ADR1D and WQP-NM-Nutrients 💧

A reproducible analytical benchmark for one-dimensional contaminant transport, accompanied by a provenance-preserving snapshot of public water-quality observations.

Controlled transport physics and traceable environmental records in a compact, text-native release.

🔗 Quick Links

---

📋 Table of Contents

• Overview
• Repository Structure
• ADR1D Scientific Model
• ADR1D Generation Methodology
• ADR1D Data Dictionary
• ADR1D Partitions and Leakage Prevention
• ADR1D Quick Start
• Validation and Quality Control
• Release Identity
• ADR1D Intended Uses
• ADR1D Limitations
• WQP Observational Component
• How to Cite
• Scientific References
• Research Team
• Industry Partners
• License and Rights
• Acknowledgments
• Contact and Support
• FAQ

---

🌟 Overview

This data release contains two separate and complementary components for data-driven studies of contaminant transport and water quality:

• ADR1D is a synthetic one-dimensional benchmark with complete transport parameters, analytical space-time concentration fields, and noisy virtual sensors.
• WQP-NM-Nutrients is a curated snapshot of publicly accessible nutrient observations from streams in New Mexico, retrieved from the Water Quality Portal (WQP).

ADR1D supports controlled forward, inverse, reconstruction, and early-arrival experiments. WQP-NM-Nutrients provides environmental observations with source, organization, analytical-method, qualification, and station metadata. The two components are not samples from the same system and must not be concatenated as if they shared a physical or statistical population.

🔧 Key Capabilities

• Controlled benchmarking: analytical concentration fields and complete physical parameters for every ADR1D scenario.
• Leakage-aware evaluation: fixed train, validation, and test partitions at scenario level.
• Observation realism: noisy virtual sensors with documented random seeds.
• Environmental context: WQP records with provider, organization, method, qualification, and monitoring-location metadata.
• Auditability: plain-text tables, exact retrieval filters, processing decisions, validation results, and versioned release metadata.

Release: ADR1D version 1.0.0 and the WQP snapshot were generated or curated and validated before publication. Component-specific rights and WQP attribution are documented below.

📊 Package at a Glance

Component	Scope	Main records
ADR1D	Synthetic analytical transport benchmark	300 scenarios
ADR1D field	51 positions by 49 times per scenario	749,700 rows
ADR1D sensors	6 sensors by 49 times per scenario	88,200 rows
WQP-NM-Nutrients	New Mexico stream nutrients, query window 2022-2026	9,932 observations
WQP numeric subset	Numeric, unit-bearing, uncensored results	5,360 rows
WQP stations	Locations used by the observations	342 stations
WQP provenance	NWIS and STORET records	15 organizations
Format	Comma-separated text with header rows	6 CSV files

📆 Release Chronology

Item	Date
Scheduled development period	December 1-31, 2024
Initial public repository release	July 20, 2026
Last documentation update	July 2026

The scheduled development period and the public release record describe different stages of the project. Retrieval, validation, repository, and citation dates retain the dates on which those operations actually occurred.

---

📂 Repository Structure

.
|-- README.md
|-- CITATION.cff
|-- LICENSE
|-- data_processed/
|   |-- synthetic_adr1d_scenarios.csv
|   |-- synthetic_adr1d_field.csv
|   |-- synthetic_adr1d_sensor_observations.csv
|   |-- wqp_nm_stream_nutrients_2022_2026_observations.csv
|   |-- wqp_nm_stream_nutrients_2022_2026_model_ready.csv
|   `-- wqp_nm_stream_nutrients_2022_2026_stations.csv
`-- docs/
    `-- team/
        |-- gtinoco.webp
        |-- dmota.webp
        |-- jagt.webp
        |-- gpj.webp
        |-- eci.webp
        |-- jlgf.webp
        |-- cnmb.webp
        `-- mgfl.webp

The three ADR1D files form one relational dataset:

• synthetic_adr1d_scenarios.csv is the parent table, with one row per physical scenario.
• synthetic_adr1d_field.csv contains 2,499 analytical values per scenario (51 spatial positions by 49 times).
• synthetic_adr1d_sensor_observations.csv contains 294 observations per scenario (6 sensors by 49 times).

Use scenario_id to join the tables. The split column is repeated in all three files for convenience and is constant within each scenario.

The three WQP files form a second relational component:

• wqp_nm_stream_nutrients_2022_2026_observations.csv preserves all 9,932 processed results, including censored and nonnumeric records.
• wqp_nm_stream_nutrients_2022_2026_model_ready.csv is a first-stage subset of 5,360 numeric, unit-bearing, uncensored results.
• wqp_nm_stream_nutrients_2022_2026_stations.csv contains the 342 monitoring locations used by the observations.

Use monitoring_location_id to join WQP observations to stations. The model_ready table is a subset of observations, not an independent table.

This repository is a data-focused distribution. The equations, sampling rules, random seeds, WQP queries, processing rules, variable definitions, validation results, and rights notices needed to audit the release are documented below.

