SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2608.08321 · 2026

Spatial Heterogeneity-Aware Multi-Hazard Susceptibility and Risk Mapping at Regional Scale

Aswathi Mundayatt, Siddharth Anil, Hitanshu Seth, Jaya Sreevalsan-Nair

arXiv · PDF · Open in the Atlas

Code that ran

11 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 9 failures of our instrument; 0 no recorded run), measured 2026-11-21; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2608.08321.json · badge: /api/ran/2608.08321.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
GVCL/Susceptibility-Mapping-FL-Hetero canonical (classifier label)
classifier basis: census
11 ran · 0 no recorded run
FunctionStatusWhere it lives
coerce_numeric_features Ran
ran, honoured the contract we drafted
GVCL/Susceptibility-Mapping-FL-Hetero/src/feature_selection/importance_selection.py
code served (permissive licence) · get_code("a36b6856ab8695c9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
import numpy as np

# Create a small DataFrame with mixed types and missing values
# This will test the coercion to numeric and filling with median
data = {
    'col1': [1.5, 2.3, np.nan, 4.1, 5.2],
    'col2': ['10', '20', 'invalid', '40', np.nan],
    'col3': [100, 200, 300, np.nan, 500],
    'col4': ['1.1', '2.2', '3.3', '4.4', '5.5']
}

X = pd.DataFrame(data)

ARGS = {
    'X': X
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_vif Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/feature_selection/vif_filter.py
code served (permissive licence) · get_code("3d012693a889ff38")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import logging
import numpy as np
import pandas as pd

logging.basicConfig(level=logging.WARNING)

# Create a sample DataFrame with numeric and non-numeric columns
np.random.seed(0)
n_rows = 10
n_cols = 4

# Create numeric data with some variation
numeric_data = {
    'feature_1': np.random.randn(n_rows),
    'feature_2': np.random.randn(n_rows) * 2 + 5,
    'feature_3': np.random.randn(n_rows) * 0.5,
    'feature_4': np.random.randn(n_rows) * 3,
}

# Add a non-numeric column to test that it's ignored
mixed_data = {
    **numeric_data,
    'category': ['A', 'B', 'C', 'D', 'A', 'B', 'C', 'D', 'A', 'B']
}

X = pd.DataFrame(mixed_data)

ARGS = {
    'X': X
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, statsmodels — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
drop_high_vif_features Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/feature_selection/vif_filter.py
code served (permissive licence) · get_code("1e1769dc4cc6d54e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import logging
import numpy as np
import pandas as pd

logging.basicConfig(level=logging.INFO)

# Create a small test DataFrame with numeric features that have varying VIF values
np.random.seed(42)
n_samples = 20

# Create features with some multicollinearity
X_data = {
    'feature_1': np.random.randn(n_samples),
    'feature_2': np.random.randn(n_samples),
    'feature_3': np.random.randn(n_samples) * 0.1 + 2.0,  # Nearly constant-ish
    'feature_4': np.random.randn(n_samples),
}

X = pd.DataFrame(X_data)

# Add a feature with high correlation to create high VIF
X['feature_5'] = X['feature_1'] * 0.95 + np.random.randn(n_samples) * 0.1

ARGS = {
    'X': X,
    'threshold': 5.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, statsmodels — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
drop_highly_correlated_features Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/feature_selection/correlation_filter.py
code served (permissive licence) · get_code("11a0cf0b412fe1e7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import logging
import numpy as np
import pandas as pd

logging.basicConfig(level=logging.INFO)

# Create a small DataFrame with some correlated features
np.random.seed(0)
n_samples = 10
feature_a = np.random.randn(n_samples)
feature_b = feature_a + np.random.randn(n_samples) * 0.1  # highly correlated with a
feature_c = np.random.randn(n_samples)  # independent

df = pd.DataFrame({
    'feature_a': feature_a,
    'feature_b': feature_b,
    'feature_c': feature_c,
    'feature_d': np.random.randn(n_samples)
})

ARGS = {
    'df': df,
    'threshold': 0.9,
    'method': 'pearson'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
evaluate_classifier Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/modeling/evaluation_metrics.py
code served (permissive licence) · get_code("dab56270eb927ac6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from sklearn.linear_model import LogisticRegression

# Set random seed for reproducibility
np.random.seed(0)

# Create small synthetic binary classification dataset
n_train = 10
n_test = 8
n_features = 4

X_train = np.random.randn(n_train, n_features).astype(np.float32)
y_train = np.random.randint(0, 2, n_train)

X_test = np.random.randn(n_test, n_features).astype(np.float32)
y_test = np.random.randint(0, 2, n_test)

# Fit a simple logistic regression classifier
model = LogisticRegression(random_state=0, max_iter=1000)
model.fit(X_train, y_train)

ARGS = {
    'model': model,
    'X_train': X_train,
    'y_train': y_train,
    'X_test': X_test,
    'y_test': y_test,
    'compute_auc': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
exclude_non_predictive_columns Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/feature_selection/feature_columns.py
code served (permissive licence) · get_code("f06e4788cfa4d1a3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
from collections.abc import Sequence

