Management of cracked soils for water saving during land preparation for rice cultivation

Transcription

1 Soil & Tillage Research 56 (2000) 105±116 Management of cracked soils for water saving during land preparation for rice cultivation Romeo J. Cabangon, T.P. Tuong * Soil and Water Sciences Division, The International Rice Research Institute, MC P.O. Box 3127, Makati City 1271, Philippines Abstract High water loss during land preparation of soils for rice (Oryza sativa L.) production results from bypass ow through cracks. It was hypothesized that the losses can be reduced by measures that minimize crack development during the soil drying period or impede the ow of water through these cracks. The effect of straw mulching and shallow surface on crack formation during the fallow period, and on water ow components during land preparation was investigated in eld experiments on an Epiaqualf and a Pellustert in the Philippines. Cracks did not completely close upon rewetting, resulting in high loss (152± 235 mm of water) during land preparation of the control (i.e. no soil management treatment) plots. Straw mulching helped conserve moisture in the soil pro le, and reduced the mean crack width by 32% of the control. Mulching did not signi cantly reduce mean crack depth and the amount of water used in land preparation. Shallow formed small soil aggregates which made the crack water ow discontinuous and impeded groundwater recharge from the water ow through cracks, reduced total water input for land preparation by 31±34%, equivalent to about 120 mm of water. The average surface irrigation water ow advanced faster and less time was needed for land preparation in the shallow plots compared to the control. Shallow offers a practical means for improving water-use ef ciency of irrigation systems. In rainfed areas, it may facilitate early crop establishment and, thus, reduce the risk of late-season drought. # 2000 Elsevier Science B.V. All rights reserved. Keywords: Bypass ow; Irrigation; Shallow ; Mulching; Water-use ef ciency 1. Introduction Over the next 30 years, rice production must increase by 70% from the present production to avoid rice shortage (IRRI, 1995). Rice is known to be less water ef cient than many other crops, and water for irrigation is becoming increasingly scarce because of escalating demand for non-agricultural uses. Improving water-use ef ciency of rice culture is a pre-requisite for food security in Asia. * Corresponding author. Tel.: ; fax: address: t.tuong@cgiar.org (T.P. Tuong). The rst step in lowland rice production is land preparation. Water is applied to rice elds until the topsoil is saturated and a ponding water layer of 10± 50 mm depth is maintained (land soaking) for 2 days or more on the eld. Land soaking is followed by plowing and harrowing several times under saturated condition to puddle the topsoil to a depth of 10±20 cm. After transplanting or direct seeding, the eld is kept ooded to a depth of about 50 mm throughout the growing season. To facilitate harvesting, irrigation is often stopped 2 weeks before the crop reaches maturity (De Datta, 1981). Fields often are left fallow and allowed to dry before the next crop. Drying of a puddled soil usually results in soil shrinkage and cracking. Cracks are especially promi /00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved. PII: S (00)

2 106 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105±116 nent if expanding clay minerals are present, but they may also be clearly noticeable in kaolinitic soils and can reach depths of 20±65 cm (Moorman and van Breemen, 1978; Ishiguro, 1992; Wopereis et al., 1994; Ringrose-Voase and Sanidad, 1996; Tuong et al., 1996). Irrigation for land preparation of the next rice crop, thus, involves water application to cracked soils and results in bypass ow losses (water that ows through cracks to the subsoil). Tuong et al. (1996) reported that bypass ow accounted for 41± 57% (equivalent to about 100 mm of water) of the total water applied in the eld during land soaking. Water loss throughout the period of land preparation may be much greater than this, because cracks may not close after rewetting (Moorman and van Breemen, 1978; Ishiguro, 1992; Wopereis et al., 1994; Tuong et al., 1996), and bypass ow may continue until soil is repuddled. This might explain the very high percolation losses during land preparation, accounting for up to 40% of the total water supplied for growing a rice crop (Wickham and Sen, 1978). Reducing these losses will contribute greatly to improving water-use ef ciency of rice. Tuong et al. (1996) quanti ed the ow process when ood irrigation is applied to cracked soils. Irrigation