Survey data analysis
The survey data were analysed using a multivariate approach. Specifically, the polythetic divisive classification procedure Two-Way Indicator Species Analysis (TWINSPAN: Hill 1979; Gauch 1982) was applied to simultaneously group samples, and species present in the samples. This strategy allowed classifying samples based on similar vegetation types and identified species that tend to occur in the same sample, producing a species classification that identifies the main assemblages present in the dataset. Moreover, the analysis distinguished indicator species which characterize sample-groups produced by the classification (Ali et al. 2000; Al-Shehabi & Murphy 2017).
Then, ANOVA tests with mean separation by Tukey’s least significant difference test (significant outcomes only: P<0.05) were applied to examine differences between plant and environmental variables, both between geographical areas, and between sets of samples making up sample-groups identified by TWINSPAN classification. Datasets of both the plant-related data, and environmental data were tested for normality (Ryan-Joiner test) and log10-transformed if necessary prior to the use of ANOVA. In cases that the data could not be normalized by transformation, the non-parametric Kruskal-Wallis test was applied instead of ANOVA.

Field experiment: growth of Haloxylon salicornicum in oil-polluted soils

A controlled field experiment was carried out in in 2013 with the aim to: (1) determine the tolerance of H. salicornicum to desert soil contaminated with petroleum hydrocarbons, (2) assess factors important for growing native plant species in contaminated desert soils such as ease of handling for operational purposes, and (3) determine requirements for environmental amendment(s) to assist plant establishment and growth.
Experimental plots were located at two of the areas where oil-polluted soils were found during an earlier vegetation survey. One site was located in north of Kuwait (Bahra) and the other in south (Burgan). Two-year old H. salicornicum plants (grown in a greenhouse of the Public Authority of Agriculture and Fisheries Resources (PAAF) Kuwait) was utilized as a representative native desert plant species. In both locations, plants were grown according to an identical experimental setup that comprised ten randomly located plots (Hurlbert 1984) within a fenced area of weathered contaminated soil (in an old dry oil lake). Three replicates were used per plot. As a control, a second set of plants were similarly planted at a nearby similar-sized fenced area of clean soil at both locations. Plants were planted in a 20-30 cm deep hole (depending on specimen size). Fencing was required to keep out grazing animals, especially camels and goats. Each plot was watered prior to planting, and a 1.0 litre supply of “Driwater™” (http://driwater.com/what-is-driwater/) was installed per plot. “Driwater™” is a cellulose gum source of water that dissolves at a constant rate and when installed into the soil adjacent to a plant. It releases water to the root system for up to 70 days.
Soil samples were collected from both oil-polluted plots (n = 10 per sampling area) and were analysed using a standard GC analytical procedure, which was carried out at the Kuwait University Laboratory for Bioremediation Research. The field experiment began 31 January 2013 in Bahra and ended 28 March 2013 with 9 field trips to collect data. In Burgan the start date was 5 February 2013, and ended 3 April 2013 after 10 sampling trips. In both cases, sampling was undertaken during 7 - 9 day intervals and plant height was measured to determine incremental growth rate. At the end of the experiment, the plants were carefully removed from the soil and transported to the lab for measuring their biomass (shoot, root and total dry weight per plant). The TPH in plant tissues were quantified using a procedure described above.
Results

The distribution of plants and hydrocarbon contamination

The semi-quantitative Oil Damage Score (ODS), shown in Fig. 2, indicates substantial variations between the examined sites in pollution (Kruskal-Wallis test, P<0.004). Um Al-Aish, Sabriya, Burgan, and Um Ghadair oil fields were highly impacted by massive oil pollution, whereas Bahra and Sabah Al-Ahmad Protected Area are less influenced, and Um Al-Rros Military Base was generally unaffected. The variation in hydrocarbon contamination was strongly associated with the geographical occurrence of the three dominant species (Haloxylon salicornicum, Cyperus conglomeratus, and Rhanterium epapposum), seen in the survey. All three species were found in both hydrocarbon contaminated areas (often in oil-polluted soils) and undamaged areas (Table. 1, Fig 3).