---

📚 ADR1D Scientific Model

Governing equation

ADR1D represents transport in a homogeneous one-dimensional medium with constant parameters within each scenario:

$$ R\frac{\partial C}{\partial t} =D\frac{\partial^2 C}{\partial x^2} -v\frac{\partial C}{\partial x} -\lambda R C. $$

Symbol	Meaning	Unit
$C(x,t)$	Dissolved concentration	mg L-1
$x$	Distance from the inlet	m
$t$	Elapsed time	s
$v$	Effective advective velocity	m s-1
$D$	Longitudinal hydrodynamic dispersion coefficient	m2 s-1
$R$	Linear equilibrium retardation factor	dimensionless
$\lambda$	First-order decay coefficient	s-1

The initial condition is

$$ C(x,0)=0. $$

At the inlet, a rectangular concentration pulse is prescribed:

$$C(0,t)=C_0\mathbf{1}_{[t_0,t_0+\tau)}(t).$$

where $C_0$ is the source concentration, $t_0$ is the pulse start time, and $\tau$ is its duration. The indicator equals one while the source pulse is active and zero otherwise. The far-field condition is

$$ \lim_{x\rightarrow\infty}C(x,t)=0. $$

The model therefore represents a semi-infinite domain. The interval from 0 to 1,000 m in the files is an observation window, not a finite domain with an outlet boundary at 1,000 m.

Analytical reference solution

The concentration field is evaluated with a reactive extension of the Ogata-Banks step response [1]. Define

$$ u=\sqrt{v^2+4\lambda R D}. $$

For $t>0$, the response to a unit step at the inlet is

$$ \begin{aligned} S(x,t)={}&\frac{1}{2} \exp!\left(\frac{(v-u)x}{2D}\right) \mathrm{erfc}!\left( \frac{Rx-ut}{2\sqrt{DRt}} \right)\\ &+\frac{1}{2} \exp!\left(\frac{(v+u)x}{2D}\right) \mathrm{erfc}!\left( \frac{Rx+ut}{2\sqrt{DRt}} \right), \end{aligned} $$

with $S(x,t)=0$ for $t\leq0$. Linearity permits a finite pulse to be written as the difference between two step responses:

$$ C(x,t)=C_0\left[S(x,t-t_0)-S(x,t-t_0-\tau)\right]. $$

At $x=0$, the prescribed pulse is imposed directly. This analytical construction avoids attaching the reference labels to a particular numerical mesh, time integrator, or truncation error.

Ogata and Banks [1] provide the classical semi-infinite longitudinal-dispersion solution. Chen and Liu [2] are cited as a complementary analytical treatment of one-dimensional transport with linear equilibrium sorption, first-order decay, and time-dependent inlet conditions. ADR1D uses the explicit semi-infinite response stated above; it does not use the finite-domain integral solution from Chen and Liu.

⚙️ ADR1D Generation Methodology

The processed release was produced in six deterministic stages:

1. Construct a separate Latin hypercube for each transport regime.
2. Map unit-hypercube coordinates to the physical parameter ranges and impose the neutral values for inactive retardation or decay.
3. Derive dispersivity, dimensionless numbers, travel time, and pulse end time.
4. Assign complete scenarios to stratified training, validation, and test partitions.
5. Evaluate the analytical pulse response on the field grid and at the six virtual sensors.
6. Add seeded sensor noise, apply the reporting-limit rule, serialize the three CSV tables, and run structural and physical validation.

Transport regimes

The benchmark contains 75 scenarios in each of four regimes. Inactive effects are fixed exactly at their neutral values.

Regime	Retardation factor $R$	Decay coefficient $\lambda$	Train	Validation	Test
conservative	1	0	53	11	11
decay_only	1	sampled, >0	53	11	11
retardation_only	sampled, >1	0	52	12	11
retardation_and_decay	sampled, >1	sampled, >0	52	11	12
Total			210	45	45

Parameter design

Within each regime, seven continuous dimensions were sampled with a 75-point Latin hypercube design [3]. Each dimension was divided into 75 equiprobable strata, one uniformly jittered value was drawn from every stratum, and the values were independently shuffled before the dimensions were combined.

Parameter	Symbol	Design range	Scale
Advective velocity	$v$	0.05-0.35 m s-1	linear
Dispersion coefficient	$D$	1-20 m2 s-1	logarithmic
Retardation factor, when active	$R$	1.1-3.0	linear
Decay coefficient, when active	$\lambda$	10-7-2×10-5 s-1	logarithmic
Source concentration	$C_0$	0.2-5.0 mg L-1	logarithmic
Pulse start time	$t_0$	0-3,600 s	linear
Pulse duration	$\tau$	1,800-10,800 s	linear

Because Latin-hypercube values are jittered within open strata, the nominal design bounds are generally not attained exactly. The realized ranges in version 1.0.0 are:

Quantity	Minimum	Maximum
velocity_m_s	0.0502795	0.349056
dispersion_m2_s	1.00072	19.8959
retardation_factor	1	2.99246
decay_rate_s_1	0	1.98670×10-5
source_concentration_mg_L	0.200193	4.99548
source_start_s	9.27389	3,590.29
source_duration_s	1,870.87	10,793.29
dispersivity_m	3.25712	303.427
peclet_number	3.29568	307.019
damkohler_number	0	0.416173
advective_travel_time_s	2,869.31	50,211.5

The deterministic parameter and partition seed is 20260714. Sampling was implemented with Python's standard random module; no external sampling library was used. Floating-point outputs were serialized with 12 significant digits.

