Water Balance in a Tropical Eucalyptus plantations in the Doce River Basin , Eastern Brazil

The rapid expansion of Eucalyptus plantations in Doce river basin, eastern Brazil, by changing the grassland and the natural surface cover in the savanna ecosystem, can potentially cause significant changes to water resources of the region. Especially for the higher amount of water transpired by the trees. The objective of this work was to model the water balance in an area cultivated with clonal E. grandis × urophylla in the Doce river basin, state of Minas Gerais, Brazil. Between October 2007 and September 2010, the water balance model estimated the daily variation of available soil water as a function of the water loss via evapotranspiration. Components of the evapotranspiration process were estimated by modifying the stomatal resistance of the Penman-Monteith equation. Evapotranspiration (ET) and precipitation (P), during the three years, were 3,467 mm and 3,439 mm, respectively. The evapotranspiration/precipitation ratio (ET/P) was of 1.01. Precipitation input was approximately balanced by water losses to evapotranspiration, without significant changes to water stored in soil and groundwater.


Introduction
Eucalyptus is the most widely planted hardwood genus in the world, covering more than 19 million hectares (Albaugh, Dye, & King, 2013).Approximately 5.7 million hectares of the Brazilian territory in 2016 is cover with species of the genus Eucalyptus (IBÁ, 2017).Minas Gerais state alone accounts close 24% of the total planted area in Brazil, most planted in Doce river basin.In addition to existing plantations, the increase in the area planted with Eucalyptus forests in the Doce river basin is expected for the coming years, especially in areas currently destined for pasture land.
However, any change in land use and land cover promotes changes in the energy and water cycles of a basin, especially in the evapotranspiration and flow components of the watercourses (Zhou et al., 2015, Almeida et al., 2016).As fast-growing species exhibit a high water consumption through the transpiration process, the increase in planted area is expected to cause a significant increase in total evapotranspiration in the basin (Maier et al., 2004(Maier et al., , 2017)), which can compromise the water resources.
The many countries around the world there is much concern about their water consumption of Eucalyptus plantations, since forests play an important role in the capture and distribution of rainwater in the watersheds (Huang & Zhao, 2014;Jones et al., 2017).In areas cultivated with Eucalyptus plantations, we expect to see higher evapotranspiration rates, resulting in decreased runoff and therefore the supply of water input to aquatic ecosystems (Reichert et al., 2017).Studies have shown that reforestation of degraded areas with Eucalyptus will often increase water loss by evapotranspiration, reducing water flow in streams and rivers (Reichert et al., 2017).
Modeling studies that assess components of the water balance in field cultivated with planted forests (mainly eucalyptus) are being conducted in various parts of the world, including South America (Smethurst, Almeida, & jas.ccsenet.

Loos, 201 major inpu several mo genus Euc studies con balance co
The aim o Eucalyptus

Study A
The study the munic portion of Minas Ger The topog meters, an the E. gra meters, res conservati The region from Octo temperatur (Table 1).

