2.1. Study area description
This study was conducted on Yabelo forest and its surrounding. The forest is located in Yabelo district of Borana Zone, southern Oromia. The total area of the study sites is 26,956.7ha and its altitude range between 1438 and 2354 m. m.s.l (SRTM, 2013). The absolute location of this area lies between 4° 45' 0" N to 4° 57' 0" N and 38° 1' 0" E to 38° 10' 30" E (Fig. 1). Yabello district has a bi-modal rainfall pattern with the main rainy season between March and May with the high precipitation in April (Coppock 1994). The short rainy season extends from September to November with a high precipitation in October. The region has a semi-arid savannah landscape, marked by gently sloping lowlands and flood plains vegetated predominantly with grass and bush land. According to Coppock (1994), this study area comprise three main soils types, 53% loam soil, 30% black clay and volcanic light colored silty clay and 17% silty. Five major land use/cover, namely; grassland, woody vegetation, cultivated, settlement and bare land were identified in the district (Gifawesen et al. 2020; Jaleta 2010). Evergreen and semi evergreen bush land and thickets, rangeland dominated by shrubby Acacia, Commiphora and allied genera, and dwarf shrub grassland to shrub grassland are also found in the area (Zeleke 2009). The district has a total population of about 128,762 (64,692 Male and 64,070 Female) (CSA, 2013).
2.2 Data sources
In the present study, Landsat 8 OLI/TIRS with 30-meter spatial resolution (www.usgs.gov) was used as the data source to calculate NDVI and used to generate LULC, from the year 2014 to 2022. Total of twenty-two Landsat data were downloaded during the short rainy season (September to November) and for each month cloud free satellite were downloaded and the mean of NDVI were used for each year. This season was selected because of its higher rainfall variability (Worku, et al. 2022). Additionally, the most common statistical methods applied to time series of NDVI and precipitation, such as simple linear correlation or regression analyses; produce inaccurate results if seasonality is not taken into account (Ji and Peters 2003). Therefore, this research was undertaken to demonstrate a statistical modeling technique that would account for seasonal timing effects of precipitation on forest.
The CHIRPS rainfall data was downloaded from the Famine Early Warning Network (FEWS). CHIRPS is a quasi-global (50ºS–50ºN) gridded products available from 1981 to near present at 0.05º spatial resolution (∼5.3km) and at daily, pentadal, dekadal, and monthly temporal resolution. it is developed by the U.S. Geological Survey (USGS) and the Climate Hazards Group (CHG) at the University of California (Funk et al. 2015).
Data of short rainy season (September to November) during 2014 to 2022 were downloaded from CHG web page (http://chg.geog. ucsb.edu/data/chirps/index.html). Bayissa et al. (2019) used CHIRPS rainfall data for validating drought monitoring tool in food-insecure regions of Ethiopia. In the same way, CHIRPS rainfall data was validated and showed the best performance than other rainfall data sources in Ethiopia (Ayehu et al. 2018; Bayissa et al. 2017). In this study, CHIRPS rainfall data was used for linear regression analysis between NDVI and precipitation.
2.3 Data analysis
2.3.1 NDVI calculation;
During the early 1980s, the NDVI was defined and developed by scientists at NASA’s Goddard Space Flight Center, Greenbelt, Md. for monitoring vegetation health based on the difference between absorption and reflectance of green leaves of the red and near-infrared band of visible light, respectively (Tucker 1979). The value of NDVI of each pixel is estimated by dividing the reflectance difference by the sum between NIR and Red band. Its value range between −1 and +1, and lower values/near to zero representing stressed vegetation, and negative values representing open water, or high moisture content and +1 indicating healthy vegetation cover, respectively. Seasonal NDVI (September to November) were used in this study. In order to calculate NDVI indices from Landsat data, there are the pre-requested procedure which is guided by USGS, these include;
VNIR spectral radiance data were converted to top of atmosphere planetary reflectance using the reflectance rescaling coefficient provided in the Landsat Metadata file. The following equation is used to convert DN values to the top of atmosphere reflectance for Landsat image.
