Meteorological scenario of Ethiopian floods in 2006–2007

Abstract

The meteorological scenario of Ethiopian highlands floods is studied. Daily rainfall in the period 1997–2007 reveals two peaks: 23–28 July 2006 and 26–31 July 2007. National Center for Environmental Prediction (NCEP) composites suggest that anomalous southerly monsoon flow over the West Indian Ocean is re-directed by an anomalous Arabian ridge westward across the Red Sea and Ethiopia. A tongue of moisture stretches from the Congo towards the highlands, but westerly equatorial wind anomalies are absent. Anomalous sinking motions and dry conditions are evident over the West Indian Ocean. Diurnal analysis reveals northwesterly flow over eastern Sudan during afternoon hours, whilst back-trajectory analysis highlights a Red Sea source and lifting over the eastern escarpment of Ethiopia. The upper level tropical easterly jet connects Indian and Ethiopian rainfall at intra-seasonal (~40 days) time scale; whilst low-level meridional flow convergence is evident during flood events. Hovmoller analysis on 10°N reveals cyclonic signals propagating westward from the Arabian Sea at 500 km day⁻¹ that produces a 10-day cycle in Ethiopian rainfall. The floods in 2006–2007 occurred at the peak of the annual cycle, with diurnal controls inducing ¾ of rain in the late evening. Whilst cold surges from southern Africa played a role in the 2006 flood, bursts in the northern Hadley cell are a more general determinant. The convection associated with the 2007 flood went on to become a destructive Atlantic hurricane.

1 Introduction

Although Africa is often plagued by drought, heavy rainfall in 2006 and 2007 over the Ethiopian highlands (6–14°N, 35–40°E; Fig. 1a, b) caused flooding of the Blue Nile River. The Ethiopian Meteorological Service reported July totals >400 mm across western districts, with daily totals >100 mm and seasonal totals up to 2,000 mm (Table 1). The United Nations reported 620 fatalities, ~35,000 displaced and ~118,000 people ‘affected’ by flooding in 2006. During 2007, floods again spread across the region with ~20,000 displaced and ~103,000 affected. All major rivers overflowed and drainage into Sudan resulted in 122 fatalities and ~200,000 homeless. The floods destroyed crops and livestock across Northeast Africa, where most people depend on farming for their livelihoods. The World Health Organization reported in September 2007 that ~3.5 million people along the Sudan–Ethiopian border were at risk of flood-related epidemics, such as dysentery and malaria.

Fig. 1

(a) Topographic map of northeastern Africa with national borders; box outlines area used in statistical analyses. (b) Close-up view of Ethiopian highlands with shaded contours at 500 m and rivers (blue)

Table 1 Ethiopian highland station rainfall extremes for 2007 in descending rank

Wegiel and Herbster (1998) and Hill and Lin (2003) have demonstrated how convective outbreaks over the Ethiopian highlands form into African easterly waves. Through satellite imagery, synoptic observations and model simulations, evening thunderstorm clusters merge into a vortex downstream over Sudan and propagate westward. Hill and Lin (2003) characterized the meteorological conditions over the Ethiopian highlands in late July 2000 as having favorable horizontal and vertical easterly wind shear (relative vorticity >10⁻⁴ s⁻¹) and equivalent potential temperature >345 K in the 850–300 hPa layer induced by moist inflows from southern Sudan and the Red Sea. Matthews (2004) identified zonal propagating intertropical waves as a source of cyclonic shear on the equatorward flank of the African easterly jet, leading to enhanced convection at synoptic and intra-seasonal time scales. Studies relating the regional meteorology to global climate have shown that Ethiopian rainfall and Blue Nile flow is enhanced by La Niña conditions and a more intense Indian monsoon (Camberlin 1997) which modulates the tropical circulation (Eltahir 1996; Potter et al. 2004; Korecha and Barnston 2007; Segele et al. 2009). The Atlantic zonal overturning circulation that often develops during Pacific La Niña has upper easterly/lower westerly flow that brings widespread uplift (descent) to Africa (South America), as outlined by Yeshanew and Jury (2007) and Jury (2009). Conway et al. (2004) and Seleshi and Camberlin (2006) studied daily rainfall records for Ethiopia and found little trend in extreme events, despite a weak down-trend in the Nile River flow in the 20th century (Conway et al. 2009). Heavy rains are an annual occurrence that generally benefit agriculture and water resources. Blue Nile flow from central Ethiopia (Fig. 1b) is typically ~50 × 10⁹ m³/yr (Vowinckel and Orvig 1979; Ahmed and Ismail 2008). About 30% of Blue Nile flow in Sudan derives from July rainfall over Ethiopia, with ~1-month delay. Run-off efficiency averages ~30%, whilst the coefficient of variability is ~20% (Dettinger and Diaz 2000). In 2006–2007, exceptional run-off contributed to numerous landslides and loss of infrastructure.