water moves rapidly in the crack networks, ahead of the surface water front. Part of this crack water in ltrates into the subsoil, bypassing the topsoil, thus, recharging the groundwater. When the subsoil is permeable, about 70% of this bypass ow may be lost to the surroundings through lateral drainage. Since the ow processes are dominated by water ow in cracks, measures that in uence crack geometry or the water ow in the cracks may affect the amount of water loss. Straw mulching, by reducing evaporation from the soil surface (Hundal and Tomar, 1985) can minimize soil shrinkage, lessen crack development during the fallow period before land soaking and, therefore, may reduce bypass ow losses. Shallow can also reduce evaporation loss. It may also form soil aggregates which block the cracks and impede water ow into them. Wopereis et al. (1994) showed that shallow reduced the crack bypass ow in undisturbed cores by 45±60%. Tillage effectiveness in reducing water loss in the eld conditions has not been tested. This study was carried out to assess straw mulching and shallow, as possible measures to reduce water loss during land preparation of dry, cracked rice soils. The processes by which these measures affects crack development during the fallow period and water balance components during the land preparation were quanti ed in eld experiments. 2. Methodology 2.1. Experimental sites The study was conducted in rice elds (i) at the International Rice Research Institute (IRRI), Los Banos, Laguna ( N, E) during the 1993 and 1994 dry seasons; (ii) in the Angat River Irrigation System, Bulacan ( N, E) during the 1993 wet season and (iii) in Munoz, Nueva Ecija ( N, E), during the 1995 dry season. All sites have two distinct seasons, i.e. wet from June to November and dry from December to May. According to soil taxonomy (Soil Survey Staff, 1992), the soil at IRRI was classi ed as an Aquandic Epiaqualf, at Bulacan Typic Epiaqualf (Tuong et al., 1996) and at Nueva Ecija Entic Pellustert (Raymundo et al., 1989). Some major soil characteristics of the sites are presented in Tables 1 and 2. At the time of land soaking, water table depths were 0.85 m (1993) and 1.1 m (1994) at the IRRI elds, 0.5 m at the Bulacan site, and 1.1 m at Nueva Ecija Treatments At IRRI The experiments were conducted in 6 m11 m plots surrounded by bunds (30 cm width and 25 cm height). The plots were hydraulically isolated by polyethylene sheets installed to 0.7 m depth along the center of the bunds. At the start of the experiment (18±20 April 1993 and 23±25 January 1994), all plots were ooded, plowed, puddled to depth of 10 cm and leveled to attain similar conditions before the treatments were applied. Surface water was drained 1 day after land leveling. All elds were allowed to dry under the sun during the fallow period until irrigation for land soaking for the next rice crop was carried out on 31 May 1993 and 17 May During the fallow period, three soil management treatments were imposed in a randomized complete block design with four replications:

3 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105± Table 1 Texture and vertical saturated hydraulic conductivity of the experimental elds at IRRI, Bulacan and Nueva Ecija a Depth (m) Clay (g kg 1 ) Silt (g kg 1 ) Sand (g kg 1 ) Saturated conductivity (m per day) IRRI b 0± ± 0.2± Farmers' fields, Bulacan b 0± (3, 9) 560 (3, 2) 150 (3, 7) ± 0.2± (3, 9) 440 (3, 2) 150 (3, 2) ± 0.3± (3, 9) 420 (3, 3) 170 (3, 5) 0.1 (3, 004) Farmers' fields, Nueva Ecija 0± (10, 36) 340 (10, 44) 130 (10, 41) ± 0.2± (10, 39) 300 (10, 48) 140 (10, 44) ± 0.4± (10, 38) 310 (10, 57) 140 (10, 40) (6, 0004) a Number of observations and standard deviations are indicated in parentheses. b Source: Tuong et al. (1996). Straw mulching. Dry rice straw (5 Mg ha 1 ) was broadcast over the plot 1 day after drainage (DAD) and remained on the soil surface during the fallow period. Shallow surface. Plots were rototilled to 5± 10 cm depth by two passes of a IRRI-manufactured rototiller at 18 DAD in 1993 and 14 DAD in The schedule of rototilling was decided by the operator based on the adequacy of soil bearing capacity to support the implement. Shallow produced topsoil aggregates of about 1±5 cm diameter.. No treatment was applied during the fallow period. In the 1994 experiment, the straw mulching treatment was not included In Bulacan The amount of water used in six farmers' elds (measured 80±90 m120±165 m) were monitored from 17 June until land preparation activities were completed. After harvest of the previous rice crop on second week of March 1993, three elds were left Table 2 Bulk density and soil water content at saturation and