The experimental sites at MRC and JRC were under native rainforest before being deforested for cultivation in the early 1950’s, and after deforestation it has been under intensive sugarcane crop production systems until now. Before the mid 1980’s a trash-burn sugarcane production system was in operation, this changed to a green-blanket system which is being continued until now in MRC and JRC. At TRC a similar sugarcane production was in operation until 1990 and this changed to banana production until now. The N-fertiliser input for banana at TRC during the study period ranged from 310 to 525 kg N/ha.yr as urea and/or diammonium phosphate and muriate of potash at 400 to 500 kg/ha∙yr The N-fertiliser was split applied (with between 8 and 20 applications annually). The N-fertiliser applications during the dry months (July-November) were fertigated. At MRC and JRC the N-fertiliser input during the study period ranged from 100 to 120 kg N/ha.yr as urea or diammonium phosphate. It was applied once at planting in June-August or for the ratoon crop and another dressing in December before the onset of rainy-season.

We chose the descriptive parameters mean, standard error, minimum, maximum, and the coefficient of variation (CV) to statistically characterize the temporal changes in watertable levels and solute concentrations within and between rainy seasons [21] [22] . Simple linear regression analysis was performed to determine the associations between variables. The [23] software package was used for the statistical analysis of the data.

In this paper the difference between the depth to watertable from ground surface before a given rainfall event (X in m) and that immediately following it (Y in m) is considered as the inflow depth = (1)

In Equation (4), 0.15 is the freely drainable porosity [24] , and the recharge is in L/ha.

Similarly, the outflow depth in meters is the difference between the depth to watertable from ground surface after a given rainfall event (P in m) and that before the subsequent rainfall event (Q in m) = (5)

In Equation (7), 0.15 is the freely drainable porosity, and the recharge is in L/ha.

Nitrate import after a given inflow event is the difference between the concentrations after and before the inflow = M ug/L. (8)

Nitrate export is the difference between the concentration after a given rainfall and that before the next rainfall event = N ug/L. (9)

The average import in given season in kg/ha = Equation (12) ´ Equation (4) (14).

The average export in given season in kg/ha = Equation (13) ´ Equation (7) (15).

The influence of the rainfall received during the monitoring periods on watertable fluctuations in the three catchments shown in Figure 2, indicate the watertable rose in response to the rainfall received beginning in January, receded between rainfall events, fluctuated temporally throughout rainy season, rapidly receded in May/June, and thereafter gradually receded to levels that existed before it began to rise again in January. The behaviour of the temporal fluctuations were simular across the catchments. For example, in JRC it rose from 9.82 m

below the surface during the first week in January 2004 to 4.96 m in late March, then receded to 6.78 m in early April, and rose again to 4.27 m, and thereafter it began to recede continuously to 9.45 m in early November (Figure 2). In between the aforementioned periods there were temporal fluctuations, and similar trends were observed during the 2003 and 2005 rainy seasons and in the other two catchments also (Figure 2). However, magnitude of the GW rise or decline for similar rainfall events varied between the catchments, suggesting temporal fluctuations were overrided by site-specific system variable impact (Figure 2). Detailed discussions on the spatiotemporal changes in watertable fluctuations in the three catchments as impacted by rainfall distributions for the same rainy seasons as in this paper have been already been reported in [9] [10] [11] , therefore in this paper the emphasis will on inflow/outflow.

The data shown in Figure 4 for nitrate, chloride and the electrical conductivity (EC) for the three catchments show the import into and the export of the solutes from GW varied spatiotemporally within and between rainy seasons across the catchments. For example, the imported nitrate concentration temporally varied

from 0 to 1470 µg/L in the JRC compared with 13 to 561 µg/L in MRC and 8 to 2439 µg/L in TRC with mean values of 561 µg/L, 158 µg/L, and 767 µg/L, respectively, for the three catchments. The temporal variations are characterised by the large CV values of 86% for JRC, 122% for MRC, and 101% for TRC. The mean values indicate the highest import occurred in TRC followed by JRC, and MRC. Higher N-fertiliser input of 310 to 525 kg N/ha∙yr in TRC compared with 100 to 120 kg N/ha∙yr in JRC or MRC is the primary reason for the higher import in TRC than JRC or MRC.