Four derived quantities are included in the scenario table:

$$ \alpha=\frac{D}{v},\qquad \mathrm{Pe}=\frac{vL}{D},\qquad \mathrm{Da}=\frac{\lambda R L}{v},\qquad t_{\mathrm{adv}}=\frac{RL}{v}, $$

where $L=1{,}000$ m is the sampled length, $\alpha$ is dispersivity, $\mathrm{Pe}$ is the domain Péclet number, $\mathrm{Da}$ is an advective Damköhler number, and $t_{\mathrm{adv}}$ is the retarded advective travel time over $L$.

Space-time output grid

The analytical solution is evaluated at 51 uniformly spaced positions from 0 to 1,000 m ($\Delta x=20$ m) and 49 uniformly spaced times from 0 to 86,400 s ($\Delta t=1{,}800$ s). These coordinates are output locations only; they are not nodes of a numerical solver.

Virtual-sensor model

Six fixed sensors are evaluated analytically at every one of the 49 output times:

Sensor	Position (m)
S01	100
S02	250
S03	400
S04	600
S05	800
S06	1,000

Sensor values are evaluated directly from the analytical solution, not interpolated from synthetic_adr1d_field.csv. This distinction matters for S02 at 250 m, which does not coincide with the 20 m field spacing.

For a true concentration $C_{\mathrm{true}}$, independent Gaussian noise is generated with

$$ \epsilon\sim\mathcal{N}(0,\sigma^2),\qquad \sigma=\max\left(0.002,,0.03C_{\mathrm{true}}\right) \ \text{mg L}^{-1}. $$

The noisy value is first constrained to be nonnegative:

$$ C_{\mathrm{raw}}=\max(0,C_{\mathrm{true}}+\epsilon). $$

A synthetic detection limit of 0.01 mg L^-1 is then applied. If $C_{\mathrm{raw}}<0.01$ mg L^-1, the reported concentration is 0.005 mg L^-1 and is_below_detection_limit=true; otherwise the reported value is $C_{\mathrm{raw}}$. The independent noise seed is 20260715.

Overall, 73,405 of 88,200 observations (83.23%) fall below the synthetic detection limit. This high proportion is expected because the complete time window includes periods before plume arrival and after pulse passage. It must be considered when designing regression, classification, or early-warning experiments. No scenario is fully censored: the number of reported detections per scenario ranges from 9 to 221, with a median of 40.

🗄️ ADR1D Data Dictionary

All files are comma-separated UTF-8 text with a header row and a period as the decimal separator. Every listed field is required. The release contains no blank data cells and uses no missing-value sentinel codes. Boolean values are stored as lowercase true and false.

synthetic_adr1d_scenarios.csv

One row describes one complete physical scenario.

Column	Type	Unit	Description
scenario_id	string	-	Stable identifier, from ADR1D-0001 to ADR1D-0300.
split	category	-	Scenario-level partition: train, validation, or test.
regime	category	-	conservative, decay_only, retardation_only, or retardation_and_decay.
domain_length_m	numeric	m	Length of the sampled interval.
spatial_nodes	integer	-	Number of sampled spatial coordinates.
spatial_step_m	numeric	m	Distance between adjacent field coordinates.
final_time_s	numeric	s	End of the sampled time window.
time_nodes	integer	-	Number of sampled times.
time_step_s	numeric	s	Time between adjacent outputs.
velocity_m_s	numeric	m s-1	Effective advective velocity $v$.
dispersion_m2_s	numeric	m2 s-1	Longitudinal dispersion coefficient $D$.
retardation_factor	numeric	dimensionless	Linear retardation factor $R$.
decay_rate_s_1	numeric	s-1	First-order decay coefficient $\lambda$.
source_concentration_mg_L	numeric	mg L-1	Inlet concentration $C_0$ during the pulse.
source_start_s	numeric	s	Pulse start time $t_0$.
source_duration_s	numeric	s	Pulse duration $\tau$.
source_end_s	numeric	s	Derived pulse end, $t_0+\tau$.
dispersivity_m	numeric	m	Derived dispersivity $D/v$.
peclet_number	numeric	dimensionless	Domain Péclet number $vL/D$.
damkohler_number	numeric	dimensionless	Advective Damköhler number $\lambda RL/v$.
advective_travel_time_s	numeric	s	Retarded travel time $RL/v$.

synthetic_adr1d_field.csv

Each row is one analytical concentration at a scenario, time, and position.

