATUS: Good crop diversity maintained")
}
Prepare analysis results for integration with farm management systems:
# Create farm-ready data export
create_farm_export <- function(analysis_results, field_boundaries) {
# Compile key metrics for farm management
farm_data <- list(
summary_statistics = list(),
recommendations = list(),
alerts = list()
)
# Extract key metrics
if (exists("veg_stats") && length(veg_stats) > 0) {
farm_data$summary_statistics <- veg_stats
}
# Generate management recommendations
if (exists("alerts")) {
if (alerts$drought_risk > 20) {
farm_data$recommendations <- c(farm_data$recommendations,
"Increase irrigation frequency")
}
if (alerts$disease_risk > 15) {
farm_data$recommendations <- c(farm_data$recommendations,
"Monitor for disease symptoms")
}
}
return(farm_data)
}
# Export field-level results
if (exists("field_analysis") && inherits(field_analysis, "sf")) {
field_summary <- sf::st_drop_geometry(field_analysis)
# Add coordinates for GPS guidance
field_centroids <- sf::st_centroid(field_analysis)
field_coords <- sf::st_coordinates(field_centroids)
field_summary$centroid_lon <- field_coords[, 1]
field_summary$centroid_lat <- field_coords[, 2]
print("Field Summary for Farm Management:")
print(field_summary)
}
Calculate economic indicators from vegetation analysis:
# Estimate economic value based on vegetation health
calculate_economic_metrics <- function(ndvi_data, crop_prices) {
if (!exists("ndvi_values")) return(NULL)
# Simplified yield prediction based on NDVI
# (In practice, would use crop-specific models)
predicted_yield <- ifelse(ndvi_values > 0.7, "High",
ifelse(ndvi_values > 0.5, "Medium", "Low"))
yield_table <- table(predicted_yield)
# Economic projections (simplified)
economic_zones <- list(
high_yield = sum(predicted_yield == "High", na.rm = TRUE),
medium_yield = sum(predicted_yield == "Medium", na.rm = TRUE),
low_yield = sum(predicted_yield == "Low", na.rm = TRUE)
)
return(economic_zones)
}
# Calculate economic zones
if (exists("ndvi_values")) {
economic_analysis <- calculate_economic_metrics(ndvi_values,
list(corn = 5.50, soybeans = 13.00))
if (!is.null(economic_analysis)) {
total_pixels <- sum(unlist(economic_analysis))
print("Economic Potential Analysis:")
for (zone in names(economic_analysis)) {
pixels <- economic_analysis[[zone]]
percentage <- round((pixels / total_pixels) * 100, 1)
print(paste(zone, ":", pixels, "pixels (", percentage, "%)"))
}
}
}
Implement quality assurance for agricultural monitoring:
# Vegetation index quality assessment
quality_check <- function(vegetation_indices) {
qc_results <- list()
if (inherits(vegetation_indices, "SpatRaster")) {
for (i in 1:terra::nlyr(vegetation_indices)) {
index_name <- names(vegetation_indices)[i]
values <- terra::values(vegetation_indices[[i]], mat = FALSE)
qc_results[[index_name]] <- list(
total_pixels = length(values),
valid_pixels = sum(!is.na(values)),
coverage_percent = round((sum(!is.na(values)) / length(values)) * 100, 1),
value_range = range(values, na.rm = TRUE),
outliers = sum(values < -1 | values > 1, na.rm = TRUE) # For normalized indices
)
}
}
return(qc_results)
}
# Perform quality check
if (exists("agricultural_indices")) {
qc_results <- quality_check(agricultural_indices)
print("Data Quality Assessment:")
for (index in names(qc_results)) {
qc <- qc_results[[index]]
print(paste(index, "- Coverage:", qc$coverage_percent, "%, Outliers:", qc$outliers))
}
}
Support ground truth collection and validation:
# Create sampling points for field validation
create_validation_points <- function(field_boundary, n_points = 10) {
# Generate random points within field boundary
if (inherits(field_boundary, "sf")) {
sample_points <- sf::st_sample(field_boundary, n_points)
validation_sf <- sf::st_sf(
point_id = paste0("VP_", 1:length(sample_points)),
validation_type = "Ground_Truth",
geometry = sample_points
)
return(validation_sf)
}
return(NULL)
}
# Create validation points for first field
if (exists("fields_sf")) {
validation_points <- create_validation_points(fields_sf[1, ], n_points = 5)
if (!is.null(validation_points)) {
print("Validation Points Created:")
print(sf::st_drop_geometry(validation_points))
# Extract vegetation index values at validation points
if (exists("agricultural_indices")) {
validation_extracted <- universal_spatial_join(
source_data = validation_points,
target_data = agricultural_indices,
method = "extract",
verbose = FALSE
)
print("Extracted Values at Validation Points:")
validation_summary <- sf::st_drop_geometry(validation_extracted)
print(head(validation_summary))
}
}
}
Demonstrate a complete agricultural analysis workflow:
# Complete agricultural analysis workflow
run_agricultural_workflow <- function(spectral_data, cdl_data = NULL,
region_boundary = NULL) {
workflow_results <- list()
# Step 1: Calculate vegetation indices
message("Step 1: Calculating vegetation indices...")