The simple correlation between nitrate import and inflow indicate the imports were strongly correlated with inflow in JRC and MRC catchments as indicated by R² values of 0.75 and 0.84 (Table 3). Even though the correlation was significant for TRC, the import was controlled by 49% of the inflow compared with 75% to 84% in the other two catchments. The gross differences are attributed partially to the differences in system variable impact between the catchments as indicated in the inflow/outflow.

The exported nitrate concentration in JRC ranged from 200 to 1670 µg/L

All possible correlations were performed and only those that were found significant are reported in the table. The superscript “a” denotes significant at P < 0.10 and the others significant at P < 0.05. The best fits were forced through zero, assuming no import/export if there were no inflow/outflow.

compared with 24 to 401 µg/L in MRC, and from 50 to 2412 µg/L in TRC. The exported ranges show qualitatively similar trends to the import, implying varied spatiotemporally. The ranges are characterised by CV values of 73% for JRC, 72% for MRC, and 94% for TRC. However, these CVs are less than the corresponding CVs obtained for the import, suggesting less rapid temporal changes in export than import. The average export from JRC was 723 µg/L compared with 128 µg/L in MRC and 761 µg/L in TRC. The correlations between nitrate export and outflow indicate 77% of the export in JRC was accounted for by outflow compared with 44% in MRC and 49% in TRC (Table 3). The gross differences are attributed partially to the differences in system variable impact between the catchments as indicated in the inflow/outflow.

The average seasonal import and export for each one of the three catchments and the corresponding average recharge and discharge are used to compute the average seasonal import and export as described in the materials and methods section and listed in Table 2. The average import in JRC was 25.3 kg/ha compared with 10.5 kg/ha in MRC and 33.8 kg/ha in TRC. The corresponding exports were 24.5 kg/ha, 7.5 kg/ha, and 26.3 kg/ha in JRC, MRC, and TRC, respectively. It seems that ~97% off the imported nitrate was exported in the outflows of JRC compared with 72% in MRC and 78% in TRC. The highest import in TRC is attributed to the largest N-fertilizer input, as much as thrice as that in MRC and JRC. Even though the fertilizer-N inputs were similar in JRC and MRC, the lesser import in the latter cannot be attributed to seasonal rainfall differences as it was higher, by ~407 mm, in the MRC than JRC (Table 1). The higher import in JRC than MRC is attributed partially to the favourable system variable impact, the presence of fissures and cracks in the basaltic regolith that facilitated more rapid bypass flow of rain-water than through the alluvium of MRC. It seems that major proportion, 72% to 97%, of the leached out nitrate was exported during the rainy season itself. This suggests the discharge and the corresponding export should be included in the total loads transported in the streams that discharge into sensitive surface water bodies (SWB) in the humid tropics. However, it should be noted the discharged baseflow and the nitrate in it are only that exited the well. The discharge to reach and mix with stream flow depends primarily on the half-life of nitrate [32] and the transit velocity [33] . Furthermore, the import vs. export balance of ~3% to ~27% may undergo adsorption in the basaltic JRC regolith [34] and mineralization reactions.

Chloride and electrical conductivity’s (EC) temporal dynamical behaviour are believed to be good indicators of GW flow directions [35] [36] , so this aspect in this paper is given due consideration to account for solute import into and export out of the GW wells. The large min-max ranges for Cl in the three catchments indicate the concentration varied rapidly over-time. The Cl import in JRC ranged from 0 to 3100 µg/L compared with 70 to 10,700 µg/L in MRC and 250 to 9510 µg/L in TRC with CVs of 82%, 140%, and 86%, respectively, for the three catchments (Figure 4 and Table 2). The mean imports were 1146 µg/L, 1994 µg/L, and 3128 µg/L, respectively for JRC, MRC, and TRC. The relatively high mean value for TRC is a reflection of the large quantity of KCl input for banana compared with moderate input to sugarcane in the other two catchments (See materials and methods). Chloride exports ranged from 100 to 3100 µg/L, 40 to 11,800 µg/L, and 500 to 22,374 µg/L in JRC, MRC, and TRC, respectively, with CVs of 81%, 128%, and 128% for the three catchments, suggesting rapid changes in Cl export qualitatively similar to their import cousins. The mean exports were 1325 µg/L, 2515 µg/L, and 3791 µg/L, respectively, in JRC, MRC, and TRC. The mean exports from the catchments were higher than the corresponding import by 16%, 26%, and 21%, respectively in JRC, MRC, and TRC. Higher export than import may be due to dissolution of regolith material for which there is no direct evidence to my knowledge.