Column	Type	Unit	Description
scenario_id	string	-	Foreign key to the scenario table.
split	category	-	Scenario-level data partition.
time_s	numeric	s	Elapsed time.
x_m	numeric	m	Distance from the inlet.
concentration_mg_L	numeric	mg L-1	Noise-free analytical concentration $C(x,t)$.
normalized_concentration	numeric	dimensionless	$C(x,t)/C_0$ for the corresponding scenario.

synthetic_adr1d_sensor_observations.csv

Each row is one virtual-sensor observation.

Column	Type	Unit	Description
observation_id	string	-	Stable identifier combining scenario, sensor, and time index.
scenario_id	string	-	Foreign key to the scenario table.
split	category	-	Scenario-level data partition.
sensor_id	category	-	Fixed virtual-sensor identifier, S01-S06.
x_m	numeric	m	Sensor position measured from the inlet.
time_s	numeric	s	Synthetic sampling time.
concentration_true_mg_L	numeric	mg L-1	Noise-free analytical concentration.
noise_std_mg_L	numeric	mg L-1	Gaussian standard deviation used for the observation.
concentration_observed_mg_L	numeric	mg L-1	Nonnegative noisy value after applying the detection-limit rule.
detection_limit_mg_L	numeric	mg L-1	Synthetic reporting limit, fixed at 0.01.
is_below_detection_limit	boolean	-	Whether the value before reporting substitution was below the limit.

🔀 ADR1D Partitions and Leakage Prevention

Splitting was performed at the scenario level and stratified by transport regime. All space-time points and sensor observations from a scenario belong to the same partition. This prevents information from one physical realization from appearing in both model fitting and evaluation.

Split	Scenarios	Field rows	Sensor rows
train	210	524,790	61,740
validation	45	112,455	13,230
test	45	112,455	13,230
Total	300	749,700	88,200

Use the supplied partitions for comparable benchmark results. Any alternative cross-validation scheme should group by scenario_id, never by individual rows.

🚀 ADR1D Quick Start

Clone the repository and enter its root directory:

git clone https://github.com/gstinoco/ADR1D.git
cd ADR1D

The CSV files can be read without specialized software. The following example uses pandas:

import pandas as pd

root = "data_processed"

scenarios = pd.read_csv(f"{root}/synthetic_adr1d_scenarios.csv")
field = pd.read_csv(f"{root}/synthetic_adr1d_field.csv")
sensors = pd.read_csv(
    f"{root}/synthetic_adr1d_sensor_observations.csv",
    converters={
        "is_below_detection_limit": lambda value: value.lower() == "true"
    },
)

# Join exact fields to their physical parameters.
field_with_parameters = field.merge(
    scenarios,
    on=["scenario_id", "split"],
    how="left",
    validate="many_to_one",
)

# Preserve the scenario-level split.
train = field_with_parameters[field_with_parameters["split"] == "train"]
validation = field_with_parameters[
    field_with_parameters["split"] == "validation"
]
test = field_with_parameters[field_with_parameters["split"] == "test"]

Recommended analytical practice

• Fit scalers, transformations, and feature-selection procedures on the training split only.
• Use scenario_id, time_s, and x_m as explicit keys; do not rely on row order when reshaping fields.
• Treat the 300 scenarios as the independent physical realizations. The 749,700 field rows are repeated space-time outputs, not 749,700 independent samples.
• Treat concentration_true_mg_L as a noise-free target and concentration_observed_mg_L as a reported synthetic measurement.
• Do not interpret the substituted value 0.005 mg L^-1 as an exact measurement. Retain is_below_detection_limit in analyses of censored data.
• Report performance by transport regime in addition to aggregate metrics.
• Account for the 83.23% below-limit proportion in arrival-detection or classification tasks.
• State clearly whether a study uses the exact field, noisy sensors, or both.

✅ Validation & Quality Control

Before this processed-data distribution was assembled, the canonical generation workflow checked:

• file dimensions, column order, identifiers, and scenario uniqueness;
• scenario-level partition separation and regime constraints;
• derived parameters and dimensionless numbers;
• the initial condition and prescribed inlet pulse;
• 14,700 independent analytical recalculations from the field table;
• the differential-equation residual at 1,982 interior points using centered differences;
• all 88,200 noise-free sensor concentrations;
• the Gaussian noise model, nonnegativity rule, detection limit, and reporting substitution; and
• the dimensions and integrity of the distributed files.

The canonical validation completed with status ok. The maximum absolute error among the independently recalculated analytical values was below 1.7×10^-10 mg L^-1. The initial-condition and inlet-boundary errors were zero at the stored output points. The maximum absolute PDE residual was 4.79×10^-7 mg L^-1 s^-1, and the 99th percentile of the relative residual was 3.03×10^-4.

🔒 Release Identity

The files under data_processed/ constitute the dataset snapshot distributed with ADR1D version 1.0.0. Their integrity was checked before publication. Repository history and the release version identify this snapshot without duplicating low-level integrity values in the public documentation.

Any modification to values, scenario assignments, serialized precision, source records, or row ordering should be published under a new version.

