The Measurement Problem: Why a Single Index Is Never Enough
Ask ten climate scientists how to measure drought and you will get ten different answers. The Standardized Precipitation Index (SPI) captures rainfall deficits but ignores heat. The Palmer Drought Severity Index (PDSI) accounts for evapotranspiration but was calibrated on North American soil moisture regimes that bear little resemblance to the tropical monsoon belt of Southeast Asia. The Vegetation Condition Index (VCI) reveals crop stress visible from satellite but arrives weeks after the damage has begun.
Each index illuminates one dimension of a fundamentally multi-dimensional phenomenon. A drought in Thailand's Chao Phraya basin behaves differently from one in the Mekong Delta of Vietnam or the Kelantan river corridor in Malaysia — different rainfall regimes, different land-use pressures, different groundwater recharge rates. A single-variable index applied uniformly across this diversity produces false negatives that leave governments and enterprises blind to emerging risk.
This is the core problem that GlobMaps designed the Multi-Dimensional Index (MDI) to solve. Today, as we launch live province-level drought intelligence across 156 provinces in Thailand, Vietnam, and Malaysia, we want to open the hood and show exactly how MDI works — from raw satellite and reanalysis data to the risk scores delivered through our API.
The Data Foundation: ERA5 Climate Reanalysis
Every MDI calculation begins with ERA5, the fifth-generation global climate reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). ERA5 provides hourly estimates of atmospheric, land, and oceanic climate variables at 31 km horizontal resolution, covering the full globe from 1940 to near-real-time. For Southeast Asia — a region with notoriously sparse ground-based observation networks — ERA5 is the closest thing to a complete historical climate record available at provincial scale.
Our data pipeline downloads monthly ERA5 aggregates for each province centroid in our coverage area. The key variables extracted are: total precipitation (tp), 2-metre temperature (t2m), surface net solar radiation (ssr), and soil water volume in layers 1 through 4 (swvl1–swvl4). These are the physical inputs from which all downstream indices are computed.
The bridge runs on the 8th of each month. Within approximately 30 minutes, ERA5 data for the previous month is pulled, processed through the MDI pipeline, and written to the production database. This monthly cadence reflects ERA5's own release schedule: the preliminary near-real-time dataset (ERA5T) typically achieves full global coverage within five days of the end of the preceding month, giving us a reliable update window.
Computing the Index: From Raw Variables to MDI
MDI is a composite index built from three independently computed sub-indices, each capturing a different dimension of drought stress. The final MDI score is a weighted average calibrated to the climatic characteristics of Southeast Asia's two principal Köppen–Geiger zones: the tropical monsoon climate (Am) dominating Thailand's central plains and Vietnam's coastal lowlands, and the tropical rainforest climate (Af) covering peninsular Malaysia and parts of the Vietnamese highlands.
Sub-Index 1: Precipitation Deficit (SPEI-3)
The Standardized Precipitation-Evapotranspiration Index calculated over a 3-month accumulation window (SPEI-3) is the precipitation dimension of MDI. Unlike SPI, SPEI incorporates potential evapotranspiration — the atmospheric demand for water — computed via the Hargreaves equation from ERA5 temperature and radiation data. SPEI-3 captures the short-to-medium-term rainfall deficits most relevant to agricultural and municipal water supply stress.
Values are standardized against a 1981–2010 baseline climatology for each province, producing a dimensionless Z-score where 0 represents median historical conditions, −1.0 indicates moderate drought, −1.5 severe drought, and −2.0 or below extreme drought. This standardization is critical: a −1.5 SPEI-3 in Chiang Rai, Thailand carries the same probabilistic interpretation as −1.5 in Đắk Lắk, Vietnam, even though their absolute rainfall totals differ by a factor of three.
Sub-Index 2: Vegetation Stress (VCI-derived)
Precipitation deficits become economically material when they translate into vegetation stress. MDI incorporates a Vegetation Condition Index analog derived from ERA5 soil moisture anomalies in the root zone (layers 2 and 3, representing 7–100 cm depth). Root-zone soil moisture is a more direct driver of vegetation stress than surface precipitation because it integrates antecedent conditions over several weeks — a wet preceding month can buffer a dry current month at the root level even as surface SPEI shows deficit.
The soil moisture anomaly is computed relative to the same 1981–2010 baseline, expressed as a percentile rank. A province scoring below the 20th percentile of its historical root-zone soil moisture is flagged as under significant vegetation stress — a threshold empirically validated against historical rice yield loss data in Thailand's northeastern provinces (Isaan), where MDI pilot testing showed 0.78 correlation with NESDC agricultural output reports.
Sub-Index 3: Thermal Amplification (Heat Stress Modifier)
The third dimension captures a dynamic that purely precipitation-based indices miss entirely: temperature amplification of drought. During El Niño years and under the warming trend accelerating across mainland Southeast Asia — mean temperatures in Thailand have risen approximately 0.3°C per decade since 1980 — evapotranspiration demand increases even when rainfall is near-normal. The result is a "hot drought" that depletes soil moisture far faster than historical relationships between precipitation and soil water would predict.
