End-to-end Nitrogen Nutrition Index pipeline from Sentinel-2 L2A

Combines the outputs of estimate_biophysical with a crop- specific critical-N dilution curve (crop_params_NNI) to produce a pixel-level NNI map and the derived agronomic zones (deficient / optimal / excessive) in a single call.

Usage

compute_NNI_from_S2(
  lai_rast,
  cm_rast,
  cnc_rast,
  cnc_layer = c("CNC_Cprot", "CNC_Cab"),
  crop,
  params = NULL,
  fvc = NULL,
  scl = NULL,
  scl_keep = c(4L, 5L, 6L, 7L),
  nni_thresholds = c(0.9, 1.1),
  return_intermediates = TRUE
)

Arguments

lai_rast

SpatRaster or path to the LAI layer produced by estimate_biophysical.

cm_rast

SpatRaster or path to the Cm layer.

cnc_rast

SpatRaster or path to the canopy N layer.

cnc_layer

Character. Either "CNC_Cprot" (default) or "CNC_Cab". Controls which conversion is applied.

crop

Character. Crop name, resolved via crop_params_NNI.

params

Optional list overriding the defaults returned by crop_params_NNI (e.g. for locally calibrated alpha_leaf or k_NChl).

fvc

A SpatRaster (or path) with fractional vegetation cover, optional. If provided the pipeline drops pixels with FVC < params$fvc_min.

scl

A SpatRaster (or path) with the Sentinel-2 SCL layer, optional. Pixels not in scl_keep are masked.

scl_keep

Integer vector of SCL classes to keep (default c(4, 5, 6, 7) = VEGETATION, NOT_VEGETATED, WATER, UNCLASSIFIED).

nni_thresholds

Numeric length-2 vector with the lower and upper NNI thresholds (default c(0.90, 1.10)).

return_intermediates

Logical. If TRUE the returned list also contains the intermediate W, N_total and N_actual rasters.

Value

A named list of SpatRaster layers:

• W (t DM ha\(^{-1}\)) – aboveground biomass

• N_actual (pct DM) – plant N concentration

• N_crit (pct DM) – critical N from dilution curve

• NNI – Nitrogen Nutrition Index (dimensionless)

• zones – integer raster (1 deficient / 2 optimal / 3 excessive)

• mask – logical raster of retained pixels

Returned items are NULL when return_intermediates = FALSE for any intermediate variable not requested.

Details

The pipeline implements the classical Lemaire & Gastal (1997) formulation: $$W_{leaf} = LAI \cdot C_m \cdot 10^4 \quad (\mathrm{g\ m^{-2}})$$ $$W = W_{leaf} / \alpha_{leaf} \quad (\mathrm{g\ m^{-2}})$$ $$N_{total} = k \cdot CNC\_Cab \quad \mathrm{or} \quad N_{total} = CNC\_Cprot$$ $$N_{actual}(\%) = 100 \cdot N_{total} / W$$ $$N_c(\%) = a \cdot (W_{tha})^{-b}$$ $$NNI = N_{actual} / N_c$$ where \(\alpha_{leaf}\) is the leaf-to-total aboveground allocation coefficient and \(k\) the N-to-chlorophyll ratio (both crop-specific, see crop_params_NNI). \(W_{tha}\) is W rescaled to t DM ha\(^{-1}\).

Two N-content pathways are supported:

• Protein path (cnc_layer = "CNC_Cprot", default). Uses the GPR retrieval trained on canopy N from protein content directly — the most internally consistent path, no extra coefficient is needed.

• Chlorophyll path (cnc_layer = "CNC_Cab"). Uses the canopy chlorophyll retrieval and multiplies it by a canopy-level N:Chl ratio \(k\) (default ~55-60, requires local calibration — see crop_params_NNI). Preferred only when CNC_Cab is the best-trained model for the target crop or when the calibration work has already been done.

Masking

Two masks are applied before the NNI computation:

• If an fvc raster is supplied, pixels with FVC < crop_params$fvc_min are removed (default threshold 0.50). Below this level the BRDF is contaminated by the soil background and the GPR retrievals degrade.

• Pixels with W below crop_params$w_min are masked because the critical-N dilution curve is not defined at the seedling stage (default 1 t DM ha\(^{-1}\)).

Optional Sentinel-2 SCL (Scene Classification Layer) classes can be dropped via scl_keep (vector of valid classes; defaults to the standard vegetation-safe classes 4, 5, 6, 7 = VEGETATION, NOT_VEGETATED, WATER, UNCLASSIFIED).

Zones

The NNI is discretised into three classes following Lemaire & Gastal (1997):

Zone	NNI range	Code
Deficient	NNI < 0.90	1
Optimal	0.90 <= NNI <= 1.10	2
Excessive	NNI > 1.10	3

Thresholds are user-configurable via the nni_thresholds argument.

References

Lemaire G, Gastal F. N uptake and distribution in plant canopies. In: Diagnosis of the Nitrogen Status in Crops, Springer, 1997.

Houles V, Guerif M, Mary B. Elaboration of a nitrogen nutrition indicator for winter wheat based on leaf area index and chlorophyll content. Eur J Agron 2007;27:1-11.

See also

estimate_biophysical, compute_NNI, diagnose_N_status, crop_params_NNI, variable_rate_N

Examples

if (FALSE) { # \dontrun{
maps <- estimate_biophysical("scene.tif", "out",
  variables = c("LAI", "Cm", "CNC_Cprot", "FVC"),
  model_dir = "models_json")
nni <- compute_NNI_from_S2(
  lai_rast = maps$LAI,
  cm_rast  = maps$Cm,
  cnc_rast = maps$CNC_Cprot,
  cnc_layer = "CNC_Cprot",
  crop     = "maize",
  fvc      = maps$FVC
)
terra::writeRaster(nni$NNI,   "NNI.tif",   overwrite = TRUE)
terra::writeRaster(nni$zones, "zones.tif", overwrite = TRUE)
} # }