🎯 ADR1D Intended Uses

• Surrogate modeling of $C(x,t)$ from transport and source parameters.
• Inverse estimation of $v$, $D$, $R$, $\lambda$, $C_0$, $t_0$, or $\tau$.
• Full-field reconstruction from sparse virtual sensors.
• Plume-arrival and early-detection experiments.
• Comparison of numerical solvers with machine-learning or deep-learning models.
• Evaluation of models across conservative, reactive, and retarded regimes.

⚠️ ADR1D Limitations and Out-of-Scope Uses

• ADR1D is synthetic and is not calibrated to a particular aquifer, river, soil profile, contaminant, or monitoring network.
• The medium is one-dimensional, homogeneous, and stationary within each scenario.
• Spatial heterogeneity, transient flow, nonlinear reactions, multispecies chemistry, and coupled flow are not represented.
• The sampled interval ends at 1,000 m, but the analytical model is semi-infinite; the dataset must not be interpreted as imposing an outlet boundary at the last sensor.
• Parameter ranges were selected for methodological coverage rather than as a probability model for a real site.
• Parameters were sampled independently within the stated design, except for regime constraints and derived quantities.
• The effective number of independent physical realizations is 300; pointwise rows within a scenario are strongly related through the analytical model.
• Sensor errors are independent and Gaussian, without drift, temporal correlation, calibration bias, or position uncertainty.
• A single idealized detection limit and half-limit substitution rule are used.
• The large proportion of below-limit observations can dominate unstratified error metrics.
• The training, validation, and test scenarios share the same parameter ranges. They support interpolation tests within the design space, not claims of extrapolation or out-of-distribution generalization.
• Analytical labels contain no numerical-solver error; performance on ADR1D does not by itself demonstrate transfer to discretized or field data.

The dataset is suitable for methodological benchmarking. It should not be used for site-specific environmental decisions without independent field data, calibration, and a model appropriate to the site.

🌊 WQP Observational Component

Source and retrieval

WQP-NM-Nutrients is a processed snapshot of records made publicly accessible through the Water Quality Portal, a cooperative service of the National Water Quality Monitoring Council, USGS, and EPA. The extract covers stream locations in New Mexico, the WQP Nutrient characteristic type, and a result-query window from January 1, 2022, through July 14, 2026.

The canonical files were retrieved on July 14, 2026, with these requests:

• Result query
• Station query
• Monitoring-location summary query

WQP is a live service. Repeating these requests can return a different snapshot as providers revise or add records. The tables included with ADR1D version 1.0.0 constitute the processed snapshot distributed here.

Curation method

The source result and station files are preserved in the internal archive. The public tables were produced without imputing concentrations or converting units:

1. Select and rename fields needed for provenance, sampling context, qualification, analytical method, and source status.
2. Generate a stable local observation_id while retaining WQP activity, organization, provider, and monitoring-location identifiers.
3. Parse ResultMeasureValue into result_value when numeric and preserve the original text in raw_result_value.
4. Flag records with detection or quantitation conditions instead of assigning replacement concentrations.
5. Create the model_ready subset only from numeric records with a reported unit and no detection or quantitation condition.
6. Retain one station record for each monitoring location used by the observations.

The name model_ready means that the first structural filter has been applied. It does not mean that analytes, units, chemical bases, sampling fractions, methods, or source statuses have been harmonized for a particular model.

Coverage and quality summary

Item	Value
Processed observations	9,932
Numeric source values	5,456
First-stage numeric subset	5,360
Monitoring locations used	342
Locations missing coordinates	0
Contributing organizations	15
Earliest / latest sample	2022-01-05 / 2026-05-20

Provider coverage is not equally current:

Provider	Rows	Earliest sample	Latest sample
NWIS	2,290	2022-02-01	2023-07-20
STORET	7,642	2022-01-05	2026-05-20

The legacy WQP interface used for the canonical download does not include recent USGS records. The observed NWIS maximum above is therefore a property of this snapshot, not evidence that USGS sampling stopped in 2023.

Source result status	Rows
Final	7,510
Preliminary	1,467
Accepted	955

Detection or quantitation condition	Rows
None reported	5,471
Not Detected	2,477
Present Below Quantification Limit	1,945
Detected Not Quantified	39

The most frequent reported characteristics are:

Characteristic	Rows
Phosphorus	1,636
Nitrate	1,074
Nitrite	1,018
Nitrate + Nitrite	1,008
Ammonia-nitrogen	876
Total Kjeldahl nitrogen (Organic N & NH3)	737
Kjeldahl nitrogen	724
Orthophosphate	410
Ammonia	369
Total Phosphorus, mixed forms	315
Ammonia and ammonium	312
Nitrogen	306
Total Nitrogen, mixed forms	268
Nitrogen, mixed forms (NH3), (NH4), organic, (NO2) and (NO3)	267
Organic Nitrogen	214

Contributing organizations

The organization fields are part of the data provenance and should be retained in redistributed or derived tables.

