Description of study area
This study was conducted in Waghimra Zone Easter part of Amhara National Regional state, Ethiopia. It is located 435 km far from Bahir Dar, and 720 Km from Addis Ababa. Geographically Waghimra Zone is located between 12°15´and 13°16´ N latitude and 38°20´and39°17´ E longitude (Fig.1). The most common features of the zone is its rugged topography characterized by mountains, steep escarpments and deeply incised valleys(Tadesse, 2015). It has a mean annual rainfall of 150 to 700 mm in which the highest rainfall occur during summer season which starts in mid-June and ends in early September. The rainfall pattern in the area is relatively erratic and unpredictable.
Data Source and methods of data collection
For this study an expedited MODIS (eMODIS) NDVI Terra image at 250m spatial resolution were used to monitor vegetation condition. Since this study aims to assess agricultural drought only data for crop growing season months from June to September for the 17 years period (2000 to 2016) were downloaded from (https://earthexplorer.usgs.gov website). Enhanced/expedited/expandable MODIS (eMODIS) data provides separate Geostationary Earth Orbit Tagged Image File Format (GeoTIFF) for each product in a 10 day interval, allowing the users to download only the files they need. For example, the eMODIS NDVI imagery for the month of June 2015 includes NDVI data from June 1st-10th, 6th- 15th, 11th-20th, 16th-25th, 21st-30th, and 26th-July 5th (Zhumanova et al., 2018). In this study 21st to 30th day interval of eMODIS NDVI imagery were taken for analysis purpose for the growing season of crops.
Monthly rainfall data recorded for 17 years were collected from Ethiopia national metrological service agency. Rainfall data was used to analyze relation between NDVI with variability of rainfall to drive standard precipitation index (SPI). In addition seasonal rainfall map was prepared from latitude/longitude files of those stations (Table 1 and Figure 2). To validate rainfall and satellite derived indices agricultural production yield data was collected from Waghimra Zone Agricultural office from the period 2000 to 2016.
Table 1. List of weather stations and their geographic coordinate
No
|
Station Name
|
Easting
|
Northing
|
1
|
Tsiketema
|
38.80
|
12.78
|
2
|
Amdework
|
38.71
|
12.43
|
3
|
Asketema
|
39.02
|
12.41
|
4
|
Chilla
|
38.84
|
12.41
|
5
|
Kewazba
|
38.92
|
12.48
|
6
|
Lugmura
|
39.16
|
12.40
|
7
|
Sekota
|
39.03
|
12.63
|
8
|
Yechila
|
38.99
|
13.28
|
9
|
Tekeza Hydro power
|
38.77
|
13.36
|
10
|
Wedisemro
|
39.34
|
12.76
|
11
|
Lalibela
|
39.04
|
12.04
|
12
|
Guhala
|
38.05
|
12.24
|
13
|
Chenek/Semen terra
|
38.18
|
13.27
|
14
|
Kobbo
|
39.63
|
12.33
|
Data processing and analysis
One weekly or 10 day’s composite eMODIS data sets include NDVI, quality, acquisition image, acquisition table and metadata files. In this study, NDVI and quality data has been used to calculate NDVI metrics. Quality files have been used to get the reliability of eMODIS NDVI image product which is computed in ArcGIS 10.5 spatial analysis tool (eq 1).
Where, reliable NDVI=reliable NDVI image which have values range from 0 to 10000, QC = quality image which have values from 0 to 10 where 0 is good values and 10 is fill values, NDVI is image which have values ranges from -2000 to 10,000 where -2000 is fill values and -1999 to 10,000 is valid range. After applying scale factor (the scale factor is 0.0001) NDVI values range from-0.2 to1.0 where valid/normal valid or normal NDVI ranges from 0.0 to 0.1(Zhu et al., 2013)(eq2).
Time series NDVI variation was derived from the calculation of NDVI using the eMODIS NDVI data set for the year 2000 to 2016 and also used to generate the maximum, minimum and average NDVI values of each season for the year 2000 to 2016 using ArcGIS 10.5 environment spatial analysis tool. Based on the threshold value Vegetation Condition Index and Normalized Vegetation Index anomaly was computed. To determine average value of monthly and seasonal composites of NDVI values, float (math) and cell statistics toolset of ArcGIS 10.5 were applied.