Where ρλ=TOA planetary reflectance without correction for the solar angle.
Mρ= represents a band-specific multiplicative rescaling factor from the Metadata Reflectance-Mult-BandX where X is band number.
Qcal is quantized and calibrated standard product pixel value (DN).
Aρ is represents band-specific additive rescaling factor from Metadata Reflectance-Add-Band-X, where X is band number. The next procedure is correcting reflectance value with sun angle.
The formula for correcting this value is;
Where, ρλ= TOA-planetary reflectance
ɵSE= local sun elevation angle, the scene center sun elevation angle in degrees is provided in metadata (sun elevation)
local zenith angle is calculated as eq (3)
In order to obtain more accurate results, its unit should be converted from radian to degree. This procedure is work for all bands that used in this study which operates in optical region of electromagnetic spectrum (band 4 and 5). Finally, after these corrections were completed bands were extracted into the study area by using an ArcGIS extraction tool. Then NDVI was generated by using an ArcGIS raster calculator.
2.3.2 Vegetation Condition Index
VCI is useful for vegetation condition assessment, as it assesses changes in NDVI through time since vegetation is water-stressed due to water deficiency such as during drought as stated by Kogan (1995). According to Kogan-VCI is measured as a percentage with values ranging between 0 (lowest) and 100 (highest) as in eq.5 with values equal to or below 40% considered as drought to varying degrees of severity (table 1).
Therefore, in the present study drought severity during the growing season (short rainy season) was measured using the NDVI-based Vegetation Condition Index (VCI) for the period of 8-years, for the same period using ArcGIS raster calculator tool as in Equation (5).
Where;
NDVI=the value of NDVI at the time of observation
NDVImin=Absolute minimum NDVI value of 8 years
NDVImax=Absolute maximum NDVI value of 8 years
Table 1 VCI and Drought severity class
VCI range
|
Drought severity class
|
>40
|
No drought
|
30_40
|
Mild drought
|
20_30
|
Moderate drought
|
10_20
|
Severe drought
|
0_10
|
Extreme drought
|
2.3.3 Forest cover change
In order to access drought impact on the forest, forest cover change was evaluated during the study period. Accordingly, land use/ land cover of 2021 was used to evaluate variability of forest cover change as the result of drought as the reference. This year was selected because of large portion of the region was covered by forest and to minimize contribution of settlement and agricultural activity in the result of this finding. Recently, settlement and agricultural activity are rapidly increasing. Therefore, in this study settlement and others human activities were max out from the study area so as to access drought impact on forest region through omitting impact of anthropogenic activities. Land use/ land covers were classified based on NDVI classification range as used by (Athick et al. 2019) at twelve weredas of Ethiopia, which belong to the same drought prone area. These NDVI ranges were >=-0.3 <=0.084 is water body, >0.084 < 0.18 is bare land, >= 0.18 <0.3 is sparse vegetation and >=0.3 <0.6 is dense vegetation. Based on these NDVI ranges spatial-temporal forest covers were evaluated over 8 time series years using ArcGIS image analysis.
2.3.4. Correlation analysis between precipitation and NDVI
A simple linear regression model was applied to examine how the precipitation–NDVI, relate to each other based on their mean values using “Microsoft excel”. The mean values for both variables for each year were evaluated.
Besides, the Pearson correlation matrix was applied to evaluate the relationships of an index. The linear regression model was computed as follows.
Where Yi = NDVI, for the ith period, Xi = precipitation, β0 + β1Xi = linear relationships between Yi and Xi, β0= mean of Yi when X = 0 (intercept), β1= change in mean of Y when X increases by 1 (slope), εi = random error term.
The results of simple linear regression or Pearson correlation can be positive or negative. This ranges between 0 and +1. Regression or Pearson correlation values close to zero indicate no relationship between the indices. However, if one index increases as the other index (e.g., NDVI) decreases, then the relationship is negative.