TERRAGEM – Soil Physicochemical Analysis Dataset

Dataset of 29 soil samples used as functional test cases for TERRAGEM, a Flutter mobile application that uses the Gemini AI API to generate agronomic diagnoses and fertilization recommendations. App repository: https://github.com/CarolayDS/TERRAGEM_APP Affiliation: Universidad Nacional de Cañete (UNDC), Lima, Peru Dataset Columns Column Description Unit Sample_ID Sample identifier — Textural_Class Full USDA textural class (Sandy Loam, Loam, Loamy Sand, Sandy Clay Loam, Silt Loam) — Soil_Type_App Simplified soil type for app selection (Loam / Sandy / Clay / Peat) — Sand_pct Sand content % Silt_pct Silt content % Clay_pct Clay content % CaCO3_pct Calcium carbonate % CE_dS_m Electrical conductivity dS/m pH Soil pH — Available_P_ppm Available phosphorus ppm Organic_Matter_pct Organic matter % Available_K_ppm Available potassium ppm Ca_meq Exchangeable calcium mEq/100g Mg_meq Exchangeable magnesium mEq/100g Na_meq Exchangeable sodium mEq/100g K_meq Exchangeable potassium mEq/100g Al_H_meq Exchangeable aluminum + hydrogen mEq/100g PAI_pct Aluminum saturation index % CICE_meq Effective cation exchange capacity mEq/100g Cu_ppm Available copper ppm Zn_ppm Available zinc ppm Mn_ppm Available manganese ppm Fe_ppm Available iron ppm B_ppm Available boron ppm pH_Level pH interpretation Acid / Neutral / Alkaline Salinity_Level Salinity level Normal / Slightly Saline / Saline Notes: N/A values in M-28 and M-29 indicate parameters not reported in the original lab analysis (special cases: calcareous and saline-sodic soils). CaCO3_pct shows <0.01 for samples below detection limit. PAI_pct in M-29 shows PSI:67.62 (Sodium Percentage Index) as it is a saline-sodic soil. How to Use This Dataset in the TERRAGEM App You can use any row from this dataset to test the TERRAGEM app manually. Step 1 – Install the app Clone and run the app from: https://github.com/CarolayDS/TERRAGEM_APP Step 2 – Select a crop Choose any crop available in the app: Sweet Potato, Corn, Avocado, or Grape. (crops shown in English in the app) The dataset does not include crop information — the samples are soil-only analyses. Step 3 – Enter your plot area Enter any value (e.g. 0.5 hectares). Step 4 – Select soil type Use the Soil_Type_App column value from your chosen row: Soil_Type_App value Select in app Loam Loam Sandy Sandy Clay Clay Peat Peat Step 5 – Enter soil parameters Copy the values from your chosen row into the app fields: App field Dataset column pH pH Electrical conductivity CE_dS_m Organic matter Organic_Matter_pct Available phosphorus Available_P_ppm Available potassium Available_K_ppm Calcium Ca_meq Magnesium Mg_meq Sodium Na_meq Effective CEC CICE_meq Copper Cu_ppm Zinc Zn_ppm Manganese Mn_ppm Iron Fe_ppm Boron B_ppm Step 6 – Get the diagnosis Press Save and View Analysis. The Gemini API will generate a soil diagnosis, identify the main problems, and provide fertilization recommendations. Suggested Test Cases Sample Notable characteristic Expected diagnosis M-01 Very acid pH (3.75), high Al saturation (68%) Severe acidity, aluminum toxicity M-12 Acid pH (4.45), low Al saturation (10%) Moderate acidity, good base saturation M-20 Near-neutral pH (4.68), high K (236 ppm) Mild acidity, good potassium levels M-28 Alkaline pH (8.48), high CE (20.46 dS/m) Calcareous soil, salinity stress M-29 Near-neutral pH (7.87), extreme CE (178.4 dS/m) Severe salinity, sodic soil Potential Use Cases for Modeling Soil textural class classification from physicochemical parameters pH level prediction (Acid / Neutral / Alkaline) from granulometry and exchangeable cations Clustering of soil profiles by fertility characteristics Feature importance analysis for agronomic diagnosis License & Attribution This dataset is published under Creative Commons Attribution 4.0 International (CC BY 4.0). Data has been anonymized in compliance with source confidentiality restrictions. Original lab analyses were used exclusively as functional test cases for the TERRAGEM system. If you use this dataset, please cite this repository: DOI: : https://doi.org/10.5281/zenodo.19103772