at different dates in farmers' elds; Bulacan and Nueva Ecija a Depth (m) Bulk density (mg m 3 ) Soil moisture content (m 3 m 3 ) Shallow Saturated Start of monitoring b Start of land soaking c Shallow Shallow Shallow Farmers' fields, Bulacan 0± (4, 0.05) 0.90 (4, 0.04) 0.63 (4, 0.03) 0.65 (4, 0.03) 0.42 (9, 0.02) 0.40 (9, 0.01) 0.52 (9, 0.02) 0.48 (9, 0.03) 0.2± (9, 0.04) 1.06 (9, 0.04) 0.61 (9, 0.01) 0.61 (9, 0.03) 0.49 (9, 0.01) 0.51 (9, 0.01) 0.57 (9, 0.02) 0.59 (9, 0.01) 0.3± (9, 0.04) 1.04 (9, 0.04) 0.61 (9, 0.02) 0.61 (9, 0.05) 0.50 (9, 0.01) 0.51 (9, 0.01) 0.56 (9, 0.02) 0.57 (9, 0.01) Farmers' fields, Nueva Ecija 0± (4, 0.01) 1.17 (4, 0.02) 0.52 (4, 0.02) 0.55 (4, 0.12) 0.26 (4, 0.05) 0.28 (4, 0.02) 0.2± (4, 0.04) 1.32 (4, 0.04) 0.50 (4, 0.01) 0.50 (4, 0.01) 0.41 (4, 0.03) 0.43 (4, 0.04) 0.4± (4, 0.05) 1.35 (4, 0.05) 0.48 (4, 0.01) 0.48 (4, 0.01) 0.46 (4, 0.02) 0.46 (4, 0.01) a Number of observations and standard deviations are indicated in parentheses. b 17 June 1993 in Bulacan. c 20±26 December 1994 in Nueva Ecija; 10±11 July (control treatment) and 28 June±10 July (shallow treatment) in Bulacan.

4 108 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105±116 fallow until land soaking was carried out on 10±11 July. Land was prepared and nal land leveling completed on 15±20 July. In the other three elds, farmers used tractor-powered rototillers to rototill the land to a depth of about 10 cm at the onset of the rainy season (12, 20, and 29 May 1993). In these elds, land soaking was carried out on 28 June, 6 and 10 July and land leveling accomplished on 4±17 July In Nueva Ecija An experiment was conducted in eight farmers' elds (measured 16±27 m220±610 m) from December 1994 to 15 January The previous rice crop was harvested on the second week of November Two treatments were imposed during the land preparation period in a randomized complete block design with four replications. In each replication, the dimensions of the elds were almost similar. The treatments were:. Fields were kept fallow until land soaking on 20±26 December 1994, plowed on 21±28 December and harrowed using comb-tooth harrow on 24±29 December. Final leveling was completed on 3±12 January Shallow surface. Fields were dry rototilled to a depth of 10 cm on 15±16 December 1994, using four-wheel tractor drawn rototillers. Land soaking was carried out from 21±26 December The elds were harrowed using comb-tooth harrows and nal leveling was completed on 4±13 January Measurements of soil water content and physical properties All measurements in the IRRI plots were taken from walk-boards installed in each plot to minimize the disturbance to the soil and crack formation. Samples, collected using 100 cm 3 cylinders, for bulk density and volumetric moisture content of the puddled and/or the tilled layer (approximately 0.1 m thick), and layers at depths 0.1±0.2, 0.2±0.3 and 0.3±0.5 m were taken at 2±3 days intervals during the fallow and land soaking periods. Similar samples were used to determine the saturated water content (Tuong et al., 1996). Vertical saturated conductivity of the 0.3±0.5 m layer at the sites were determined using constant head method and encased soil columns 0.25 high and 0.20 m in diameter (Wopereis et al., 1994). In Bulacan and Nueva Ecija soil moisture contents were monitored by the same method as at IRRI, at the start of the monitoring program before and after land soaking at four stations along the center transect of each eld. Sampling depths varied depending on the soil pro le of each site: 0±0.1 m (tilled layer), 0.1±0.2, 0.2±0.3, and at 0.3±0.5 m in Bulacan and 0±0.1 m (tilled layer), 0.1±0.2, 0.2±0.4, and 0.4±0.6 m in Nueva Ecija. For simplicity and for the presentation purpose, the above depths (and else where in this paper) refer to distances from the original soil surface. The puddled or tilled layer, however, might change its thickness due to shrinking and swelling during the fallow and land soaking periods. To ensure that the same layers were sampled at various times, all subsoil samplings used the bottom of the tilled layer (0.1 m from the original soil surface) as the datum. For example, for the layer at depth 0.2±0.3 m, the sample was taken from 0.1±0.2 m below the bottom of the tilled layer Measurement of crack dimensions At the IRRI elds, crack depth and width were monitored at 1±2 days intervals during the fallow period in a 1 m1 m subplot in each of the control and mulched plots. In the mulched plots, straw was removed before and replaced after each measurement. Crack dimensions of the shallow plots before the shallow were assumed to be the same as those in the control plot. In the farmers' elds, crack dimensions were monitored in two 1 m1 m subplots per eld 2±3 days before land soaking. Methods of measuring and computing crack depth, width, volume and the surface area of soil islands (soil masses distinctively separated by cracks) are presented in Tuong et al. (1996) Water application and monitoring Irrigation was applied from one end of each eld, through pipes (7.5±10 cm diameter) at IRRI and Nueva Ecija; and from irrigation channels via 15 cm culverts in Bulacan. Application rates were measured using 908V-notch weirs (at IRRI) and trapezoidal weirs (in farmers' elds). Discharge per unit eld width was 0.14±0.26 l s 1 m 1 at IRRI, 0.06± 0.12 l s 1 m 1 in Bulacan and 0.28±0.60 l s 1 m 1 in