The objective of this paper is to understand the regional atmospheric circulation and thermodynamic conditions that modulated the 2006–2007 Ethiopian flood events through analysis of meteorological patterns on flood days and statistical analyses of filtered daily time series. It is hypothesized that a number of factors contribute to African flooding (Asnani 1993): (1) equatorial moisture is drawn from the Congo, (2) low-level winds undergo orographic and diurnal forcing, (3) the upper level tropical easterly jet (TEJ) surges in response to thermal gradients and (4) cold surges from southern Africa accelerate the southern Hadley circulation. The statistical analysis seeks to establish the relative influence of different meteorological forcings, whilst composite, case study and Hovmoller analyses illustrate the regional atmospheric controls. This work will not delve into the wider issues of inter-annual variability and hemispheric scale controls, but will instead focus on synoptic scale events over the Ethiopian highlands. In Section 2, the data and methods are outlined. Section 3 provides results in statistical context, as composites, and case studies. Conclusions are given in Section 4.

2 Data and methods

To study flood events over the Ethiopian highlands, daily rainfall from the Global Precipitation Climatology Project (GPCP) was obtained in the period 1997–2007 as area averages: 6–14°N, 35–40°E via the International Research Institute (IRI) climate library (iridl.ldeo.columbia.edu). The GPCP rainfall product is based on satellite infrared radiance threshold as described by Adler et al. (2003), interpolated to a 1° grid. Rainfall data were similarly extracted over the Indian monsoon (10–25°N, 75–100°E) to explore the potential for intra-seasonal connections (Editorial MWR 1900; Hoyos and Webster 2007). Other rainfall data used here include the 3-hourly cMorph multi-satellite product (Joyce et al. 2004) and the daily Famine Early Warning System (FEWS) gauge-interpolated data (Love et al. 2004). The three rainfall products are consistent (r > 0.80) over the 2-year study period. Flood events were identified by ranking 3-day area-averaged GPCP rainfall in the period 1997–2007 that contains diurnal and annual cycles.

To investigate the meteorological scenario, daily and 6-hourly re-analyzed 2° resolution atmospheric fields (Kanamitsu et al. 2002) from National Center for Environmental Prediction (NCEP) were obtained for the Ethiopian highland region: 700 hPa meridional wind, 150 hPa zonal wind, 500 hPa relative vorticity, 150 hPa divergence, 700 hPa divergence and precipitable water. Statistical associations between the area-averaged meteorological variables and Ethiopian 3-day rainfall in the period 1997–2007 were analyzed. Cross-correlation values >|0.42| were deemed significant at 95% confidence, taking into account a reduced degrees of freedom from annual and synoptic cycles. The amplitude and period of oscillations in 2006–2007 were studied using wavelet spectral analysis, and statistical significance with respect to red noise was determined.

After objectively ranking the Ethiopian highlands GPCP rainfall, two periods—23–28 July 2006 and 26–31 July 2007—had greatest rainfall. NCEP re-analysis fields for these 12 flood days were averaged, and the mean was subtracted to produce composite anomaly maps and cross-sections over Northeast Africa and the adjacent Indian Ocean 15°S–35°N, 15°E–60°E, following methods similar to those described by Fontaine et al. (1995) and Rudari et al. (2004). To study the diurnal cycle, 0800 hours LST fields were subtracted from 14:00 hours LST for surface air temperature and 700 hPa vector winds. Thus, the annual cycle was removed in the composites. Individual days within the two flood periods were further analyzed using 1° resolution Global Forecast System (GFS) numerical model analyses. Output fields for winds at 700 and 150 hPa, convergence, precipitable water, boundary layer height, convective available potential energy, surface air temperature and soil moisture were obtained, to understand localized and orographic forcing during peak floods. In addition, Hovmoller plots were drawn to understand temporal forcing and zonal wave propagation. cMorph rainfall estimates at 3-h intervals were analyzed for the two flood periods to quantify diurnal amplitude and air parcel back-trajectories were studied using the Hysplit model (http://www.arl.noaa.gov/documents/reports/arl-224.pdf) with an end-point at 10°N, 37°E, 2,000 m above ground from www.ready.noaa.gov. The methodology aims to provide a regional description of multi-variate factors contributing to Ethiopian flood events, and not to single out a certain influence, nor to imply cause–effect relationships.

