← Back to the catalog
Private

ECOSTRESS rock all axis 5a744ba8

ecostress · other

ECOSTRESS rock all axis 5a744ba8. v2.0 standardized NIRS package: 1 spectral source(s), 2 declared target(s). Auto-generated from dataset_card.json (verify before publication).

nirv2ecostress
🔒
Private dataset. Full metadata and metrics are shown, but the bytes are not redistributed here — exporting the data requires a Dataverse token. The identity card carries no spectra, only descriptive statistics.
4
samples
2,844
wavelengths
1
sources
2
targets
27
metadata
other
family

Dataset property explorer

Mean profile risk0.46
Highest axisArtefacts locaux · 1.00
Diagnostics8
Sources profiled1
ECOSTRESS rock all axis 5a744ba8 property profile0.250.50.751integritynoiseartefactsbaselinePCA outliersreferencerepeatabilitystructureECOSTRESS rock all axis 5a744ba8 profileintegrity: 0.00noise: 0.00artefacts: 1.00baseline: 1.00PCA outliers: 0.31reference: 0.98repeatability: 0.00structure: 0.40ECOSTRESS rock …0 center · 1 outer ring · outward = stronger anomaly / heterogeneity signal

Profile axes

Intégrité0.00
Artefacts locaux1.00
Bruit0.00
Outliers PCA0.31
Distance à la référence0.98
Répétabilité0.00
Baseline / forme1.00
Structure multi-régimes0.40
Diagnostic hypotheses00.250.50.751hypothesis scoreSplice / raccord détecteursSplice / raccord détecteurs: 0.700.70Erreur calibration / référenc…Erreur calibration / référence blanche: 0.670.67Fond différentFond différent: 0.590.59Différence de sonde / géométr…Différence de sonde / géométrie: 0.520.52Erreur interpolation / réécha…Erreur interpolation / rééchantillonnage: 0.500.50Signature VERA25-likeSignature VERA25-like: 0.500.50Mélange feuille + fondMélange feuille + fond: 0.430.43Spectre hors domaine valideSpectre hors domaine valide: 0.380.38
DiagnosticScoreForceSignauxInterprétation probable
Splice / raccord détecteursX0.70moyenneSpike rate 1.00, SNR non dégradé 1.00, RMS/SAM référence 0.98Rupture aux jonctions de détecteurs, calibration locale ou sonde différente.
Erreur calibration / référence blancheX0.67moyenneBaseline/mean/area 1.00, artefacts locaux 1.00, RMS/SAM référence 0.98Décalage systématique entre campagnes, instruments ou référence blanche.
Fond différentX0.59moyenneBaseline/mean/area 1.00, RMS/SAM référence 0.98, Mahalanobis / T2 0.31Effet systématique du support, blanc/noir, transflectance ou environnement de mesure.
Différence de sonde / géométrieX0.52moyenneBaseline/mean/area 1.00, RMS/SAM référence 0.98, Mahalanobis / T2 0.31Modification de l'illumination, collecte, angle ou distance sonde-échantillon.
Erreur interpolation / rééchantillonnageX0.50moyenneSpike rate 1.00, SNR normal/élevé 1.00, Noise RMS faible 1.00Artefacts numériques ou traitement spectral incorrect.
Signature VERA25-likeX0.50moyenneSpike rate 1.00, RMS/SAM référence 0.98, Jump rate 0.53Combinaison possible changement de sonde + splice, amplifiée par géométrie, fond ou calibration.
Mélange feuille + fondX0.43moyenneBaseline/mean/area 1.00, RMS/SAM référence 0.98, Mahalanobis / T2 0.31Couverture partielle du spot; contribution du fond ou du support.
Spectre hors domaine valideX0.38faibleRMS/SAM référence 0.98, Structure PCA 0.40, Mahalanobis / T2 0.31Variété, espèce, lot ou condition différente mais physiquement plausible.