Organization ID	Organization	Rows
21NMEX_WQX	NM Environmental Dept./SWQB	2,932
USGS-NM	USGS New Mexico Water Science Center	2,290
PUEBLOISLETA	Pueblo of Isleta, New Mexico (Tribal)	1,429
POLSWATER_WQX	Pueblo of Laguna (Tribal)	838
SANDIAWQ_WQX	Pueblo of Sandia Water Quality Program (New Mexico)	693
TAOSPBLO_WQX	Pueblo of Taos (Tribal)	430
SANTACLARAPBLO	Pueblo of Santa Clara	407
PUEBLOJEMEZ	Pueblo of Jemez (Tribal)	216
PUEBLO_POJOAQUE	Pueblo of Pojoaque (Tribal)	170
DRMPWQX	Mescalero Apache Tribe DRMP (Tribal)	168
11NPSWRD_WQX	National Park Service Water Resources Division	159
ZIAPUEBLO_WQX	Zia Pueblo	103
SANFELIPE_WQX	San Felipe Pueblo (Tribal)	72
TCEQMAIN	Texas Commission on Environmental Quality	19
PUEBLO_SANTAANA	Pueblo of Santa Ana, New Mexico	6

WQP observation fields

The observations and model_ready files have the same 31 columns.

Field	Type / unit	Description
observation_id	string	Stable identifier generated for this release.
organization_id	string	Identifier of the organization that owns or supplied the source record.
organization_name	string	Formal name of the source organization.
provider	category	WQP source system: NWIS or STORET.
activity_id	string	Source sampling-activity identifier.
activity_type	string	WQP activity-type code.
sample_date	ISO date	Sampling-activity start date.
sample_time	local time	Reported activity start time, when available.
time_zone	string	Time-zone code associated with sample_time.
monitoring_location_id	string	Key used to join observations to stations.
sample_media	string	Environmental medium sampled.
sample_media_subdivision	string	Reported subdivision of the sample medium.
hydrologic_condition	string	Hydrologic condition reported for the activity.
hydrologic_event	string	Hydrologic event reported for the activity.
characteristic_name	string	Nutrient or nutrient-related characteristic.
sample_fraction	string	Fraction or filtering status reported for the result.
result_value	numeric; see result_unit	Parsed result; blank for censored, missing, or nonnumeric source values.
result_unit	string	Unit and chemical basis exactly as reported.
is_numeric_result	boolean	Whether the source value was parsed as numeric.
is_censored_or_qualified	boolean	Whether WQP reported a detection or quantitation condition.
detection_condition	string	Detection or quantitation condition.
measure_qualifier	string	Qualifier code attached to the measure.
result_status	category	Source status: Accepted, Final, or Preliminary.
result_value_type	string	WQP classification of the result value.
usgs_parameter_code	string	USGS parameter code when supplied by NWIS.
analytical_method_id	string	Identifier of the analytical method.
analytical_method_context	string	Vocabulary or organization defining the method identifier.
analytical_method_name	string	Reported analytical-method name.
laboratory_name	string	Reported laboratory name.
raw_result_value	string	Original, unparsed ResultMeasureValue text.
source_file	string	Preserved source filename from which the row was derived.

Blank cells represent information not supplied in the selected WQP profile; the files do not use a separate missing-value token.

WQP station fields

Field	Type / unit	Description
monitoring_location_id	string	Primary key matching the observation tables.
organization_id	string	Identifier of the source organization.
organization_name	string	Formal name of the source organization.
provider	category	WQP source system: NWIS or STORET.
monitoring_location_name	string	Name assigned by the source organization.
monitoring_location_type	string	WQP monitoring-location type.
huc8	string	Eight-digit hydrologic unit code.
latitude	decimal degrees	Monitoring-location latitude.
longitude	decimal degrees	Monitoring-location longitude.
horizontal_datum	string	Horizontal coordinate reference datum.
country_code	string	Reported country code.
state_code	string	Reported state code.
county_code	string	Reported county code.
drainage_area_value	numeric	Reported drainage area, when available.
drainage_area_unit	string	Unit for drainage_area_value.
source_file	string	Preserved source filename from which the row was derived.

WQP quick start

import pandas as pd

root = "data_processed"

observations = pd.read_csv(
    f"{root}/wqp_nm_stream_nutrients_2022_2026_observations.csv"
)
numeric_subset = pd.read_csv(
    f"{root}/wqp_nm_stream_nutrients_2022_2026_model_ready.csv"
)
stations = pd.read_csv(
    f"{root}/wqp_nm_stream_nutrients_2022_2026_stations.csv",
    dtype={"huc8": "string", "state_code": "string", "county_code": "string"},
)

observations_with_locations = observations.merge(
    stations,
    on="monitoring_location_id",
    how="left",
    validate="many_to_one",
    suffixes=("_observation", "_station"),
)

# Example conservative analysis subset; chemistry-aware unit harmonization is
# still required before comparing concentrations across characteristics.
final_numeric = numeric_subset[numeric_subset["result_status"] == "Final"]

WQP rights and attribution

EPA describes WQP as a public download and analysis resource, while the WQX metadata identify contributing organizations as the primary owners of their records. This release therefore follows a scoped-rights approach:

• The authors claim responsibility for query design, selection, field mapping, filtering, documentation, and validation of the processed snapshot.
• No ownership of the original WQP observations is claimed.
• No license assigned to ADR1D is extended to the WQP-derived tables.
• organization_id, organization_name, and provider must remain available so individual source records retain their attribution.
• Users should cite WQP, state the access date and query scope, and consult the source metadata before reuse in consequential analyses.