3 Results

3.1 Statistical analysis of daily rainfall

Cross-correlations of 3-day filtered continuous Ethiopian highlands rainfall and meteorological variables in the period 1997–2007 are listed in Table 2. Values are significant (after deflation for annual cycle) for precipitable water, upper zonal wind, lower meridional wind and Indian monsoon rainfall. Indian rainfall generates the TEJ (Webster and Hoyos 2004) and modulates Ethiopian precipitable water, vorticity and meridional wind. Hence, the pulse frequency and amplitude of the Indian monsoon is connected with Northeast African highlands weather (Camberlin 1997). Conducting a stepwise multi-variate regression of all variables onto Ethiopian rainfall, precipitable water and 700 hPa meridional wind provide an optimal fit, r ² = 51%. The influence of meridional wind is negative, so pulses of northerly wind correspond with increased convection. These could be generated by interaction of the northern Hadley circulation and zonal propagating waves, as will be shown later.

Table 2 Cross-correlation of 3-day smoothed Ethiopian and Indian rainfall with various meteorological variables averaged over 6–14°N, 35–40°E in the period 1997–2007

The 2-year period 2006–2007 is analyzed for spectral energy for the leading variables in Fig. 2. Ethiopian highlands GPCP rainfall displays a seasonal increase with two sharp peaks at the end of July 2006 and July 2007 (Fig. 2a). Oscillations were significant about 10 days during the rainy season, and there was a low frequency pulse denoting the mid-summer maximum. One possible explanation for the 10-day cycle is intra-seasonal oscillations moving eastward across central Africa (Mutai and Ward 2000). Co-spectral energy between the Ethiopian and Indian rainfall (Fig. 2b) was prominent around 40 days, suggesting the monsoon is modulated by Madden Julian Oscillations, which then ‘pulse’ the upper-level TEJ (Hulme and Tosdevin 1989; Rui and Wang 1990; Camberlin 1997; Hoyos and Webster 2007; Ding and Wang 2009). Both precipitable water and low-level meridional wind varied at intervals of about 10 days throughout the 2-year period (Fig. 2c, d).

Fig. 2

Wavelet spectral analysis for 2006–2007 of a Ethiopian 3-day filtered rainfall, b co-spectra of Ethiopian and Indian monsoon rainfall, c Ethiopian precipitable water and d Ethiopian 700 hPa meridional wind. Energy is shaded at 25% power intervals; cone of validity given; outline denotes statistical significance w.r.t. red noise at 90% confidence. Dashed lines highlight key cycles at 10 and 40 days; circles in a identify flood events: 23–28 July 2006 and 26–31 July 2007

cMorph rainfall for the two flood cases—23–28 July 2006 and 26–31 July 2007—is mapped in Fig. 3a, b. Highest rain rates occurred over western (2006) and northwestern (2007) Ethiopia, filling the Blue Nile River basin. Heavy rainfall extended into southeastern Sudan and the southwestern Arabian peninsula. The time series (Fig. 3c) indicate a pronounced diurnal cycle, with heaviest rainfall after 18:00 hours LST each day. The convection associated with the 2007 flood went on to become Hurricane Dean, a category 5 system that devastated parts of the Caribbean (www.nhc.noaa.gov/pdf/TCR-AL042007_Dean.pdf). The 2006 flood triggered an easterly wave that never intensified due to Saharan dust, amongst other factors. To determine air mass sources, Hysplit-model back-trajectories were analyzed from GFS analyses (Fig. 3d) using an ‘end point’ 2,000 m above ground (~700 hPa) at 10°N, 37°E. Two 5-day periods ending at 24:00 hours on 27 July 2006 and 30 July 2007 are considered. The majority of trajectories arrive from the Red Sea coast, following an anti-cyclonic arc over Eritrea. The z–x section of back-trajectories indicates that the channel leading into the Rift Valley (cf. Fig. 1b) funnels and lifts the flow in agreement with Kinuthia and Asnani (1982). Both 2006 and 2007 floods exhibit back-trajectories from the north; however, back-trajectory analysis for other Ethiopian floods (1998–2000) has inflow from the southwest (Congo).