Spectral sources

rock all

X · other · source instruments vary by sample
rock all spectra0204060051015q05-q95 envelopeq25-q75 envelopemedian spectrummedianq25–q75q05–q95wavelength / none14.011none — median 2.378 (q25–q75 1.961–2.464)13.293none — median 2.773 (q25–q75 2.166–3.445)12.614none — median 3.853 (q25–q75 3.431–4.315)12.028none — median 4.361 (q25–q75 3.838–4.719)11.469none — median 4.282 (q25–q75 4.131–4.408)10.983none — median 3.886 (q25–q75 3.467–4.336)10.516none — median 3.463 (q25–q75 3.202–3.828)10.106none — median 2.822 (q25–q75 2.743–3.005)9.7082none — median 3.081 (q25–q75 2.924–3.255)9.3577none — median 3.325 (q25–q75 3.071–3.406)9.0159none — median 3.334 (q25–q75 2.789–3.566)8.7129none — median 1.831 (q25–q75 1.443–2.228)8.4295none — median 2.056 (q25–q75 1.547–2.472)8.1512none — median 3.228 (q25–q75 3.039–3.27)7.9027none — median 5.871 (q25–q75 5.009–6.324)7.6575none — median 8.938 (q25–q75 7.782–9.577)7.4378none — median 11.59 (q25–q75 10.66–12.14)7.2203none — median 13.34 (q25–q75 11.92–14.86)7.0246none — median 14.59 (q25–q75 12.54–17.16)6.8302none — median 15.85 (q25–q75 13.36–19.01)6.6549none — median 15.81 (q25–q75 13.41–19.11)6.4802none — median 16.93 (q25–q75 15.26–19.06)6.3221none — median 18.03 (q25–q75 16.42–19.96)6.1716none — median 17.81 (q25–q75 15.71–20.48)6.0211none — median 17.72 (q25–q75 16.5–19.02)5.8844none — median 21.46 (q25–q75 20.12–22.11)5.7474none — median 23.59 (q25–q75 22.34–24.68)5.6227none — median 22.24 (q25–q75 21.48–24.45)5.4975none — median 24.93 (q25–q75 24.21–26.66)5.3833none — median 26.37 (q25–q75 25.14–28.15)5.2684none — median 27.3 (q25–q75 25.55–29.25)5.1635none — median 29.11 (q25–q75 26.91–30.42)5.0577none — median 30.26 (q25–q75 28.43–31.13)4.9609none — median 32.73 (q25–q75 30.88–33.38)4.8678none — median 35.34 (q25–q75 33.54–36.58)4.7736none — median 39.05 (q25–q75 36.99–40.1)4.6873none — median 42.85 (q25–q75 39.69–44.28)4.6none — median 45.82 (q25–q75 42.76–47.46)4.5197none — median 48.08 (q25–q75 45.08–49.93)4.4385none — median 49.55 (q25–q75 46.47–51.45)4.3638none — median 50.64 (q25–q75 47.51–52.48)4.288none — median 51.68 (q25–q75 48.49–53.44)4.2182none — median 52.53 (q25–q75 49.25–54.26)4.1506none — median 53.2 (q25–q75 49.83–54.9)4.082none — median 53.72 (q25–q75 50.25–55.42)4.0187none — median 54.01 (q25–q75 50.55–55.45)3.9544none — median 54.48 (q25–q75 50.93–55.55)3.8949none — median 54.69 (q25–q75 51.13–55.94)3.8344none — median 54.82 (q25–q75 51.3–55.93)3.7785none — median 54.85 (q25–q75 51.39–55.86)3.7216none — median 54.85 (q25–q75 51.44–55.54)3.6689none — median 54.67 (q25–q75 51.44–55.03)3.6152none — median 53.96 (q25–q75 50.87–54.41)3.5654none — median 52.96 (q25–q75 49.5–54.17)3.5171none — median 51.6 (q25–q75 47.78–53.65)3.4677none — median 50.44 (q25–q75 46.28–53.23)3.4219none — median 49.05 (q25–q75 44.84–52.26)3.3751none — median 48.31 (q25–q75 43.64–52.25)3.3317none — median 47.64 (q25–q75 42.71–52.13)3.2874none — median 46.36 (q25–q75 41.07–51.62)3.2462none — median 45.27 (q25–q75 39.62–51.03)3.2041none — median 44.51 (q25–q75 38.47–50.34)3.1649none — median 43.8 (q25–q75 37.38–49.73)3.1249none — median 42.95 (q25–q75 36.11–49.1)3.0876none — median 42.26 (q25–q75 35.15–48.53)3.0513none — median 41.59 (q25–q75 34.24–47.98)3.014none — median 40.86 (q25–q75 33.19–47.41)2.9794none — median 