📄 earth_science

Satellite-Based Assessment of Vegetation Water Stress Across West Africa Using Evapotranspiration Limitation and Independent Rainfall Evaluation from 2001 to 2023

This study presents a scalable, satellite-based framework for monitoring crop water stress across West Africa from 2001 to 2023, demonstrating that a parsimonious evapotranspiration limitation index (1 − ET/PET) outperforms temperature-based and vegetation-enhanced metrics in correlating with independent rainfall anomalies and effectively capturing the region's strong north-south moisture gradient.

Original authors: George Owus Amoah

Published 2026-07-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: George Owus Amoah

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine West Africa as a giant, living thermostat where the land is trying to cool itself down, but the water supply is playing hide-and-seek. A researcher named George Owusu Amoah decided to build a super-smart, satellite-powered "stress detector" to see exactly how thirsty the crops and plants are across this massive region, from the rainy, lush south to the dry, dusty north. He didn't just guess; he used a massive digital playground called Google Earth Engine to crunch data from 2001 all the way to 2023.

The Big Discovery: The North is Drier, the South is Greener
The most obvious thing the satellite saw was a giant temperature and moisture gradient. Think of it like a ladder. At the bottom (the humid south), the land is cool and lush, with plants drinking up water like a sponge. At the top (the semi-arid north), the land is baking hot, with plants struggling to find any moisture at all.

The data showed a stark contrast:

Temperature: The average land surface temperature in the humid south was 30.94 °C, but it skyrocketed to 40.55 °C in the semi-arid north. That's a nearly 10 °C difference!
Water Usage: The amount of water plants actually released into the air (evapotranspiration) dropped dramatically as you went north. In the south, it was a massive 1125.95 mm, but in the north, it plummeted to just 156.93 mm. That's a sevenfold drop!
Stress Levels: The new "stress meter" (which the paper calls the Crop Water Stress Index) showed that plants in the humid south were only about 0.369 stressed, while those in the dry north were nearly maxed out at 0.940.

The "Stress Meter" Trick
Usually, scientists try to measure plant stress by looking at how hot the leaves are. But in West Africa, the air is so humid and the plants are so different that just looking at temperature can be misleading. It's like trying to guess how thirsty a person is just by looking at their face; sometimes they look hot because they ran a race, not because they need water.

Instead, George built a new meter based on a simple ratio: Actual Water Lost vs. Potential Water Lost.

The Formula: He calculated 1 − ET/PET.
The Logic: If a plant is drinking and sweating as much as the air wants it to, the stress is low. If the air is screaming "Give me water!" but the plant can only give a tiny drop because the soil is dry, the stress score goes up.
The Result: This new meter worked better than the old temperature-based ones. When checked against independent rainfall data (the "truth"), the new meter matched the rain patterns much better (a correlation of -0.29) than the old temperature method (which only matched at -0.05).

What the Paper Ruled Out: The "Green Leaf" Myth
Here is where it gets interesting. Many people assume that if you see a plant looking greener (measured by something called NDVI), it must be less stressed. George tested this idea like a scientist in a lab. He asked: "If we mix our water-stress meter with a 'greenness' meter, does it get better?"

The answer was a hard no.

The paper explicitly found that adding the greenness factor actually made the stress meter worse.
The "greenness" alone had almost zero connection to how much rain was falling (a correlation of -0.01).
When they multiplied the water stress by the greenness, the signal got weaker (dropping to -0.09).
The Verdict: A plant can look green but still be thirsty, or look brown but be doing okay. The paper concludes that for this region, you don't need the greenness factor to measure water stress; the water ratio alone is the champion.

The Time Travel Test: Did Things Change?
The study looked at the last 23 years (2001–2023) to see if the situation was getting better or worse.

The Good News: The vegetation actually got greener over time. The paper says this trend is statistically significant (meaning it's a real change, not just a fluke), with a p-value of 0.001. This fits with stories of the Sahel "re-greening" after past droughts.
The Reality Check: Even though the plants got greener, the actual water stress did not change significantly. The stress index showed no statistically significant trend (p = 0.13).
The Takeaway: Just because the plants look greener doesn't mean the water situation has improved. The paper suggests that the moisture flux (the actual water moving) hasn't changed enough to lower the stress levels, even if the plants are trying to grow back.

How Sure Are We?
The paper is very confident in the spatial results (the north-south differences). The data clearly shows the gradient, and the new stress meter is proven to track rainfall better than the old temperature tricks.

However, the paper is careful about the temporal (time-based) results. It doesn't claim the region is "solved" or that drought is gone. It explicitly states that while the greening is a real, measured trend, the stress levels have remained stubbornly stable. The new method is a proven, scalable tool for data-scarce areas, but it relies on satellite estimates, not ground measurements (which are rare in this region).

In short, the paper gives us a clearer, more accurate way to see where the plants are thirsty in West Africa, proves that "green" doesn't always mean "hydrated," and shows that while the landscape is getting greener, the water stress is still a persistent, unchanging challenge in the dry north.