MDI's thermal modifier computes monthly mean 2-metre temperature anomalies relative to the 1981–2010 baseline. Positive temperature anomalies amplify the weight assigned to SPEI-3 in the final composite. During months where temperature anomaly exceeds +1.5°C, the SPEI-3 component weight increases from 0.45 to 0.60, reflecting the disproportionate evaporative demand that makes precipitation deficits more severe. This dynamic weighting is one of MDI's key differentiators from static multi-variable composites used in older drought monitoring systems.
Province Coverage: Thailand, Vietnam, and Malaysia
The current MDI release covers 156 administrative units across three countries, chosen for their combination of climate risk exposure and economic significance within GlobMaps' initial Southeast Asia mandate.
Thailand — 77 Provinces
Full national coverage of Thailand's 77 provinces, encompassing the four major climatic subregions: the humid central plains (Chao Phraya basin), the semi-arid northeast plateau (Isaan), the wet peninsular south (shared monsoon with Malaysia), and the seasonally dry northern highlands. Thailand is the highest-priority market in the initial rollout given its centrality to ASEAN agricultural trade — it is the world's second-largest rice exporter, and Isaan alone accounts for roughly one-third of national rice production. Drought risk in this region has direct implications for regional food security and commodity price volatility.
Vietnam — 63 Provinces
Vietnam's 63 provinces present the most climatically diverse challenge in our coverage area. The country spans 15 degrees of latitude, incorporating a dry tropical climate in the south (Ho Chi Minh City region), a distinct dry season in the central highlands (Đắk Lắk, Gia Lai), and a humid subtropical regime in the Red River Delta north of Hanoi. MDI calibration for Vietnam required separate baseline climatologies for five distinct climate zones, making it the most technically complex country in the initial dataset. This diversity also makes MDI valuable here: single-index systems applied nationally produce severe misclassification in the central highlands, where traditional SPEI values consistently under-represent drought stress due to the region's bimodal monsoon seasonality.
Malaysia — 16 States and Federal Territories
Malaysia's 16 administrative units present the opposite challenge from Vietnam: relatively low inter-unit climatic variance in Peninsular Malaysia, but a sharp discontinuity between the Peninsula and the Borneo states of Sabah and Sarawak. Peninsular Malaysia lies fully within the tropical rainforest climate zone and receives rainfall year-round, making drought a less frequent but higher-impact event. El Niño-driven dry spells — such as those of 2015–2016 and 2019 — caused significant reservoir drawdown, hydropower output reductions, and forest fire outbreaks in typically wet regions. MDI captures these episodic drought events precisely because its SPEI-3 component is calibrated to the local rainfall distribution, not global averages.
Reading the API Response: What MDI Scores Mean in Practice
MDI scores are delivered via the GlobMaps REST API on a 0–100 scale, where higher values indicate more severe drought conditions. The scale is designed to map directly to operational response thresholds used by government agencies and enterprise risk managers:
- 0–20 (Normal): Conditions within 1 standard deviation of the 1981–2010 baseline. No significant drought stress.
- 21–40 (Watch): Mild deficit conditions developing. Appropriate for early-warning monitoring by agricultural agencies. Typically corresponds to SPEI-3 between −0.5 and −1.0.
- 41–60 (Warning): Moderate drought. Root-zone soil moisture trending below 40th percentile. Crop yield impact probabilities above 15% for rain-fed agriculture.
- 61–80 (Alert): Severe drought. SPEI-3 below −1.5, soil moisture below 20th percentile. Reservoir drawdown, irrigation demand exceeding supply, and urban water restriction risk.
- 81–100 (Emergency): Extreme drought. SPEI-3 below −2.0 with thermal amplification. Conditions comparable to the worst drought years in the 1981–2010 record. Humanitarian and economic impact certain without intervention.
Each API response includes the composite MDI score alongside the three sub-index values (SPEI-3, soil moisture percentile, temperature anomaly), enabling downstream systems to understand which dimension is driving the risk. A province scoring 65 because of a precipitation collapse behaves differently — and requires different interventions — than one scoring 65 because of a thermal amplification event on otherwise near-normal rainfall.
What Comes Next: Expanding the Intelligence Layer
The June 2026 launch represents the production baseline for MDI. The pipeline is designed for iterative expansion — more frequent updates, probabilistic forecast horizons, and broader geographic coverage are on the roadmap as the platform grows.
The core insight driving all of this is simple: drought risk is not a binary condition, and it cannot be captured by any single number derived from any single data source. The goal of MDI is not to replace expert judgment — it is to give that judgment a richer, more dimensionally complete picture of what is happening on the ground, province by province, month by month, across one of the world's most climate-exposed regions.
If you are working in agricultural finance, government water resource management, insurance underwriting, or enterprise supply chain risk in Southeast Asia, MDI is available now through the GlobMaps platform. Contact our team to discuss API integration and enterprise access.