40.1 (q25–q75 32.13–46.92)2.9439none — median 39.55 (q25–q75 31.34–46.47)2.9108none — median 39.23 (q25–q75 30.75–46.29)2.8769none — median 39.15 (q25–q75 30.52–46.19)2.8453none — median 39.1 (q25–q75 30.33–46.26)2.8129none — median 39.42 (q25–q75 30.27–46.67)2.7827none — median 40.33 (q25–q75 31.14–47.78)2.7517none — median 41.71 (q25–q75 31.92–49.83)2.7228none — median 43.65 (q25–q75 32.64–53.12)2.6945none — median 48.12 (q25–q75 39.53–54.94)2.6654none — median 55.83 (q25–q75 51.93–57)2.6382none — median 55.63 (q25–q75 51.87–57.31)2.6103none — median 55.14 (q25–q75 51.33–57.17)2.5843none — median 54.61 (q25–q75 50.71–56.85)2.5575none — median 54.05 (q25–q75 50.16–56.47)2.5326none — median 53.42 (q25–q75 49.44–56.03)2.5068none — median 52.83 (q25–q75 48.84–55.61)2.4828none — median 52.3 (q25–q75 48.24–55.27)2.4581none — median 51.74 (q25–q75 47.62–54.86)2.435none — median 51.23 (q25–q75 47.05–54.51)2.4124none — median 50.82 (q25–q75 46.6–54.11)2.389none — median 50.41 (q25–q75 46.12–53.63)2.3672none — median 50.12 (q25–q75 45.78–53.55)2.3447none — median 49.85 (q25–q75 45.48–53.04)2.3237none — median 49.66 (q25–q75 45.27–52.86)2.302none — median 49.43 (q25–q75 45.05–53.01)2.2818none — median 49.29 (q25–q75 44.92–53.05)2.2609none — median 49.19 (q25–q75 44.81–52.86)2.2413none — median 49.13 (q25–q75 44.74–52.76)2.2211none — median 49.16 (q25–q75 44.71–52.76)2.2023none — median 49.17 (q25–q75 44.71–52.61)2.1837none — median 49.24 (q25–q75 44.77–52.57)2.1646none — median 49.3 (q25–q75 44.78–52.53)2.1466none — median 49.4 (q25–q75 44.86–52.47)2.1281none — median 49.47 (q25–q75 44.91–52.39)2.1108none — median 49.58 (q25–q75 44.94–52.37)2.0929none — median 49.72 (q25–q75 45.07–52.34)2.064none — median 49.51 (q25–q75 44.93–51.92)1.98none — median 47.77 (q25–q75 43.53–49.48)1.9none — median 46.9 (q25–q75 43.17–47.74)1.82none — median 46.68 (q25–q75 43.56–47.12)1.736none — median 45.23 (q25–q75 41.63–46.96)1.656none — median 43.56 (q25–q75 39.42–46.69)1.572none — median 41.62 (q25–q75 37.51–45.65)1.492none — median 40.22 (q25–q75 36.86–42.74)1.408none — median 38.6 (q25–q75 36.3–39.4)1.328none — median 36.36 (q25–q75 33.55–37.35)1.244none — median 33.46 (q25–q75 30.59–35.03)1.164none — median 31.68 (q25–q75 28.93–33.11)1.08none — median 27.45 (q25–q75 26.19–28.44)1none — median 26.24 (q25–q75 24.46–27.21)0.92none — median 26.9 (q25–q75 24.2–28.94)0.836none — median 30.61 (q25–q75 27.44–31.62)0.789none — median 32.97 (q25–q75 29.43–33.26)0.768none — median 33.56 (q25–q75 29.75–33.89)0.748none — median 33.4 (q25–q75 29.68–33.61)0.727none — median 33.76 (q25–q75 29.97–34.06)0.707none — median 33.4 (q25–q75 29.71–33.69)0.686none — median 33.25 (q25–q75 29.59–33.6)0.666none — median 32.96 (q25–q75 29.55–33.21)0.645none — median 32.93 (q25–q75 29.5–33.08)0.625none — median 32.71 (q25–q75 29.41–32.79)0.605none — median 32.15 (q25–q75 28.91–32.26)0.584none — median 31.59 (q25–q75 28.37–31.9)0.564none — median 31.43 (q25–q75 28–31.79)0.543none — median 30.22 (q25–q75 26.61–30.85)0.523none — median 29.04 (q25–q75 25.3–30.12)0.502none — median 27.72 (q25–q75 24.27–28.66)0.482none — median 26.24 (q25–q75 23.04–27.28)0.461none — median 24.73 (q25–q75 21.86–25.81)0.441none — median 24.41 (q25–q75 21.37–25.64)0.42none — median 23.33 (q25–q75 20.4–24.45)0.4none — median 21.71 (q25–q75 19.25–22.57)

