LANDMauro Esposito

Method and sources

How the satellite analysis was done, what it shows and what it doesn't, and where every figure in the paper comes from.

What was measured

The analysis compares two strips of land along the wall: the strip the wall occupies, 40 metres on each side of its route, and a strip of land between 150 and 400 metres away. The second works as a reference. If a year is dry, both strips lose vegetation; if only the first one drops, something happened along the route.

For each year from 1985 to 2025, the median vegetation index (NDVI) for March and April was calculated from Landsat 4, 5, 7, 8 and 9 surface reflectance images (Collection 2), after masking clouds and cloud shadows. Spring is when natural vegetation is at its greenest, so a bare strip stands out. Values from Landsat 4, 5 and 7 were adjusted to be comparable with Landsat 8 and 9 using the coefficients of Roy et al. (2016).

What is NDVI?

An index of how much living vegetation there is in a spot, calculated by the satellite from the light reflected by the ground. Close to zero means bare soil, rock or concrete; 0.2 to 0.35, sparse vegetation; 0.35 to 0.5, grassland, olive groves and fields; above 0.6, dense vegetation.

Results

Wall strip compared with land 150–400 m away1985–20022004–20252021–2025
Qalqilya+2%−29%−30%
Tulkarm−0.3%−24%−17%

2003, the year of construction, is left out of the averages. Before the wall the two strips were practically identical. In Tulkarm the gap has narrowed in recent years, probably because some vegetation has grown back along the edges of the strip, but it has never closed.

Qalqilya

Tulkarm

Spring NDVI (March–April), 1985–2025. Analysis by the author on Landsat data (USGS/NASA).

Limits

  • The satellite measures vegetation, not biodiversity. It shows that a habitat was removed and has not come back; the effects on species come from field studies such as those by Husein and Qumsiyeh.
  • Each Landsat pixel covers 30 by 30 metres, so the strip mixes the bare ground of the wall with some vegetation along its edges. The real loss inside the cleared strip is larger than the figures show.
  • NDVI is not proportional to the amount of plants: an index 29% lower does not mean 29% fewer plants.
  • The two stretches of the wall were traced on satellite imagery and checked visually. The next step is to extend the analysis to the whole route, using OCHA's official data.

Code and data

Sources