# Create a sample DataFrame with various column types
X = pd.DataFrame({
    'id': [1, 2, 3],
    'feature_1': [0.5, 0.6, 0.7],
    'feature_2': [1.2, 1.3, 1.4],
    'cat_dummy_0': [1, 0, 0],
    'cat_dummy_1': [0, 1, 0],
    'cat_dummy_2': [0, 0, 1],
    'color_red': [1, 0, 1],
    'color_green': [0, 1, 0],
    'color_blue': [0, 0, 1],
    'continuous_feature': [2.1, 2.2, 2.3],
})

# Define columns to exclude (exact names)
columns_to_exclude = ['id']

# Define categorical prefixes to exclude
categorical_prefixes = ['cat_dummy_', 'color_']

ARGS = {
    'X': X,
    'columns_to_exclude': columns_to_exclude,
    'categorical_prefixes': categorical_prefixes,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
format_seconds Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/evaluation/local_morans_i.py
code served (permissive licence) · get_code("ec60cdd8563739f5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

# Construct arguments for format_seconds function

ARGS = {
    "seconds": 3661.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
infer_feature_group Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/feature_selection/importance_selection.py
code served (permissive licence) · get_code("50f27ea8bdda0251")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

# Test cases for infer_feature_group function
ARGS = {
    "column": "LithoGranite"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
resolve_selected_columns Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/modeling/train_zone_model.py
code served (permissive licence) · get_code("01949e9db177bf7c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Construct arguments for resolve_selected_columns
torch.manual_seed(0)

# available_factors: dict[str, str]
# Maps user-selected keys to factor names
available_factors = {
    "factor_0": "age",
    "factor_1": "gender",
    "factor_2": "income",
    "factor_3": "region"
}

# selected_keys: list[str]
# User-selected factor keys to look up in available_factors
selected_keys = ["factor_0", "factor_2"]

# columns: list[str]
# List of all column names (possibly one-hot expanded)
# These should include columns that start with the factor names
columns = [
    "age",
    "age_normalized",
    "gender_male",
    "gender_female",
    "income",
    "income_low",
    "income_high",
    "region_north",
    "region_south",
    "other_column"
]

ARGS = {
    "available_factors": available_factors,
    "selected_keys": selected_keys,
    "columns": columns
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
score_candidate Ran
ran, honoured the contract we drafted
GVCL/Susceptibility-Mapping-FL-Hetero/src/partitioning/select_grid_clusters.py
code served (permissive licence) · get_code("d945652875d1aed4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

ARGS = {
    "positive_count": 15.0,
    "negative_count": 10.0,
    "max_total_points": 50.0,
    "balance_weight": 0.6,
    "volume_weight": 0.4,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
select_prefixed_columns Ran
ran; our contract draft was wrong, not the code
GVCL/Susceptibility-Mapping-FL-Hetero/src/feature_selection/feature_columns.py
code served (permissive licence) · get_code("df572762231b28f4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
from collections.abc import Sequence

# Create a valid pd.Index with various column names
columns = pd.Index([
    'feature_a',
    'feature_b',
    'target_col',
    'metadata_x',
    'metadata_y',
    'other_field'
])

# Create prefixes to search for
prefixes = ['feature_', 'metadata_']

ARGS = {
    'columns': columns,
    'prefixes': prefixes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 11 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Floods and landslides often occur within the same regional landscapes, but their relationships with environmental controls vary across space. This study develops a spatial heterogeneity-aware framework for flood-landslide susceptibility and relative-risk mapping in the state of Kerala in India and Nepal. The framework combines 15 km x 15 km grid cells with region-specific contextual zones (from a zonation map) and compares two training strategies. These two strategies for selection of training data within this framework are proximity-gated cross-zone training (S1) and ecology-gated zone-constrained training (S2). S1 is designed to allow geographically nearby models to be assigned across contextual boundaries, whereas S2 is designed to restrict model development and assignment to the same contextual zone. Random Forest models are developed separately for flood and landslide using strategy-specific predictor sets and evaluated on spatially held-out test samples. The resulting susceptibility surfaces are integrated with CRITIC-weighted exposure and vulnerability indices to produce hazard-specific and nine-class bivariate relative-risk maps. S1 achieved higher mean accuracy, precision, recall, F1-score, AUC-ROC, and PR-AUC for both hazards and study regions. The largest difference occurred for Nepal flood susceptibility, where AUC-ROC increased from 0.728 under S2 to 0.886 under S1 and PR-AUC increased from 0.512 to 0.823. S2 nevertheless produced lower mean Brier scores for both Nepal hazards and retained zone-specific differences in predictor selection, SHAP rankings, and response patterns, particularly in Kerala. Both strategies reproduced the broad separation between flood-prone lowlands and landslide-prone uplands, but differed in the spatial patterns of susceptibility and risk classes. Overall likelihood of agreement between the bivariate risk maps was 0.521 in Kerala and 0.711 in Nepal, with allocation disagreement exceeding quantity disagreement in every S1-S2 risk-map comparison. Susceptibility-to-risk correspondence for the bivariate maps remained below 0.350, showing that exposure and vulnerability substantially changed the locations identified as priorities. The findings show that cross-zone learning supports stronger regional discrimination, while zone-constrained learning preserves differences among environmental settings. Hence, a regional multi-hazard system is recommended to combine both forms of model learning rather than rely on either strategy alone.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2608.08321")
get_code_for_paper("2608.08321")
have("2608.08321")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2608.08321.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2608.08321.svg)](https://syntology.ai/paper/2608.08321)

Connect an agent — have() is free.