Sampling

Wavelengths2,844
Axis range0.4–14.01 none
Mean spacing0.00479 none
Gridirregular
Observations4

Signal & quality

Value range0.472 – 59.3
Mean range1.67 – 53.6
Mean level34.99
Area266.2
PTP51.91
Noise RMS0.0048441
SNR7.2e+03
SNR dB8e+01 dB
Dynamic range51.9
Smoothness0.1294
Saturated0.0%
X-outliers0

Integrity & artefacts

NaN ratio0.00%
Inf count0
Zero ratio0.00%
Spike count1,544
Spike rate13.58%
Jump count60
Jump rate0.53%
Clip fraction0.02%

Shape & reference

Baseline slope-46.317
Curvature RMS0.11495
D1 RMS0.13385
RMS to mean5.381
RMS p956.9594
SAM to mean0.094686
SAM p950.17129
Affine offset p955.6202
Affine gain p95 Δ0.21651
Affine residual p956.1203
Xcorr lag p950

Outliers & repeatability

PCA Q p95/median2.2
Hotelling T2 p95/median1.4
Mahalanobis H p95/median1.2
Repeat groups0

Dimensionality (PCA)

Effective rank2.1
PCs → 95% var2
PCs → 99% var2
Top-10 cum. var100.0%
Computed metric scores 29worst 1.00
FamilleMétrique calculéeValeurScoreNiveauInterprétation datasetCauses typiquesCalcul / scoring
Intégrité des donnéesNaN ratiointegrity.nan_ratio0%0.00faibleSpectre completErreur acquisition/exportcount(isnan(X)) / X.sizealert = min(1, nan_ratio / 0.05)
Intégrité des donnéesInf countintegrity.inf_count00.00faibleNormalCalculs invalidescount(isinf(X))alert = min(1, inf_count / 1)
Intégrité des donnéesZero ratiointegrity.zero_ratio0%0.00faibleNormalExport, saturationcount(X == 0) / count(finite X)alert = min(1, zero_ratio / 0.05)
Amplitude globaleMean reflectanceamplitude.mean_reflectance34.9881.00fortValeur atypique: Trop clair / fond visible ou Trop sombreFond, géométriemean(X finite)alert reuses baseline/shape drift because absolute reflectance ranges are technology-dependent
Amplitude globaleArea under curveamplitude.area_under_curve266.171.00fortValeur atypique: Différence d'éclairement ou NormalDistance sondetrapezoid(mean_spectrum, spectral_axis)alert reuses baseline/shape drift because area scale depends on axis and units
Amplitude globalePeak-to-peak (PTP)amplitude.peak_to_peak51.9060.00faibleVariabilité forteSaturationmax(mean_spectrum) - min(mean_spectrum)alert increases when dynamic range is abnormally flat
Amplitude globaleVarianceamplitude.variance272.040.00faibleNormal ou hétérogèneMauvais contactvar(X finite)alert increases when variance/dynamic range is abnormally flat
BruitNoise RMSnoise.noise_rms0.00484410.00faibleStableLampe, détecteurmedian MAD(second derivative) * 1.4826 / sqrt(6)alert = noise_rms / signal_scale, saturated at 5%
BruitSNRnoise.snr7222.80.00faibleBon signalAcquisitionmean(abs(X)) / noise_rmsalert decreases with SNR dB; >=40 dB is treated as low alert
BruitBandwise SNRnoise.bandwise_snr_min28.4240.17faibleZone fiableDétecteurmin(abs(mean_spectrum) / local second-derivative noise)alert decreases with worst-band SNR dB; >=35 dB is treated as low alert
Artefacts locauxSpike countartefacts.spike_count1,5441.00fortArtefactsCosmic rays, splicecount robust outliers in second derivativealert follows spike_rate, saturated at 1%
Artefacts locauxSpike rateartefacts.spike_rate13.6%1.00fortSpectre suspectInterpolationspike_count / (n_samples * (n_features - 2))alert = min(1, spike_rate / 0.01)
Artefacts locauxJump countartefacts.jump_count600.53moyenRaccord détecteurSplicecount robust outliers in first derivativealert follows jump_rate, saturated at 1%