Water B
The mode The water process.M system, an soil evapor where, AS the precipi the plant tr depth as m  Almeida, 2001).In this study, total evapotranspiration (ET) values were obtained by the sum of the plant transpiration (T), soil evaporation E(s) and canopy interception (I).
The interception of rainfall was estimated with the rutter model (Rutter, Morton, & Robins, 1975).In this model, the evaporation rate of a saturated canopy is calculated by the Penman-Monteith equation (Monteith, 1965) assuming zero resistance of the canopy.The main input parameters of the Rutter model are shown in Table 1.
Values are based on the work of Steidle Neto et al. (2012) in commercial plantations of E. grandis × urophylla located in the same study area (Table 1).The estimation of transpiration (T) was based on the modified equation of Penman-Monteith for stomatal resistance, according to the methodology proposed by Carneiro et al. (2008): where λ is the latent heat of vaporization (MJ kg -1 ), T is the plant transpiration (mm h -1 ), s is the slope of the saturation vapor pressure curve (kPa °C-1 ) at mean air temperature, R n is the average daylight canopy net radiation (MJ m -2 h -1 ), ρ a is the density of air (kg m -3 ), c p is the air specific heat (MJ kg -1 °C-1 ), VPD is the vapor pressure deficit of the air (kPa), γ is the psychometric constant (kPa °C-1 ), r a is the canopy aerodynamic resistance (s m -1 ), and r s is the stomatal resistance (s m -1 ).
Canopy resistance was estimated by dividing the stomatal resistance by the leaf area index (LAI) of the plantation.The stomatal resistance was estimated with the model developed by Carneiro et al. (2008) for commercial eucalyptus plantations of the same species in eastern Brazil.In this model, r s is estimated as a function of the average air temperature (t med , °C), VPD, and the global solar irradiance (R g , W m -2 s -1 ) on an hourly basis: The LAI of the plantation was estimated non-destructively with a canopy analyzer LAI-2000 (Licor, Lincoln, Nebraska, USA).We measured LAI twice a year (in February and September) over the three years of the study.The value adopted for aerodynamic resistance was 83 s m -1 , based on the study of Hatton, Walker, Dawes, & Dunin (1992) in trees of e. maculate (Table 1).
Evaporation of the soil was estimated by the Penman-Monteith equation with soil resistance (r soil ) decreasing rapidly as the available soil water decreases, as observed by Choudhury and Monteith (1988).The aerodynamic resistance of the soil (R asoil ) used was 0.012 m s -1 , following Choudhury and Monteith (1988).The net radiation flux that reached the soil surface (R nsoil ) was estimated based on the Beer law formulation.
The main meteorological input variables of the water balance model were temperature (°C), relative humidity (%), solar radiation (MJ m -2 s -1 ), wind speed (m s -1 ), and rainfall (mm).Data were obtained by sensors, collected in a 20-minute interval, of an automatic weather station (Vaisala® HMP45AC, Wind Speed and Direction Sensor, CS300 pyranometer, Tipping Bucket Raingauge) installed on top of a 30 m tower located about 5 km from the study site.
The available soil water was measured in the field using the neutron attenuation method.Neutron count readings were performed using a probe access tube (Model 503 DR-HIDROPROBE, CPN International Inc., Martinez California USA) measuring two feet in length and 45 mm in internal diameter, located in the center of the watershed.Measurements were performed on 20 times along the hydrological year at 0.2-m soil intervals (layers) to a maximum depth of 2 m.The total available soil water in the rooting zone for the measured data was calculated by summing the water content of all layers.
We adjusted site-specific calibration equations between neutron probe counts and soil moisture (cm³ cm -3 ).Measurements of gravimetric moisture and neutron counting were performed at every 0.2 m depth of the soil profile to a depth of 2 m.In these campaigns, three conditions of soil moisture content were simulated: dry (20%), intermediate (50%) and wet (80%).For each depth, three replicates were performed with both the gravimetric method and the neutron probe.The calibration equations were adjusted based on the mean values for each depth of the soil profile.
The field-measured available soil water was compared against estimates provided by the water balance model.Validation was performed using the coefficient of determination (R 2 ), and root mean square error (RMSE).