Technical Summary: Satellite-Based Assessment of Vegetation Water Stress Across West Africa (2001–2023)

Problem Statement
Assessing crop water stress in tropical and semi-arid regions like West Africa is complicated by the complex interplay of surface temperature, vegetation condition, evapotranspiration, and atmospheric moisture demand. Conventional approaches, such as the Crop Water Stress Index (CWSI) based on canopy temperature (Eq. 1), rely on field-measured "wet" and "dry" reference temperatures. These references are rarely available across large, heterogeneous regions, rendering the standard CWSI difficult to apply at a regional scale in data-scarce environments. Furthermore, existing satellite-based stress indices are often validated against quantities derived from their own inputs (e.g., vegetation or temperature fields), leading to circular agreement rather than genuine skill. There is a critical need for a scalable, threshold-free framework that is computed entirely from open satellite data and validated against independent measures of moisture supply.

Methodology
This study developed a fully satellite-based framework within Google Earth Engine (GEE) to monitor crop water stress across West Africa from 2001 to 2023. The core methodology involves:

Data Sources: The study utilized Landsat Collection 2 Level-2 data for Land Surface Temperature (LST) and Normalised Difference Vegetation Index (NDVI), and the MODIS MOD16A2GF product for Actual Evapotranspiration (ET) and Potential Evapotranspiration (PET). CHIRPS rainfall data served as an independent validation metric.
Primary Stress Metric: Instead of canopy temperature thresholds, the study defined the CWSI directly from the evapotranspiration ratio:
$CWSI = 1 - (ET / PET)$
This formulation quantifies the limitation of actual evapotranspiration relative to atmospheric demand without requiring field-measured reference temperatures.
Hypothesis Testing on Vegetation Terms: The study explicitly tested the common assumption that combining a moisture term with a vegetation index improves stress characterization. A vegetation-weighted variant, $(1 - ET/PET) \times (1 - NDVI)$ , was evaluated against the parsimonious ET/PET formulation.
Validation Strategy: To avoid circularity, the derived index was not validated against its own inputs. Instead, it was:
1. Benchmarked against the established Temperature–Vegetation Dryness Index (TVDI).
2. Validated against an independent moisture measure: the standardised CHIRPS rainfall anomaly (z-score).
3. Trend Analysis: Temporal trends (2001–2023) were assessed using the non-parametric Mann–Kendall test and Sen's slope estimator.
Zonal Analysis: The region was divided into three climatic zones: the humid Guinea Coast (south of 8°N), the transitional Sudanian savanna (8–12°N), and the semi-arid Sahel (north of 12°N).

Key Results

Spatial Gradients: A pronounced north–south gradient was observed. Mean CWSI increased from 0.369 in the humid zone to 0.940 in the semi-arid Sahel. This corresponded with a rise in mean LST from 30.94°C to 40.55°C and a roughly sevenfold decline in ET (from 1125.95 mm to 156.93 mm).
Validation Performance: Against independent rainfall anomalies, the ET/PET-based CWSI demonstrated superior skill compared to the TVDI. The median per-pixel temporal correlation was $r = -0.29$ for the ET/PET index versus $r = -0.05$ for the TVDI.
Vegetation Term Analysis: The inclusion of a vegetation term did not improve the index. The vegetation term alone ($1 - NDVI$) was uncorrelated with rainfall ( $r = -0.01$ ), and the multiplicative combination weakened the signal of the moisture term ( $r = -0.09$ ). Consequently, the parsimonious ET/PET formulation was retained.
Temporal Trends: Over the 23-year period, NDVI showed a statistically significant increasing trend ( $p = 0.001$ ), indicating vegetation greening. However, the ET/PET-based stress index showed no statistically significant trend ( $p = 0.13$ ). This indicates that while vegetation cover has increased, the underlying moisture flux (and thus water stress) has not significantly changed, and the greening signal does not mechanically propagate into the stress index.

Significance and Claims
The paper claims three primary contributions:

Methodological Advancement: It provides a scalable, threshold-free CWSI that is computable entirely from open satellite data, removing the dependency on unavailable field reference temperatures.
Evidence-Based Design: Through direct testing against independent rainfall, the study demonstrates that adding a vegetation term to the moisture ratio does not improve, and slightly degrades, the characterization of water stress in this region. The retention of the parsimonious ET/PET index is presented as an evidence-based design choice rather than an assumption.
External Validation: The framework is benchmarked against an established index (TVDI) and validated against an external moisture reference (CHIRPS), showing that the ET/PET-based index tracks interannual moisture variability more closely than temperature-based formulations.

The study concludes that the resulting framework offers an interpretable and externally validated basis for drought monitoring in data-scarce tropical environments, highlighting that the dominant feature of the region's hydro-climatic state is a persistent spatial stress gradient rather than a significant temporal trend in water stress.

Technical Summary: Satellite-Based Assessment of Vegetation Water Stress Across West Africa (2001–2023)

More like this