Artefacts locauxJump rateartefacts.jump_rate0.528%0.53moyenProblème spectralCalibrationjump_count / (n_samples * (n_features - 1))alert = min(1, jump_rate / 0.01)
Artefacts locauxClip fractionartefacts.clip_fraction0.0176%0.02faibleNormalDétecteur saturéfraction of finite cells equal to repeated min/max extremaalert = min(1, clip_fraction / 0.01)
Forme spectraleBaseline slopeshape.baseline_slope-46.3171.00fortDériveÉclairementlinear slope of mean_spectrum over normalized axisalert = abs(slope / signal_scale), saturated at 0.5
Forme spectraleCurvature RMSshape.curvature_rms0.114950.22faibleLisseFond, splicemedian RMS(second derivative per spectrum)alert = curvature_rms / signal_scale, saturated at 1%
Forme spectraleD1 RMSshape.d1_rms0.133850.05faiblePlatBiologie ou artefactmedian RMS(first derivative per spectrum)alert = d1_rms / signal_scale, saturated at 5%
Outliers multivariésPCA Q (SPE)outliers.pca_q_ratio2.22790.28faibleConformeArtefact, mélangep95(Q/SPE residual) / median(Q/SPE residual)alert = min(1, pca_q_ratio / 8)
Outliers multivariésHotelling T²outliers.hotelling_t2_ratio1.43580.18faibleCentralVariabilité naturellep95(Hotelling T2) / median(Hotelling T2)alert = min(1, hotelling_t2_ratio / 8)
Outliers multivariésMahalanobis Houtliers.mahalanobis_h_ratio1.22640.31faiblePopulation normaleDomaine différentp95(sqrt(T2)) / median(sqrt(T2))alert = min(1, mahalanobis_h_ratio / 4)
Comparaison à référenceRMS to mean spectrumreference.rms_to_mean_spectrum_p956.95940.54moyenSpectre différentDomain shiftp95 RMS distance to dataset mean spectrumalert = RMS_p95 / signal_scale, saturated at 25%
Comparaison à référenceSpectral Angle Mapper (SAM)reference.sam_to_mean_spectrum_p950.171290.49moyenForme différenteFond, géométriep95 spectral angle to dataset mean spectrumalert = min(1, SAM_p95 / 0.35 rad)
RépétabilitéRMS intra-IDrepeatability.rms_intra_id—0.00faibleStablePositionnementmedian RMS distance to repeated-sample centroidalert = RMS_intra_ID / signal_scale, saturated at 10%
RépétabilitéSAM intra-IDrepeatability.sam_intra_id—0.00faibleStableAcquisitionmedian SAM to repeated-sample centroidalert = min(1, SAM_intra_ID / 0.15 rad)
RépétabilitéCV intra-IDrepeatability.cv_intra_id—0.00faibleStableOpérateurmedian within-ID band CValert = min(1, CV_intra_ID / 0.25)
Structure du datasetPCA score densitystructure.pca_score_density0.00221920.40faibleHomogèneLots différents1 / median kNN distance in PCA score spacealert follows density_cv/profile structure complexity, not raw density alone
Structure du datasetLocal Outlier Factor (LOF)structure.local_outlier_factor_p951.20830.10faiblePopulation normaleCas raresp95 approximate LOF from PCA-score kNN distancesalert = min(1, max(0, LOF_p95 - 1) / 2)
Structure du datasetIsolation Forest scorestructure.isolation_forest_score_p950.518750.40faibleNormalDiverses causesp95 IsolationForest anomaly score on PCA scoresalert follows structure complexity; raw score is implementation-dependent
X PCA score plot-400-2000200400-400-2000200400PC1 -268 · PC2 -235.6PC1 343.4 · PC2 -147.2PC1 3.088 · PC2 213.9PC1 -78.54 · PC2 168.9PC1 (55.9%)PC2 (43.2%)4 scores
PCA explained variance0%25%50%75%100%PC1: 55.9% (cumulative 55.9%)1PC2: 43.2% (cumulative 99.1%)2PC3: 0.9% (cumulative 100.0%)3cumulative explained variancePC variancecumulativeprincipal component · cumulative (dashed)

