Rainfall Leads Clues to Tripoli Soil Moisture

A multiple linear regression analysis of 297 monthly observations found that precipitation was the leading climatic correlate of average soil volumetric water content in Tripoli, Libya, over 2001–2024. Relative humidity and mean soil temperature across four depths were also associated with moisture in the fitted model. Together, the three retained variables accounted for about 48% of the observed variation, leaving slightly more than half of the monthly pattern unexplained by this compact statistical model.

The findings describe historical associations rather than controlled causal effects. The researchers analyzed ERA5 reanalysis data for the Tripoli area and found that average volumetric water content increased during the rainy season and declined during the dry season. This seasonal pattern was consistent with the correlations reported for rainfall, humidity and temperature. The study examined one regional series and did not test whether its fitted relationships remain stable under future climatic conditions or apply equally to individual fields, land uses or soil types.

Monthly ERA5 data were obtained from the Copernicus Climate Change Service Climate Data Store. The dependent variable was mean volumetric soil water content averaged across four depths. Candidate climatic variables initially considered for regression included relative humidity, precipitation, mean soil temperature, air temperature, wind speed, potential evaporation and groundwater depth. The workflow comprised descriptive analysis, construction of a correlation matrix, multicollinearity testing with the variance inflation factor, multiple linear regression, graphical examination of model assumptions, interpretation of coefficients and evaluation of prediction quality. The statistical work was performed in R.

The first model specification was simplified because several candidate predictors carried overlapping information. Air temperature had a variance inflation factor of 132, mean soil temperature had a VIF of 89 and potential evaporation had a VIF of 12, all above the threshold of less than 5 used by the researchers. These variables were excluded from the reduced model, which retained relative humidity, precipitation and mean soil temperature averaged over four depths. The screening addressed unstable parameter estimates caused by multicollinearity, but it did not establish that the excluded variables were unrelated to soil moisture or that the final model captured every relevant influence.

Relative humidity showed the strongest positive bivariate correlation with average volumetric water content, with r = 0.638. Precipitation was positively correlated with moisture at r = 0.528, whereas mean soil temperature was negatively correlated at r = −0.559. The retained predictors were themselves correlated: humidity and precipitation had a coefficient of 0.510, humidity and soil temperature had a coefficient of −0.742, and precipitation and soil temperature had a coefficient of −0.345. These bivariate relationships show that wetter and cooler periods tended to coincide in the record, but they do not by themselves demonstrate which physical process produced each correlation.

The estimated regression was written as average volumetric water content equal to 0.01378 plus 0.000070 multiplied by relative humidity, plus 0.000422 multiplied by precipitation in millimeters, minus 0.000098 multiplied by mean soil temperature in degrees Celsius. All three coefficients were statistically significant at p < 0.001. At face value, the precipitation estimate corresponds to an increase of 0.000422 in average volumetric water content for each additional 1 millimeter of precipitation when the other retained variables were held constant. This is a model estimate from the historical series, not an experimentally established response to rainfall.

Precipitation was identified by the researchers as the most influential variable in the reduced model, with a reported coefficient of β = 4.22 × 10−4. The coefficient for relative humidity was 7.0 × 10−5, and the coefficient for mean soil temperature was −9.8 × 10−5. The authors associated the positive humidity relationship with lower evaporative loss and the negative temperature relationship with greater evaporation and seepage. Those interpretations are consistent with the direction of the fitted associations, but the regression analysis did not independently isolate evaporation, infiltration or seepage as causal mechanisms.

Model-level statistics indicated that the retained climatic variables were jointly related to the monthly moisture measurements. The adjusted coefficient of determination was 0.477, indicating that the fitted model accounted for approximately 48% of the variation in average volumetric water content. The overall F-test was highly significant, with an F value of about 91, 293.3 degrees of freedom and p < 2.2 × 10−16. That result supports the model’s explanatory fit within the analyzed record, while the adjusted R² also defines a clear limit on its performance because more than half of the observed variation remained associated with omitted variables, model error or other unmeasured influences.

Residual diagnostics broadly supported the assumptions of the linear regression. In the residuals-versus-fitted plot, most points were scattered around the zero line without an obvious pattern, which the authors interpreted as consistent with homoscedasticity. A Q–Q plot showed an approximately normal residual distribution, with minor deviations at the extremes. Some outliers were present, but the researchers reported that they did not significantly affect the model’s suitability. The study nevertheless did not report independent out-of-sample validation, leaving its capacity as a forecasting tool untested.

Within the study’s limits, rainfall and relative humidity emerged as especially useful retained indicators of average soil water content, while soil temperature provided a negative association. The researchers recommend continued climatic-data collection to support analysis of climate change and water-resource conditions. They also advise testing time-series and nonlinear models in future work to seek improved predictive accuracy. Those steps are relevant because a regression fitted to 2001–2024 observations cannot by itself establish how relationships may change under projected precipitation decreases or temperature increases. The clearest result is therefore a historical statistical association: monthly soil moisture in Tripoli tracked rainfall, humidity and soil temperature together, with precipitation providing the largest reported contribution after the initial model was reduced to limit multicollinearity.

Subject of Research: Historical climatic associations with Tripoli soil volumetric water content

Article Title: Modeling soil volumetric water content based on climatic factors in the Tripoli region (Libya) using multiple linear regression

Article References: Ashargawi, A. S., & Miloud, H. M. B. (2026). Modeling soil volumetric water content based on climatic factors in the Tripoli region (Libya) using multiple linear regression. Discover Geoscience, 4(1), Article 408. https://doi.org/10.1007/s44288-026-00773-2

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00773-2

Keywords: soil moisture, Tripoli, Libya, precipitation, relative humidity, soil temperature, multiple linear regression, multicollinearity, ERA5, water resources, Modeling, soil