The EPA citation guidance and the USGS dataRetrieval citation record support the following source citation:

National Water Quality Monitoring Council (2026). Water Quality Portal. https://doi.org/10.5066/P9QRKUVJ. Accessed July 14, 2026. New Mexico; site type: Stream; characteristic type: Nutrient; result-query window: January 1, 2022-July 14, 2026.

WQP limitations

• These are discrete observational water-quality records, not a PDE-complete transport benchmark.
• Velocity, discharge, dispersion, porosity, initial conditions, and boundary conditions are generally unavailable in the result records.
• Units and chemical bases are preserved as reported. For example, mg/L, mg/l as N, and mg/l as P are not interchangeable without a chemistry-aware conversion.
• Nondetects and other qualified results require censored-data methods; the blank numeric cells in observations are not zeros.
• The model_ready subset excludes qualified results and can introduce selection bias if used without the complete observation table.
• Preliminary records are included and may be revised by their providers.
• NWIS and STORET have different temporal coverage in this snapshot.
• Station coordinates and measurements are reproduced from source records and have not been independently field-verified by the curators.
• WQP is dynamic; later retrievals may not reproduce the same rows.

📝 How to Cite

ADR1D

Machine-readable citation metadata are provided in CITATION.cff. The formal dataset citation includes the three principal researchers; student contributors are recognized in the research-team section below.

Please cite the repository as:

Tinoco-Guerrero, G., Domínguez-Mota, F. J., & Guzmán-Torres, J. A. (2026). ADR1D and WQP-NM-Nutrients: A reproducible contaminant-transport benchmark and curated water-quality snapshot (Version 1.0.0) [Data set]. Universidad Michoacana de San Nicolás de Hidalgo. https://doi.org/10.5281/zenodo.21499528

BibTeX:

@misc{TinocoGuerrero2026ADR1D,
  author    = {Tinoco-Guerrero, Gerardo and
               Domínguez-Mota, Francisco J. and
               Guzmán-Torres, J. Alberto},
  title     = {{ADR1D} and {WQP-NM-Nutrients}: A Reproducible
               Contaminant-Transport Benchmark and Curated
               Water-Quality Snapshot},
  year      = {2026},
  publisher = {Universidad Michoacana de San Nicolás de Hidalgo},
  note      = {Data set, version 1.0.0},
  doi       = {10.5281/zenodo.21499528},
  url       = {https://doi.org/10.5281/zenodo.21499528}
}

Cite the methodological sources below when the analytical model or sampling design is material to the study.

WQP-NM-Nutrients

When using the curated WQP tables, cite both the processed release and the original portal:

Tinoco-Guerrero, G., Domínguez-Mota, F. J., Guzmán-Torres, J. A. (2026). WQP-NM-Nutrients: Curated New Mexico stream-nutrient observations, 2022-2026 query window [Data set]. Universidad Michoacana de San Nicolás de Hidalgo. https://github.com/gstinoco/ADR1D. Source data: National Water Quality Monitoring Council, Water Quality Portal, https://doi.org/10.5066/P9QRKUVJ, accessed July 14, 2026.

Publications should also state the actual WQP filters and retain the source-organization attribution in derived archives.

📚 Scientific References

1. Ogata, A., and Banks, R. B. (1961). A solution of the differential equation of longitudinal dispersion in porous media. U.S. Geological Survey Professional Paper 411-A, pp. A1-A7. https://doi.org/10.3133/pp411A
2. Chen, J.-S., and Liu, C.-W. (2011). Generalized analytical solution for advection-dispersion equation in finite spatial domain with arbitrary time-dependent inlet boundary condition. Hydrology and Earth System Sciences, 15, 2471-2479. https://doi.org/10.5194/hess-15-2471-2011
3. McKay, M. D., Beckman, R. J., and Conover, W. J. (1979). Comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics, 21(2), 239-245. https://doi.org/10.1080/00401706.1979.10489755

---

🧑‍🔬 Research Team

🌟 Meet the Team

Researchers and students advancing traceable environmental data and reproducible transport benchmarks

👥 Main Researchers

Photo	Researcher	Affiliation	Contact
	Gerardo Tinoco-Guerrero Numerical methods and environmental modeling
	Francisco J. Domínguez-Mota Applied mathematics and numerical methods
	J. Alberto Guzmán-Torres Engineering applications and artificial intelligence

🎓 Ph.D. Research Students

Photo	Student	Institution	Contact
	Gabriela Pedraza-Jiménez
	Eli Chagolla-Inzunza

🎓 M.Sc. Research Students

Photo	Student	Institution	Contact
	Jorge L. González-Figueroa
	Christopher N. Magaña-Barocio

🎓 Undergraduate Research Students

Photo	Student	Institution	Contact
	Maria Goretti Fraga-Lopez

Student contributors are recognized for their participation in the broader research program. Formal dataset citation and copyright attribution remain limited to the three principal researchers listed in the citation section and the license file.