Metric interpretation reference

Metric catalog 29
FamilleMétriqueCe qu’elle détecteForte valeur =Faible valeur =Causes typiquesCalcul / score
Intégrité des donnéesNaN ratioDonnées manquantesSpectre corrompuSpectre completErreur acquisition/exportcount(isnan(X)) / X.sizealert = min(1, nan_ratio / 0.05)
Intégrité des donnéesInf countValeurs infiniesCorruptionNormalCalculs invalidescount(isinf(X))alert = min(1, inf_count / 1)
Intégrité des donnéesZero ratioColonnes ou cellules nullesSpectre tronquéNormalExport, saturationcount(X == 0) / count(finite X)alert = min(1, zero_ratio / 0.05)
Amplitude globaleMean reflectanceNiveau moyenTrop clair / fond visibleTrop sombreFond, géométriemean(X finite)alert reuses baseline/shape drift because absolute reflectance ranges are technology-dependent
Amplitude globaleArea under curveIntensité globaleDifférence d'éclairementNormalDistance sondetrapezoid(mean_spectrum, spectral_axis)alert reuses baseline/shape drift because area scale depends on axis and units
Amplitude globalePeak-to-peak (PTP)DynamiqueVariabilité forteSpectre platSaturationmax(mean_spectrum) - min(mean_spectrum)alert increases when dynamic range is abnormally flat
Amplitude globaleVarianceVariabilité spectraleNormal ou hétérogèneSpectre platMauvais contactvar(X finite)alert increases when variance/dynamic range is abnormally flat
BruitNoise RMSBruit haute fréquenceBruitéStableLampe, détecteurmedian MAD(second derivative) * 1.4826 / sqrt(6)alert = noise_rms / signal_scale, saturated at 5%
BruitSNRQualité signalBon signalMauvais signalAcquisitionmean(abs(X)) / noise_rmsalert decreases with SNR dB; >=40 dB is treated as low alert
BruitBandwise SNRBruit localiséZone fiableZone problématiqueDétecteurmin(abs(mean_spectrum) / local second-derivative noise)alert decreases with worst-band SNR dB; >=35 dB is treated as low alert
Artefacts locauxSpike countPics étroitsArtefactsSpectre propreCosmic rays, splicecount robust outliers in second derivativealert follows spike_rate, saturated at 1%
Artefacts locauxSpike rateDensité de picsSpectre suspectNormalInterpolationspike_count / (n_samples * (n_features - 2))alert = min(1, spike_rate / 0.01)
Artefacts locauxJump countDiscontinuitésRaccord détecteurContinuSplicecount robust outliers in first derivativealert follows jump_rate, saturated at 1%
Artefacts locauxJump rateFréquence de sautsProblème spectralNormalCalibrationjump_count / (n_samples * (n_features - 1))alert = min(1, jump_rate / 0.01)
Artefacts locauxClip fractionSaturationClippingNormalDétecteur saturéfraction of finite cells equal to repeated min/max extremaalert = min(1, clip_fraction / 0.01)
Forme spectraleBaseline slopePente globaleDériveStableÉclairementlinear slope of mean_spectrum over normalized axisalert = abs(slope / signal_scale), saturated at 0.5
Forme spectraleCurvature RMSCourbureForme inhabituelleLisseFond, splicemedian RMS(second derivative per spectrum)alert = curvature_rms / signal_scale, saturated at 1%
Forme spectraleD1 RMSVariabilité localeSpectre structuréPlatBiologie ou artefactmedian RMS(first derivative per spectrum)alert = d1_rms / signal_scale, saturated at 5%
Outliers multivariésPCA Q (SPE)Non expliqué par PCASpectre atypiqueConformeArtefact, mélangep95(Q/SPE residual) / median(Q/SPE residual)alert = min(1, pca_q_ratio / 8)
Outliers multivariésHotelling T²Extrême dans PCAExtrême mais cohérentCentralVariabilité naturellep95(Hotelling T2) / median(Hotelling T2)alert = min(1, hotelling_t2_ratio / 8)
Outliers multivariésMahalanobis HDistance au nuageOutlier globalPopulation normaleDomaine différentp95(sqrt(T2)) / median(sqrt(T2))alert = min(1, mahalanobis_h_ratio / 4)
Comparaison à référenceRMS to mean spectrumDistance moyenneSpectre différentTypiqueDomain shiftp95 RMS distance to dataset mean spectrumalert = RMS_p95 / signal_scale, saturated at 25%
Comparaison à référenceSpectral Angle Mapper (SAM)Différence de formeForme différenteSimilaireFond, géométriep95 spectral angle to dataset mean spectrumalert = min(1, SAM_p95 / 0.35 rad)
RépétabilitéRMS intra-IDReproductibilitéMauvaise répétabilitéStablePositionnementmedian RMS distance to repeated-sample centroidalert = RMS_intra_ID / signal_scale, saturated at 10%
RépétabilitéSAM intra-IDVariation de formeInstableStableAcquisitionmedian SAM to repeated-sample centroidalert = min(1, SAM_intra_ID / 0.15 rad)
RépétabilitéCV intra-IDVariabilité interneMauvais contrôleStableOpérateurmedian within-ID band CValert = min(1, CV_intra_ID / 0.25)
Structure du datasetPCA score densityClustersSous-populationsHomogèneLots différents1 / median kNN distance in PCA score spacealert follows density_cv/profile structure complexity, not raw density alone
Structure du datasetLocal Outlier Factor (LOF)Anomalie localeSpectre isoléPopulation normaleCas raresp95 approximate LOF from PCA-score kNN distancesalert = min(1, max(0, LOF_p95 - 1) / 2)
Structure du datasetIsolation Forest scoreAnomalie globaleSpectre atypiqueNormalDiverses causesp95 IsolationForest anomaly score on PCA scoresalert follows structure complexity; raw score is implementation-dependent
Technology-specific extensions