Results
The seasonal behavior of the meteorological variables was similar for the three years, with the highest values coinciding with the months of the rainy season (Figure 2).Total rainfall during the study period was approximately 3439 mm.The first and second experiment years presented the lowest (785 mm) and highest (1517 mm) total precipitation values, respectively.December 2009 was the month with the highest precipitation (328 mm), accounting for about 21% of the annual precipitation.Precipitation in third was 1136 mm, which is close to the historical average (1991-2012) of 1163 mm.
The average annual air temperature (Figure 2a) during the three years of analysis was approximately 23 °C.June and July presented the lowest monthly air temperatures.The highest temperatures were concentrated from December to March, where the highest value was observed for February 2010 (27 °C).The average global solar irradiance was 18 MJ m -2 h -1 (Figure 2b).Energy availability was always higher during the wet period of the hydrological years, from October to March.The annual average vapor pressure deficit (VDP) was approximately 0.8 kPa.The highest VPD value was recorded for October 2007 (1.62 kPa).
Transpiration was the major avenue of water loss in the plantation in all months, with an average of 77, 75 and 88 mm month -1 for years 1, 2, and 3, respectively (Figure 3).In the second hydrologic year, (Oct/2008-Sep/2009), the monthly behavior of transpiration differed from the other years, with a decrease in the second month and subsequent increase in the following months.From December to April, the monthly transpiration was constant and close to 90 mm, a consequence of the high precipitation values during this period (Figure 3b).Another interesting fact observed is the strong correlation between monthly rainfall and the amount of water that is intercepted by the canopy.The nopy architecture, LAI, morphological characteristics of the trunk and the number of trees per hectare, and may contribute with up to 24% of incident rainfall (Reichert et al., 2017).
The evaporation of water from the soil to the atmosphere redistributed about 12% of the water input, also playing an important role in the evapotranspiration process.This value was similar to that found by Soares and Almeida (2001) in an experimental watershed with a nine-year-old plantation of Eucalyptus grandis, located in the state of Espírito Santo, Brazil.The low rainfall caused a negative balance (-12 mm) of available water in the soil, even in the rainy season of the hydrological year.Under these conditions, the water demand of forest species can be met by the capillary rise, reversing the direction of the flow of groundwater, and removing water from the deeper layers of the soil (Soares & Almeida, 2001).Also, as a consequence of the low precipitation values, the ratio of evapotranspiration to precipitation (ET P -1 ) was always greater than one during the year, even in the wet season (1.02).This indicates that there was no equilibrium between the input and output of water in the system, which differs from results by Almeida and Soares (2003).
The second year showed the highest precipitation (1517 mm), with events well distributed among months and periods considered.The total precipitation in the rainy season alone (Oct/2008-Mar/2009) was higher than the total rainfall observed in the other two hydrological years (Oct/2007 to Sep/2008 and Oct/2009-Sep/2010), resulting in accumulation of water stored in soil and groundwater.In the second hydrologic year analyzed (Oct/2008-Sep/2009), the average daily evapotranspiration was 3.08 mm day -1 , corresponding to 74% of the rainfall in that year.The total transpired in the year by the plantation corresponded to 59% of the rainfall.During the wet season, this number reduced to 40%.The canopy and trunk of the trees intercepted only 8% of the annual rainfall.Soil evaporation was responsible for the consumption of only 6.5% of the rainfall.The variation in water storage in the soil was always positive, regardless of the period of the year, resulting in ET P -1 ratios of less than one.
With total precipitation near the historical average of 1163 mm, the average daily evapotranspiration in the third year was 3.38 mm day -1 .Transpiration was responsible for the consumption of 93% of the rainfall.When considering the entire year (Oct/2009-Sep/2010), an equilibrium is observed between inputs and outputs of water.This suggests that in years with rainfall close to the historical average, evapotranspiration consumes nearly 100% of the incident water in this site.
When considering the entire period, of the total rainfall (3439 mm), whereas 84% left the system via transpiration, 9% via soil evaporation and 8% was intercepted by leaves and trunks (Table 2).In this same period, the ET P -1 ratio was approximately one, indicating that there is a balance between the components of the water balance in this site.Evaluating the water consumption by commercial eucalyptus plantations on the east coast of Brazil, Almeida and Soares (2003) found ET P -1 ratios close to one, even when the hydrological year presented total precipitation greater than the historical average of the region.
The model accurately estimated values of available soil water (R 2 = 0.9, RMSE = 5.8 mm or 11%), with performance similar to that observed in the work of Almeida, Ribeiro, and Leite (2013), and in Reis et al. (2014), in which growth and water balance were modeled in plantations of E. grandis located on the east coast of Brazil.The good performance of the water balance model indicates that the estimated values of evapotranspiration were not overestimated.The available soil water came close to the permanent wilting point (451 mm) only in months with a low occurrence of rainfall (June to September).

Conclusions
1).In the hydrological years with precipitation lower than average and during the dry seasons the ratio of ET/P was greater than 1.
2).A detailed analysis of the components of the water balance shows that the natural water availability in the soil was not compromised during the hydrological years analyzed, guaranteeing water storage in the soil.
3).The simple model was able to estimate the water balance components satisfactorily in the area with eucalyptus which ensures its application in future studies.
Processes such as capillary rise, deep drainage, surface runoff, and base flow were not considered in the model, as studies have shown negligible values in areas cultivated with eucalyptus plantations in Brazil(Soares &

Table 1 .
Parameters values used in the study.The parameters in italic (13-19) were taken from the literature, as described in the text

Table 2 .
A