Fig. 3

Multi-satellite rainfall maps for a 23–28 July 2006, b 26–31 July 2007 and c their respective 3-h rainfall time series averaged over 6–14°N, 35–40°E. d Hysplit back-trajectory plan view and z–x section analysis from 10°N, 37°E, 2,000 m above ground for two 5-day periods up to 27 July 2006 and 30 July 2007

3.2 Flood composite anomalies

Composite maps for flood days are given in Fig. 4. The 700-hPa wind anomalies exhibit a cyclonic vortex centered on Ethiopia (Fig. 4a). The southerly monsoon off the Somali coast is 5 m s⁻¹ stronger than usual, and draws easterly flow from the equatorial Indian Ocean. This anomalous flow extends toward the Arabian Peninsula and turns westward across the Red Sea then southward toward the Blue Nile catchment. Thus, part of the ‘Somali jet’ is briefly deflected back towards Africa. This appears to be a key process underlying the peak-season Ethiopian floods of 2006–2007. The 700 hPa specific humidity field (Fig. 4b) reveals anomalous dry conditions over the West Indian Ocean and enhanced moisture over the southern Arabian Peninsula/Gulf of Aden—a ‘source-sink’ pattern. Humid air also prevails from the Congo toward Kenya and Sudan. The composite anomaly ‘snap-shots’ indicate a cyclonic vortex over Ethiopia during the floods, whilst the back-trajectories highlight northerly inflow as an important element in the precursor period.

Fig. 4

Composite anomalies of National Center for Environmental Prediction (NCEP) reanalysis fields for 12 peak flood days in 2006–2007: a 700 hPa winds with bold arrows >3 m s⁻¹, b 700 hPa specific humidity

Vertical section composite anomalies are shown in Fig. 5, based on NCEP re-analyses averaged on 32–37°E and 8–13°N. Temperatures exhibit a deep warm anomaly to the north (Fig. 5a) that strengthens the upper easterly jets. This warming reaches +6°C through the 800–300 hPa layer near 25°N (Egypt). Because of its structure, the warming is not only related to surface heating but indicative of an upper ridge over the eastern Mediterranean. Specific humidity increased over the equatorial zone and over the Ethiopian highlands from the surface up to 500 hPa. Moisture was depleted over Egypt as illustrated in the N–S section (Fig. 5b). The meridional wind and vertical motion pattern shows a tilted zone from 900 to 500 hPa of anomalous northerly rising flow (Fig. 5c) in the latitudes 10–17°N, consistent with the statistical results (r = −0.45 for 700 V, Table 2). Anomalous sinking motion over Egypt feeds into the northerly winds over Sudan. Zonal winds and vertical motions are analyzed as an east–west section in Fig. 5d. Anomalous rising motion prevails over the highlands and Sahel, as expected. Anomalous sinking motions are evident over the west Indian Ocean (55°E) particularly below 600 hPa. The TEJ accelerates downstream over Sudan. These circulation patterns are consistent with Hill and Lin (2003; their Fig. 5a) for easterly wave initiation over Ethiopia in July 2000.

Fig. 5

Composite NCEP reanalysis fields for 12 peak flood days in 2006–2007 as vertical sections averaged over Ethiopian highlands: a temperature anomaly; b specific humidity anomaly; c V,W wind on north–south slice; d U,W wind on east-west slice. Vector key is given; W component × 10

Diurnal amplitude is studied on flood days by subtracting 0800 hours from 14:00 hours LST fields for 2 m air temperature and 700 hPa winds; the resultant composite differences are mapped. Air temperatures increase 12°C over the northeastern Sahara desert (18–30°N) and the highlands of southern Africa (Fig. 6a). In the equatorial band 0–10°N, there is little change from morning to afternoon, according to NCEP re-analysis, due to clouds inhibiting surface heating and to evaporative cooling by rainfall. The wind response to diurnal heating is evident over southern Egypt and northern Sudan, with northwesterly differences >3 m s⁻¹ (Fig. 6b) that enhance orographic lifting during afternoon hours along Ethiopia’s NW border, consistent with back-trajectories (Fig. 3d) and the helical circulation (Fig. 5c).

