Holwerda et al. (2025) A comparison of drought indices for crop yield loss detection: The role of green-up onset alignment and spatial resolution

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2025
• Date: 2025-11-26
• Authors: F. Holwerda, Diego Salazar-Martínez, Thomas Holmes, Christopher Hain, Martha C. Anderson
• DOI: 10.1016/j.ejrh.2025.102939

Research Groups

• Instituto de Ciencias de la Atm´osfera y Cambio Clim´atico, Universidad Nacional Aut´onoma de M´exico, Mexico City, Mexico
• Sustainable Agricultural Water Systems Unit, USDA-ARS, Davis, CA, United States
• NASA Goddard Space Flight Center, Greenbelt, MD, United States
• NASA Marshall Space Flight Center, Huntsville, AL, United States
• Hydrology and Remote Sensing Laboratory, USDA-ARS, Beltsville, MD, United States

Short Summary

This study compares six drought indices for detecting rainfed crop yield loss in the Central Plateau of Mexico, finding that aligning indices to satellite-derived green-up onset improves performance, with ALEXI-based Evaporative Stress Index and MODIS NDVI anomaly showing the strongest relationships with yield anomalies.

Objective

• To examine the performance of different drought indices for detecting agricultural drought in the Central Plateau of Mexico, specifically evaluating the impact of green-up onset alignment and spatial resolution on their ability to detect crop yield loss.

Study Configuration

• Spatial Scale: 82 municipalities in the Central Plateau of Mexico (18°N to 21°N), ranging from 4 km² to 492 km² in area. Drought indices were analyzed at their native spatial resolutions (e.g., 0.05°, 0.1°, 500 m) and at a common 0.1° grid.
• Temporal Scale: 2003–2023, with data from 2012 and 2013 excluded due to the MasAgro program's impact. Crop yield data had a yearly resolution, while drought indices were calculated using monthly or 4-week moving average time steps.

Methodology and Data

• Models used:
  • Evaporative Stress Indices (ESIs) derived from:
    • ALEXI (Atmosphere-Land Exchange Inverse model) relative evapotranspiration fraction (fRET)
    • MOD16 (MODIS) actual evapotranspiration (ET) and potential evapotranspiration (PET)
    • GLEAM (Global Land Evaporation Amsterdam Model) evaporative stress factor (S)
  • MODIS Normalized Difference Vegetation Index (NDVI) anomaly (NDVIanom)
  • Standardized Precipitation Index (SPI)
  • Mexico Drought Monitor (MDM)
• Data sources:
  • Crop yield data: Municipal-level agricultural production and sown/harvested area from the Agrifood and Fisheries Information Service (SIAP), Mexico.
  • Remote sensing ET data:
    • ALEXI fRET: MODIS MYD11C1 (land surface temperature), NCEP CFSv2 (potential temperature gradient, longwave radiation), CERES SYN1deg (incoming shortwave radiation), MODIS MCD43B3 (albedo).
    • GLEAM v4.2a: Satellite soil moisture (ESA CCI), vegetation optical depth (VODCA), leaf area index (MODIS), air temperature and vapor pressure deficit (reanalysis), multisource precipitation.
    • MOD16A2GF v061: MODIS albedo, fraction of photosynthetically active radiation (FPAR), land cover type, leaf area index (LAI), daily meteorological reanalysis.
  • Remote sensing vegetation index data: MODIS MOD13A1 v061 (NDVI).
  • Precipitation data: Rain gauge data from the National Meteorological Service of Mexico (SMN) climatological stations, spatially interpolated using Inverse Distance Weighting (IDW).
  • Drought severity classes: Mexico Drought Monitor (MDM) from SMN-CONAGUA, a composite product integrating various indicators.
  • Land surface phenology data: MODIS MCD12Q2 version 6.1 (green-up onset date from EVI2).

Main Results

• Significant delays in satellite-derived green-up onset were observed during drought years, correlating with postponed sowing dates.
• Aligning drought variables to green-up onset improved the temporal synchronization of crop growth stages across years.
• For corn, the strongest relationships between yield anomalies and cumulative NDVI anomaly/evaporative stress indices occurred during the mid-season growth stage (four to five months after green-up onset).
• Degrading the spatial resolution of remote sensing data generally weakened the drought index-yield anomaly relationships, particularly for the NDVI anomaly.
• At their native spatial resolutions, the ALEXI-based Evaporative Stress Index (ESIALEXI) and MODIS NDVI anomaly (NDVIanom) showed the strongest relationships with crop yield anomalies:
  • For corn: ESIALEXI (r² = 0.75), NDVIanom (r² = 0.64).
  • For all crops: NDVIanom (r² = 0.81), ESIALEXI (r² = 0.67).
• ESIMODIS showed strong relationships for corn (r² = 0.78) but moderate for all crops (r² = 0.44).
• ESIGLEAM showed moderate correlations with yield anomalies (r² = 0.54 for corn, r² = 0.61 for all crops).
• The Standardized Precipitation Index (SPI) and the Mexico Drought Monitor (MDM) displayed the weakest relationships with crop yield anomalies (SPI: r² = 0.27 for corn, r² = 0.60 for all crops; MDM: r² = 0.33 for corn, r² = 0.35 for all crops).
• The MasAgro program (implemented in 2012) resulted in substantial yield increases (approximately 24% for all crops and 33% for corn) and reduced unharvested areas during drought years.

Contributions

• This study provides a comprehensive intercomparison of six diverse drought indices, including multiple remote sensing-based evapotranspiration products, for agricultural drought detection in the Central Plateau of Mexico.
• It demonstrates the critical importance of aligning drought variable time series to satellite-derived green-up onset dates to improve the accuracy and relevance of within-season drought impact assessments on crop yields, particularly for corn.
• The research quantifies the impact of spatial resolution on drought index performance, highlighting that using native, finer-resolution data generally leads to stronger relationships with crop yield anomalies.
• It identifies the ALEXI-based Evaporative Stress Index and MODIS NDVI anomaly as the most effective indices for detecting agricultural drought in the study region, while SPI and the Mexico Drought Monitor showed weaker performance.

Funding

• D.S.M. was supported by a graduate scholarship from CONACYT, Mexico (number 595847).

Citation

@article{Holwerda2025comparison,
  author = {Holwerda, F. and Salazar-Martínez, Diego and Holmes, Thomas and Hain, Christopher and Anderson, Martha C.},
  title = {A comparison of drought indices for crop yield loss detection: The role of green-up onset alignment and spatial resolution},
  journal = {Journal of Hydrology Regional Studies},
  year = {2025},
  doi = {10.1016/j.ejrh.2025.102939},
  url = {https://doi.org/10.1016/j.ejrh.2025.102939}
}