TechnologieAdaptations / métriquesAnomalies cibléesCommentaire pratique
UV-Vis 300-1000 nmBaseline, pente globale, dérive aux bords 300-350 et 900-1000; métriques par zonesLumière parasite, mauvais blanc, saturation, faible signal aux extrémitésLes bords sont souvent instables; calculer aussi des scores edge/middle.
UV-Vis 300-1000 nmSaturation / clipping proche absorbance max ou réflectance maxSignal écrêtéTrès important si absorption forte.
UV-Vis 300-1000 nmRed-edge, position de maximum, ratios de bandes si végétalDécalage biologique ou artefact optiqueAide à distinguer changement réel et problème d'acquisition.
UV-Vis 300-1000 nmSmoothness / roughness indexBruit haute fréquenceSouvent plus informatif que le SNR seul.
MIR / ATR-FTIRATR contact quality index: intensité globale, aire totale, profondeur des bandes clésMauvais contact cristal-échantillonCrucial: beaucoup d'anomalies viennent du contact ATR.
MIR / ATR-FTIRCO2 / H2O atmospheric bandsMauvaise correction atmosphériquePics parasites fréquents.
MIR / ATR-FTIRBaseline curvature / rubber-band residualDiffusion, contact, dérive baselineTrès utile avant PCA.
MIR / ATR-FTIRPeak position shiftMauvais alignement spectral / calibrationImportant en FTIR car de petits shifts comptent.
MIR / ATR-FTIRBand area ratios sur bandes connuesSpectre chimiquement incohérentÀ adapter par matrice: polysaccharides, protéines, lipides, etc.
HS-MSTotal Ion Current (TIC), Base Peak Intensity (BPI)Injection faible, ionisation instableÉquivalent MS du niveau global spectral.
HS-MSNombre de pics détectésSpectre pauvre ou trop bruitéTrop peu = mauvais signal; trop = bruit/contamination.
HS-MSMass accuracy / m/z driftProblème calibration masseFondamental en HRMS.
HS-MSRetention time drift si LC/GC-MSDérive chromatographiqueÀ suivre sur standards/QC pools.
HS-MSBlank contamination scoreContaminants / carry-overComparer échantillons vs blancs.
HS-MSInternal standard CVVariabilité instrumentaleTrès robuste si standards disponibles.
HS-MSMissingness par featureInstabilité de détectionCrucial pour filtrer les variables.
Avec répétitionsRMS intra-échantillonRépétabilité globaleApplicable à toutes les technologies.
Avec répétitionsSAM / corrélation intra-échantillonRépétabilité de formeTrès utile pour spectres.
Avec répétitionsCV intra-échantillon par bande / featureRépétabilité localeDétecte les zones instables.
Avec répétitionsICC ou variance componentsPart variance échantillon vs techniqueTrès utile si plusieurs répétitions par sample.
Avec répétitionsDistance au centroïde intra-IDRépétition aberrantePermet de flagger la mauvaise répétition plutôt que le sample entier.
Bug-hunting / supervised audits
Famille de bug potentielMéthodes à ajouterCe que ça détecteÉtat dans l’explorateur
Shift spectral globalCorrélation spectre moyen inter-dataset, DTW, cross-correlation, comparaison positions de picsDécalage en longueur d'onde, mauvais alignement, interpolation différentePartiellement calculé: cross-correlation lag et dispersion des positions de pics vs spectre moyen.
Baseline / offset / gainRégression chaque spectre vs spectre moyen: x = a + b ref + residual; suivi de a, b, RMS résiduelOffset additif, effet multiplicatif, dérive de baselineCalculé dans reference.affine_*.
Mélange de lignes / mauvais appariement X-M-YVérification index, hash des lignes, duplication ID, distance spectrale intra-ID, labels incohérentsLignes mélangées, metadata mal alignées, Y attribué au mauvais spectrePartiellement couvert par répétabilité intra-ID; checks index/hash à ajouter au pipeline canonical.
Fuite d'information / répétitions mal splitéesGroupKFold par sample_id vs StratifiedKFold random; audit des partitions par sample_idPerformance artificiellement bonne due aux répétitionsNécessite splits et benchmark modèle; non calculé par la carte descriptive.
Label bugsÉchantillons proches en X mais Y différents, confident learning, erreurs systématiques FP/FNY inversés, erreurs de saisie, classes ambiguësNécessite Y et/ou modèle; recommandé pour l'explorateur supervisé.
Sous-domaines cachésPCA/UMAP/t-SNE + clustering non supervisé + association avec dataset/Y/date/operatorLots, campagnes, sondes, backgrounds non renseignésPartiellement calculé par structure PCA/LOF; UMAP/t-SNE hors carte statique.
Artefacts localisés inconnusCarte wavelength x dataset: différence moyenne, différence variance, KS par longueur d'ondeRégions spectrales anormales non anticipéesÀ calculer au niveau banque quand plusieurs datasets partagent un axe spectral.
Ruptures instrumentalesDiscontinuités dans dérivées, changepoint detectionSplice, raccord détecteur, saut local non prévuCalculé par jump/spike rates; changepoint plus avancé à ajouter.
Mélange / contamination spectraleNMF / unmixing / reconstruction par convex hullComposante externe: fond, plastique, solNon calculé automatiquement; nécessite hypothèses de composants ou grande bibliothèque.
Features instables mais prédictivesImportance modèle vs instabilité QC par variableModèle qui apprend un artefact plutôt qu'un signal biologiqueNécessite modèle supervisé; recommandé pour rapports de benchmark.