Fig. 6

Composite differences of NCEP reanalysis fields for 12 peak flood days in 2006–2007 describing diurnal amplitude (1,400 hours minus 0800 hours LST) for a 2 m air temperature and b 700 hPa winds with bold arrows >3 m s⁻¹

The mean diurnal cycle was analyzed from 6-hourly NCEP re-analyses and cMorph rainfall, and results are presented for the 2007 flood event in Fig. 7. The diurnal cycle follows a sequence of (1) surface heating (insolation and heat fluxes >800 W m⁻²) by 14:00 hours LST (Fig. 7a), (2) deepening of the warm boundary layer >2,000 m above ground level (AGL) after 14:00 hours, causing an influx of northern flow (Fig. 7b). Cyclonic vorticity and vertical uplift are at a minimum during mid-day. (3) Towards sunset convective available potential energy rises and upward motions develop, (4) peak rainfall occurs from 2000 to 0200 hours (Fig. 7c) and corresponds with greater cyclonic vorticity. The late evening rainfall maximum is typical of continental highland locations in summer (Hill and Lin 2003; Singh and Nakamura 2009).

Fig. 7

Mean diurnal cycle in the period 24–31 July 2007 from 6-h NCEP reanalysis except for rainfall from cMorph product. a Surface fluxes: Qs insolation, Qe evaporation, Qh heat; b 500 hPa absolute vorticity (left) and 700 hPa meridional wind, and c rainfall (left) and 500 hPa vertical motion with Pa/s scale inverted

3.3 Analysis of July 2006 and July 2007 floods

The Blue Nile River in southeastern Sudan rose ~1 m after the floods, according to gauges at Roseries, Khartoum and Atbara. Meteorological features in the July 2006 flood are studied in Fig. 8a, b from GFS analyses. The 700-hPa wind streamline analysis of 18:00 hours LST 25 July reveals a classical neutral-point convergence, with southerly (northerly) flow over southern (northern) Ethiopia splitting eastward to the Indian Ocean and westward to the African Sahel. Meridional convergence in the lower levels was a significant factor in the floods. Above 500 hPa, flow was from ENE 20 m s⁻¹. The meteogram for the 3-day period (Fig. 8b) indicates a large diurnal cycle for planetary boundary layer (PBL) height and convective available potential energy (CAPE), with CAPE lagging >3-h with values up to 1,000 J kg⁻¹. Rainfall followed CAPE, peaking on the 25th. A total water volume >10¹⁰ m³ was generated for run-off by the end of July 2006. Satellite infrared imagery indicated that thunderstorm clusters reached −70°C over the western highlands following earlier development along the eastern escarpment (disc.sci.gsfc.nasa.gov/hurricane/) and moved westward at ~10 m s⁻¹. CloudSat reflectivity slices were collected at 23:00 hours LST on 24 July and 23:00 hours LST on 1 August 2006 (Fig. 8c, d) to describe the vertical structure of convection. In the first case, an isolated thunderstorm cluster developed near the eastern escarpment; in the second, a larger cluster was located over the NW border. There is high reflectivity above the freezing level indicative of cold cloud processes. Cluster diameters >100 km were typical. According to the GFS analysis of 18:00 hours LST 29 July 2007, surface air temperatures exceeded 35°C across the eastern lowlands of Ethiopia and Red Sea where the PBL lifted above 2,000 m. Over the highlands, surface temperatures remained near ~15°C and the PBL lifted ~500 m. The eastern escarpment on 40°E divided the two regimes. Upper easterly winds of 40 m s⁻¹ from 6° to 12°N coincided with upper divergence (4 10⁻⁴ s⁻¹). Soil moisture exceeded 40 mm over the highlands (not shown), helping to ‘anchor’ convective systems (Douville et al. 2001) and 700-hPa winds described a cyclonic vortex.