Vegetation Condition Index (VCI)
Normalize Different Vegetation Index (NDVI) has been extensively used in the past for vegetation monitoring; it is often very difficult to interpret in relation to vegetation condition, especially when comparing different ecosystems. The vegetation condition index reflects the overall effect of rainfall, soil moisture, weather and agricultural practices(Kogan, 1995).Accordingly in areas like Waghimra which have different ecosystems and non-homogenous topography VCI is important for one to compare the weather impact in areas with different ecological and economical resources, since the index captures rainfall dynamics better than the NDVI particularly in geographically non homogeneous areas.
The VCI has been used to estimate the climate impact on vegetation. This index is most useful during the growing season because it is a measure of vegetation vigor. When the vegetation is dormant (not in the summer season), the VCI cannot be used to measure moisture stress or drought. Anything that stresses the vegetation including insects, disease, and lack of nutrients will result in decreases in plant growth and therefore lower VCI values. Also, areas that have significant irrigation may not respond to precipitation deficiencies(Quiring & Papakryiakou, 2003). For each monthly and seasonal NDVI image, VCI will be processed from 2000 to 2016 using the ArcGIS raster calculator (eq 3).
Where, NDVImax and NDVImin are calculated from the long-term record for that month, and j is the index of the current month in ArcGIS cell statistics. VCI value is being measured in percentage ranging from 1 to 100. The VCI values between 50% and 100% indicates slight or optimal/normal conditions whereas VCI values close to zero percent reflects an extreme dry season (Thenkabail et al., 2004)(Thenkabail, 2004).The VCI was reclassified into five clusters (Table 2).
Table 2. Classification of VCI values in terms of drought
VCI value (%)
|
category
|
0 to 20
|
very severe drought
|
21 to 35
|
severe drought
|
36 to 50
|
moderate drought
|
51 to 60
|
slight drought
|
61 and above
|
optimum/normal
|
Standardized Precipitation Index (SPI)
Standard Precipitation Index, developed by Mckee et al. (1993) is the most widely used index for calibrating the magnitude and duration of drought events. In this study the SPI values at two time-scales, (three months SPI-3) was computed. Seasonal rainfall data have been used as an input to compute the SPI for the periods 2000 – 2016. Spatial distribution of metrological drought was prepared from latitude/longitude files of those stations (Table 1 and Fig.2).
The software which automatically calculates SPI value by using observed monthly rainfall data to detect historical drought at 1, 3, 6, 9, 12, 36 and 48 months‘ time scale. It is freely available at (https://drought.unl.edu/droughtmonitoring/SPI/SPIProgram.aspx) website.
Mathematically SPI is calculated based on following empirical formula (eq4)
Where, (Xij= is the seasonal precipitation and, Xim is its long-term seasonal mean and σ is its standard deviation). SPI results computed from seasonal rainfall data were assigned to each grid cell of the study area and reclassified based on drought severity classes (Table 3).SPI values of metrological stations have been spatially interpolated using inverse distance weight of ArcGIS spatial analysis tool box to create drought severity map of study area at multiple time scale.
Table 3.metrological drought risks classification using SPI value (McKee etal, 1993)
SPI values
|
Drought category
|
<−2.00 and less
|
Extreme drought
|
−1.50 to −1.99
|
Severe drought
|
−1.00 to −1.49
|
Moderate drought
|
0 to −0.99 Near
|
Normal or mild drought
|
Above 0
|
No drought
|
Drought frequency analysis
In this study the seasonal frequency maps derived from agricultural and metrological drought indices were reclassified into common scale based on the frequency of drought occurrence. To generate drought frequency map, each drought indices have been reclassified in to binary images for each of the drought severity class. Those maps are added to obtain the frequency of slight, moderate, severe and very severe drought occurrence at each pixel level for both agricultural and metrological drought. The resultant severity maps were then added to get agricultural and metrological drought risk maps. The probability of drought occurrence in a given area can be classified into high, moderate and low drought probability zones when drought occurs in more than 50 %, 30 to 50 %and less than 30 %of the years, respectively (Lemma ,1996). Based on these criteria, the frequency maps of each drought classes are reclassified into five classes based on the frequency of drought occurrence in study periods: 0-2 classified as no drought; 3-4 as slight drought; 5-8 as moderate drought; 9-13 as severe drought; 13-16 as very severe drought. Finally, maps from agricultural and meteorological drought frequency maps were weighted according to the percentage of influence, and then combined using weighted overly analysis (Fig 3).