indices <- calculate_multiple_indices(
spectral_data = spectral_data,
indices = c("NDVI", "SAVI", "DVI"),
output_stack = TRUE
)
workflow_results$vegetation_indices <- indices
# Step 2: Crop classification (if CDL available)
if (!is.null(cdl_data)) {
message("Step 2: Analyzing crop distribution...")
crop_mask <- create_crop_mask(cdl_data, "corn")
workflow_results$crop_mask <- crop_mask
}
# Step 3: Stress assessment
message("Step 3: Assessing crop stress...")
if (inherits(indices, "SpatRaster") && "NDVI" %in% names(indices)) {
ndvi_vals <- terra::values(indices[["NDVI"]], mat = FALSE)
stress_percent <- sum(ndvi_vals < 0.5, na.rm = TRUE) / sum(!is.na(ndvi_vals)) * 100
workflow_results$stress_assessment <- round(stress_percent, 1)
}
# Step 4: Generate recommendations
message("Step 4: Generating management recommendations...")
recommendations <- character()
if (!is.null(workflow_results$stress_assessment)) {
if (workflow_results$stress_assessment > 25) {
recommendations <- c(recommendations, "High stress detected - investigate causes")
}
if (workflow_results$stress_assessment < 10) {
recommendations <- c(recommendations, "Crop conditions appear favorable")
}
}
workflow_results$recommendations <- recommendations
return(workflow_results)
}
# Run the workflow
workflow_output <- run_agricultural_workflow(
spectral_data = spectral_stack
)
print("Complete Agricultural Workflow Results:")
print(paste("Stress assessment:", workflow_output$stress_assessment, "% of area"))
print("Recommendations:")
for (rec in workflow_output$recommendations) {
print(paste("-", rec))
}
GeoSpatialSuite provides comprehensive agricultural analysis through:
get_comprehensive_cdl_codes()
,
analyze_cdl_crops_dynamic()
analyze_crop_vegetation()
,
calculate_ndvi_enhanced()
calculate_water_index()
for irrigation planning# 1. Always validate your data
print("Data Validation Checklist:")
print("✓ Check coordinate reference systems")
print("✓ Verify date ranges match growing season")
print("✓ Validate vegetation index ranges")
print("✓ Confirm crop mask accuracy")
# 2. Use appropriate indices for your crop type
crop_index_recommendations <- list(
corn = c("NDVI", "EVI2", "SAVI", "DVI"),
soybeans = c("NDVI", "SAVI", "GNDVI", "DVI"),
wheat = c("NDVI", "SAVI", "MSAVI"),
general = c("NDVI", "SAVI", "DVI", "RVI")
)
print("Recommended indices by crop type:")
for (crop in names(crop_index_recommendations)) {
indices <- crop_index_recommendations[[crop]]
print(paste(crop, ":", paste(indices, collapse = ", ")))
}
# 3. Monitor throughout growing season
print("Seasonal Monitoring Schedule:")
print("- Planting: Establish baseline measurements")
print("- Early season: Monitor emergence and establishment")
print("- Mid-season: Peak growth assessment and stress detection")
print("- Late season: Maturity evaluation and harvest timing")
print("- Post-harvest: Residue analysis and planning")
# Precision agriculture workflow components
precision_ag_components <- list(
data_collection = c("Satellite imagery", "UAV surveys", "Ground sensors"),
analysis_methods = c("Vegetation indices", "Stress detection", "Yield mapping"),
decision_support = c("Variable rate applications", "Irrigation scheduling", "Harvest timing"),
validation = c("Ground truth sampling", "Yield monitoring", "Economic analysis")
)
print("Precision Agriculture Integration:")
for (component in names(precision_ag_components)) {
print(paste(component, ":", paste(precision_ag_components[[component]], collapse = ", ")))
}
GeoSpatialSuite’s agricultural applications provide:
The agricultural tools in GeoSpatialSuite provide everything needed for professional agricultural analysis and precision farming applications!
This work was developed by the GeoSpatialSuite team with contributions from: Olatunde D. Akanbi, Vibha Mandayam, Yinghui Wu, Jeffrey Yarus, Erika I. Barcelos, and Roger H. French.