Fig. 8

a GFS 700 hPa wind streamlines for 18:00 hours LST 25 July 2006. b GFS meteogram (at circle in a of CAPE, PBL height and 6-h rainfall in July 2006. CloudSat reflectivity slices on lines in e for 2,300 hours LST 24 July 2006 (c) and 1 August 2006 (d) with NCEP isotherms overlain

Hovmoller time-longitude plots were analyzed on 8–13°N from daily GFS analyses, and the 2007 case is given in Fig. 9. The 700 hPa meridional wind anomaly pattern shows a steady increase of northerly flow over Ethiopia during the flood and a simultaneous acceleration of southerly winds to ~10 m s⁻¹ over the West Indian Ocean (Fig. 9a). Westward propagation of 500 hPa vorticity anomalies is evident in Fig. 9b. A cyclonic signal crosses the west Indian Ocean around 25 July 2007, moves over Ethiopia and Sudan during the flood event, and continues across the African Sahel at a speed of ~500 km day⁻¹. The passage of the system over Ethiopia caused low-level northerly winds to surge (V~−7 m s⁻¹), whilst upper easterly (U) winds oscillated from −15 to −30 m s⁻¹. In the 2006 case, a cyclonic signal propagating from the Arabian Sea is not so clear.

Fig. 9

Hovmoller plots averaged over 8–13°N from 15 to 60°E for the period 10 July–20 August 2007. a 700 hPa meridional wind (shading blue −10 to red +10 m s⁻¹), b 500 hPa relative vorticity (blue −5 to red +5 10⁻⁵ s⁻¹). Circle denotes flood event, dashed lines are cyclonic vortex propagation

To understand whether cross-equatorial cold surges are important (Love 1985) to accelerating the southern Hadley circulation in Ethiopian floods, Hovmoller analyses were constructed for 700-hPa temperature, winds, vertical motion and rainfall averaged over the 30–35°E longitude from 40°S to 30°N and the case of 2006 is provided in Fig. 10. During the peak rains, a cold surge is seen to progress rapidly equatorwards as an area of increased southerly flow with T < 15°C between Zimbabwe and Kenya. Rainfall over Ethiopia follows rainfall near Durban, South Africa, as a cold front passes, and is followed by a ridging anticyclone that delivers the cold air and accelerates the lower limb of the southern Hadley circulation. In the 2007 case, such a feature is not as obvious; instead, heavy rainfall extended southward into Kenya.

Fig. 10

Hovmoller plot averaged over 30–35°E from 40°S to 30°N for the period 1 June–30 September 2006 of a 850–700 hPa temperature (shaded blue to orange with contours at 5°C) and multi-satellite rainfall (rainbow shading aqua to red, 5 to 30 mm/day); and b 850–700 hPa meridional wind and 700–500 hPa vertical motion (largest vector = 5 m s⁻¹ with W × 10). Topographic profile is plotted on lower axis. Bold tick highlights flood event.

4 Conclusions

The analysis of daily Ethiopian highlands rainfall in the period 1997–2007 revealed two periods of flooding—23–28 July 2006 and 26–31 July 2007—that mark the seasonal peak of convection. Using NCEP composites, it was seen that anomalous southerly monsoon flow over the West Indian Ocean was re-directed westward across the Red Sea to Ethiopia. Inflow from the Congo did not play a role in these events. Instead, moisture was drawn from the West Indian Ocean, leaving anomalous sinking motions and dry conditions there. The diurnal analysis revealed additional northwesterly flow over Sudan during afternoon hours. Orographic forcing was evident in the back-trajectory analysis: re-directed monsoon flow passed over the Red Sea, funneled into the Rift Valley and lifted over the highlands of Ethiopia. The TEJ acts to connect Indian and Ethiopian rainfall at intra-seasonal time scales (r = −0.74, Table 2). Hovmoller analysis revealed cyclonic signals propagating westward from the Arabian Sea at ~500 km day⁻¹ (5.8 m s⁻¹) that pulse Ethiopian rainfall, precipitable water and meridional wind; and spawn African easterly waves (Hill and Lin 2003; Lin et al. 2005). The convection associated with the 2007 flood went on to become a destructive Atlantic hurricane. A key feature underlying the floods of 2006–2007 is a diurnally-modulated overturning circulation north of Ethiopia (cf. Fig. 5c) that represents a mid-summer burst of the northern Hadley cell in conjunction with a surge of the TEJ (Fig. 5d). Further work on Ethiopian flood events could include the use of in situ data and mesoscale model simulations, to distinguish annual, intra-seasonal, synoptic and diurnal factors.