The EC data shown in Figure 2 indicate they varied spatiotemporally as did the nitrate and Cl. The ECs’ were divided by a conversion factor of ~2 (http://www.chemiasoft.com/chemd/TDS) to obtain total dissolved solutes (TDS). The TDS import temporally varied from 0 to 13.2 mg/L in JRC compared with 1.0 to 23.8 mg/L in MRC and 0 to 65.2 mg/L in TRC (Table 2). The temporal variations are characterised by 93%, 241%, and 163% CVs for JRC, MRC, and TRC, respectively. The exports’ temporal variations are qualitatively similar to the corresponding imports. The exports ranged from 2.0 to 14.2 mg/L in JRC, 0.0 to 10.5 mg/L in MRC, and 0 to 62.5 mg/L in TRC, with corresponding CVs of 67%, 83%, and 156%, respectively. The CVs indicate the imports were more temporally dynamic than the corresponding exports. The average import in JRC was 6.4 mg/L compared with 4.01 mg/L in MRC mg/L, and 10.9 mg/L in TRC. The highest import in TRC, approximately 41% and 63% more than in JRC and MRC, respectively, is attributed to the larger KCl input to banana in TRC than for sugarcane in JRC and MRC. The TDS export in JRC was ~18% less than the corresponding import compared with ~23% less in MRC and ~25% in TRC.

A comparison of the CV’s for EC vs. Cl indicate that EC's temporal variations were more rapid than Cl both in the inflow an outflow across the three catchments, except for that in the outflow of TRC. This indicates that EC is better descriptor of GW flow directions than Cl. Even though [35] showed similar results with regard to EC as tracer in stream bed, they didn’t compare it with Cl. According to [36] Cl can also be used as a tracer, however, this study shows that EC is better than Cl to characterise inflow/outflow directions. Because EC determination is relatively low-tech and cost-effective than Cl, the study recommends the use of EC to characterise GW flow directions. This may indirectly help us to improve and trace the transport direction of nitrate.

A comparison of the CVs of EC across the catchments indicates the temporal fluctuations in the GW-inflow were most rapid in MRC followed by TRC and JRC, respectively, in that order. This trend is similar to the inflow temporal fluctuations across the catchments, even though the CV for inflow was less than the corresponding CV for EC. However, such consistent trend was not observed between the EC in the outflow and the corresponding outflow itself across the catchments, suggesting further clarification through future research.

The imported nitrate plus Cl means as a percent of the corresponding TDS import in JRC was ~27%, compared with 47% in MRC and 36% in TRC (Table 2). The exported nitrate plus Cl means as a percent of the corresponding TDS export was 39%, 85%, and 55%, respectively, in JRC, MRC, and TRC. It seems the major constituents in the imported TDS were nitrate and Cl. The imported nitrate plus Cl means as a fraction of the corresponding exported TDS in JRC was 32% compared with 69% in MRC, and 48% in TRC, implying the major analytes in the exported TDS were nitrate and Cl. This again indicates the usefulness/importance of EC determination to characterise GW flow direction, consequently the nitrate flow direction. The exported nitrate mean as a fraction of the corresponding exported TDS was 11% in JRC compared with 5% in MRC and 9% in TRC. Even though the proportion of the exported nitrate in TDS is low, the hazard/risk associated with its export to sensitive SWB is substantial. Future research in nitrate import/export should consider computing the proportion of it in TDS that can be used to develop pedotransfer functions to estimate nitrate in GW-inflow/outflow where only EC data is available.

There are two issues that need attention with regard to nitrate import into shallow GW. First, whether this water is being used in households and the second is what are implications of the GW outflow and the export of the nitrate in it to streams and subsequent loading in SWB that receive steam flows. In this paper the first issue, household use, will not be undertaken. The second issue is number one priority in the far-northeast wet-tropics of Australia in relation to the sustainable health of the UN listed World Heritage Great Barrier Reef (GBR) [12] [13] . Nutrients, particularly nitrate and phosphate, pesticides, and sediment loadings in the GBR has been linked, partially, to the deteriorating health of the reef, which brings in billions of international and national tourist dollars to the local economy [12] .