Variables

Targets 2

material_name

target · categorical
material_name classesPicritePicrite: 22DiabaseDiabase: 11IjoliteIjolite: 11
n / missing4 / 0
Classes3
Balance (entropy)0.95
Imbalance ratio2
Top classPicrite (2)

subclass

target · categorical
subclass classesUltramaficUltramafic: 33MaficMafic: 11
n / missing4 / 0
Classes2
Balance (entropy)0.81
Imbalance ratio3
Top classUltramafic (3)

Metadata 4

ecostress_resource_id

metadata · categorical
n / missing4 / 0
Classes4
Balance (entropy)1
Imbalance ratio1
Top classrock.igneous.mafic.fine.all.diabase_h1.jhu.becknic.spectrum (1)

location

metadata · categorical
n / missing4 / 0
Classes4
Balance (entropy)1
Imbalance ratio1
Top classfrom St. Peters, Pennsylvania, via Ward's Scientific (Cat. No. W-127) (1)

sample_description

metadata · categorical
n / missing4 / 0
Classes4
Balance (entropy)1
Imbalance ratio1
Top classThe sample from which the powder was derived was a medium-grained dark gray rock consisting of feldspar and pyroxene.Particle size was 0-75 micrometers. Original ASTER Spectral Library name was jhu.becknic.rock.igneous.mafic.fine.diabas1.spectrum.txt (1)

notes

metadata · categorical
n / missing4 / 0
Classes4
Balance (entropy)1
Imbalance ratio1
Top classrock.igneous.mafic.fine.all.diabase_h1.jhu.becknic.ancillary.txt (1)
Constant metadata 14
  • categoryrock
  • material_typerock
  • instrumentjhu.becknic
  • acquisition_modeDirectional (10 degree) hemispherical reflectance
  • signal_typeReflectance (percent)
  • axis_unitWavelength (micrometers)
  • axis_min0.4
  • axis_max14.01
  • n_points_original2,844
  • publication_doi10.1016/j.rse.2019.05.015
  • citationMeerdink et al. 2019, Baldridge et al. 2009
  • licenseCopyright California Institute of Technology / JPL, all rights reserved
  • rights_statusmanual_review_needed
  • usage_scopeprivate_use_only

9 variable(s) omitted (no recorded values).

Alignment

Alignment levelobservation
Sample id availableyes
Samples4
Observations (total)4
Reps per samplemin 1 · mean 1 · max 1

Provenance & citation

ContributorECOSTRESS Spectral Library
Origin · url [open]https://speclib.jpl.nasa.gov/download
Origin · url [open]https://speclib.jpl.nasa.gov/
Origin · script [manual]source_to_standard.py — standardization script (maintainer-only)
Publication10.1016/j.rse.2019.05.015 — The ECOSTRESS spectral library version 1.0
Publication10.1016/j.rse.2008.11.007 — The ASTER Spectral Library Version 2.0

Governance & integrity

Tierprivate
LicenseLicenseRef-not-cleared
Permitted useResearch and benchmarking; private use only.
Access policyManual download / private-use-only per source.
RedistributionOfficial ECOSTRESS page requests citation and states copyright/all rights reserved; converted matrices are private/internal until redistribution rights are clarified.
Content version1.0.0
Schema / protocol2.0
Content hash01c4b1cdb14dfb62…
Processing hasha4ae7c5309280910…
Metadata hashfdc8032524c46988…

Load this dataset

# pip install nirs4all-datasets
from nirs4all_datasets import get

# private dataset — export requires a Dataverse token
ds = get("ecostress_rock_all_2844points_4rows", token="…")
X, y = ds.x(), ds.y()
print(X.shape, y.shape)

Metadata downloads are available for public datasets only. The dataset bytes are never served here — fetch them from the origin / DOI above.