---

🏭 Industry Partners Supporting Innovation

🌟 Academic-Industry Collaboration

Connecting reproducible environmental data with applied engineering and technology transfer

SIIIA MATH Soluciones en Ingeniería Focus areas Mathematical modeling and scientific computing Environmental data products and reproducible benchmarks Applied research, development, and technology transfer

---

📄 License & Rights

License scope must be read by component:

• ADR1D and original documentation: the three author-generated ADR1D CSV files and the original documentation are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). Reuse is permitted with appropriate attribution, a link to the license, and an indication of changes.
• WQP-NM-Nutrients: the source observations were made publicly accessible through WQP and remain attributed to their contributing organizations. The curators do not apply the ADR1D license or assert ownership over those source records. Reusers should preserve provenance, cite WQP, and consult source metadata for the intended use.

The licenses of methodological publications do not determine the rights of either dataset component.

Institutional emblems and logos under docs/partners/ remain the property of their respective organizations. They are used solely for identification and acknowledgment and are not covered by the repository's CC BY 4.0 license.

---

🙏 Acknowledgments

❤️ Special Thanks

We thank the institutions and partners whose continuing support sustains this research, its public data products, and student participation.

🏛️ Institutional Support

Universidad Michoacana de San Nicolás de Hidalgo UMSNH · Academic institution · Mexico Support provided Research infrastructure and institutional backing Academic supervision and undergraduate and graduate training	Secretaría de Ciencia, Humanidades, Tecnología e Innovación SECIHTI · Federal science agency · Mexico Support provided Financial support for scientific and technological research Promotion of knowledge generation and public dissemination
International Centre for Numerical Methods in Engineering CIMNE · Research center and international collaboration Support provided Collaboration in numerical methods and computational engineering Research exchange, training, and international scientific links	SIIIA MATH: Soluciones en Ingeniería Engineering and innovation partner Support provided Applied research and engineering development Financial support and technology-transfer perspective

:building_with_garden: Research Centers & Collaborations

Aula CIMNE Morelia Research collaboration and training space Collaboration highlights Numerical methods and computational engineering environment Research exchange and reproducible-computing training

Coordinación de la Investigación Científica CIC-UMSNH · Institutional research coordination Collaboration highlights Institutional coordination for scientific research Support for data curation, documentation, and public dissemination

💧 Environmental Data Providers

National Water Quality Monitoring Council and WQP contributing organizations Public environmental observations with record-level provenance The authors acknowledge the organizations that make their monitoring records publicly accessible through WQP. Organization names and identifiers are retained in the distributed tables so provenance remains visible at record level.

💻 Technology Communities

📦 Framework	👥 Community	⭐ Contribution
	Python Community	Reproducible data-access and analysis examples
	pandas Community	CSV inspection, filtering, and analysis

---

📧 Contact & Support

Scientific questions, data corrections, and reproducibility support

Primary Research Contact Scientific coordination and data provenance Gerardo Tinoco-Guerrero Universidad Michoacana de San Nicolás de Hidalgo Morelia, Michoacán, Mexico

Repository Support Questions, corrections, and reproducibility reports Use GitHub Issues for reproducible data or documentation problems. Include the component, file name, row identifier, and expected result. Use email for scientific collaboration or provenance questions.

---

💬 FAQ

Where are the distributed datasets located?

All six CSV files are stored under data_processed/. The three synthetic_adr1d_*.csv files form the analytical benchmark, while the three wqp_nm_stream_nutrients_*.csv files form the separate observational component.

Can ADR1D and WQP-NM-Nutrients be merged into one training table?

No, not as observations from a common population. ADR1D is a controlled synthetic transport benchmark, whereas WQP-NM-Nutrients contains field observations from multiple organizations, locations, methods, units, and dates. They support complementary experiments but require separate scientific interpretations.

Is the WQP model_ready table ready for every statistical or ML task?

No. It is a conservative first-stage subset that removes nonnumeric, unit-missing, and censored results. Users must still harmonize chemical bases, units, characteristics, spatial support, temporal support, and study-specific quality criteria.

May the WQP-derived tables be redistributed?

The source records were made publicly accessible through WQP and are retained with provider and organization attribution. Cite WQP, preserve provenance, and review source metadata for consequential reuse. The repository's CC BY 4.0 license applies only to the author-generated ADR1D component and original documentation.

Can a later WQP query reproduce this snapshot exactly?

Not necessarily. WQP is dynamic and providers may add or revise records. Use the versioned tables when exact snapshot identity matters; use the documented filters and access date when conducting an updated query.

---

Open transport benchmarks and traceable environmental data for reproducible research

If this dataset supports your research, please cite it and consider giving the repository a star.