The results from this study show that large quantities of nitrate were imported from agricultural soils into GW after rainfall events and subsequently most of it exited when the GW receded. It is also known the receded GW, most of it, was discharged into streams as baseflow. It is however not known how long it will take for the receded GW, with the nitrate in it, to reach streams. This then raises a question as to how old is the GW that is being discharged into streams or whether the recently infiltrated rainfall is being discharged. If the GW discharged is old not that percolated recently, then the nitrate in it might have undergone several biogeochemical reactions, consequently the concentration might have decreased compared to what is was when the monitoring was undertaken after rainfall events [32] . Answers to all the aforementioned questions are beyond the scope of this study. However, the recent study in MRC [33] show the transit time (TT) required to travel 280 m by employing flux-theory was ~15 years compared with ~2.7 years obtained using watertable recession approach. Shorter TT implies insufficient opportunity for the nitrate to undergo substantial biogeochemical reactions under anaerobic condition, thereby increasing the potential for its transport to streams.

[37] reported that in humid environment with conducive soil transmission characteristics, the subsurface flow contributed substantially towards stream flow. However, the large difference obtained using different methods for the estimation of TT suggest the flow mechanism and pathways may be different. The recession approach accounted for the flow through macropore bypass flow, fissures and cracks, “soil-pipe flow” and stream “gaining-loosing”, and piston-flow [33] , hence these workers concluded the TT estimates obtained using this approach is more appropriate than the piston-flow based estimates in MRC and may be in the other two catchments. [3] [4] [6] [7] [8] reported that extrapolations from point measurement (piezometer wells) of solute transport to catchment scale should take into consideration the issues of sorption in soil matrix and half-life of solutes along the transport path, including the aforementioned complexities. These workers have also indicated solutes with moderate sorption and long half-life are most susceptible to transport within an year of import. They also reported that those with high potential for rapid transport showed large annual/seasonal variability (CV). The high CVs obtained in this study (Table 2) indicate that solutes were transported relatively fast in inflow/outflow via macropore bypass flow, piston-flow, “soil-pipe” flow, and “stream gaining-loosing”, and piston-flow conditions that existed in the three catchments. It seems that extrapolation of point-measurement of solute concentration to solute mass-flux can only be undertaken provided spatiotemporal variations in GW flow direction and velocities are known. The results from this study while supporting their claim show how best to address the issue.

While acknowledging the aforementioned complexities and difficulties involved in baseflow discharges into streams, we propose to relate the concentrations of the nitrate in GW with the trigger values proposed for different aquatic ecosystems sustainable health hazard/risk in this region. The trigger values provided for the oxides of N, which is predominantly nitrate-N, provided in Table 3.3.4 of the [38] National Water Quality Management strategy report are used in this section to relate the hazard/risk associated with the nitrate in GW when it reaches sensitive SWB. The trigger values for SWB ranged from 1 to 4 μg∙L⁻¹ for offshore marine reef, 2 to 8 μg∙L⁻¹ for inshore marine reef, 10 μg∙L⁻¹ for lowland rivers, freshwater lakes, reservoirs and wetlands, and 30 μg∙L⁻¹ for upland rivers and estuaries.

The lower quartiles (LQ) for nitrate-N in this study the ranged from 300 to 1500 μg∙L⁻¹ with a LQ of 1500 μg∙L⁻¹ in JRC, 23 to 476 μg∙L⁻¹ with a LQ of 476 μg∙L⁻¹ in MRC, and 1761 to 3285 μg∙L⁻¹ with a LQ of 3285 μg∙L⁻¹ in TRC. The LQ indicate that 75% of the nitrate concentration in GW is 2 to 3 orders of magnitude higher than the triggers proposed for most of the SWB (Table 2). The minimum in the LQ range is higher than the trigger values, suggesting the nitrate in GW is a hazard/risk to most of the surface aquatic system sustainable health. The hazard to turn into risk depends on the TT, the half-life of nitrate in GW, and dilution effect when baseflow reaches streams. The half-life of nitrate in GW could range from 1 to 3 years [33] this coupled with the TT of 3 years [14] suggest the potential for 75% of the nitrate in GW to be transported to nearby streams. However, the dilution effect is to be accounted for when the baseflow discharge mixes with the stream flow. Thus we suggest the nitrate in the GW in the three catchments is a hazard/risk to the SWB receiving baseflow discharges and it should be given due consideration in order to provide more reliable N- loading estimates to sensitive SWB.