5 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105± Nueva Ecija. The applied water was spread uniformly across the width of the eld by a distribution channel as described by Tuong et al. (1996). In Nueva Ecija, the distance traveled by the advancing surface irrigation water front along the center longitudinal transect of the plots during the land soaking process was monitored at approximately 1 h interval. Groundwater table tubes were installed at 10 m apart along the same transects. The water table at each location was monitored from the start of land soaking until the water table reached the soil surface, following the procedures in Tuong et al. (1996) Computation of water ow components during land preparation Water balance calculation was carried out during land soaking at the IRRI elds. In Bulacan and Nueva Ecija, water balance during land preparation was divided into two phases, namely, land soaking phase (from rst water application to rst harrowing) and harrowing phase (from rst harrowing to nal leveling). In Bulacan, water balance prior to land soaking (from start of the monitoring, 17 June 1993 to land soaking irrigation) was also carried out to take into account the rainfall at the beginning of the wet season. For each period, the water balance for the topsoil (0± 0.2 m depth), expressed in mm of water over each eld, can be quanti ed with the following equation (Tuong et al., 1996): I R ˆ S s S c A E L (1) where I is the irrigation water, R the rainfall, S s the surface water storage, S c the crack storage, i.e. the amount of water that lls the cracks in the eld, A the water absorbed in the soil layer under consideration, E the evaporation from the eld, L the losses, i.e. the amount of water that goes beyond the topsoil, recharging the groundwater. Irrigation water was calculated from the total volume of water applied (integral of ow discharge over time) divided by the eld area and S s was the change in depth of surface water. R was monitored with rain gauges installed at the experimental sites. The S c was the volume of cracks under the surface expressed in mm of water depth. The A was computed from the difference in soil moisture contents at the beginning and end of the study period. After land soaking, soil became saturated, there was no more increase in crack storage and absorbed water. The E from the start of the monitoring period to land soaking, when the topsoil water content in the elds changed from dry to saturated, were estimated by multiplying open water evaporation (measured by Class A pan) by 0.5 (Tuong et al., 1996). Evaporation after land soaking was estimated by open water evaporation. The losses were derived from the difference between the sum of inputs (I R) and the sum of soil surface storage, crack storage, soil absorption, and evaporation losses S s S c A E : The water balance was computed separately for the tilled layer (approximately 0.1 m thick) and for the 0.1 m layer immediately below the bottom of the tilled layer, before being summed up for the topsoil under consideration (0±0.2 m depth of the original soil). 3. Results and discussions 3.1. Soil moisture content Soil moisture contents at different depths at IRRI are presented in Fig. 1. Fluctuations in soil moisture content from 10 DAD (1993) and from 44 DAD (1994) were due to intermittent rains. In the 1993 experiment, soil moisture content of the mulched plots was consistently higher than the control and the shallow tilled plots for depths 0±0.1 and 0.1±0.2 m; and for depth 0.2±0.3 m until about 18 DAD. The difference among treatments at the later stage, being affected by intermittent rains, was not signi cant. Moisture content at 0.3±0.5 m depth did not differ among treatments (data not shown). The results con rmed previous ndings that straw mulch was effective in reducing soil surface evaporation (Hundal and Tomar, 1985). Water content of the 0.1 m top layer of the tilled plots did not differ signi cantly from that of the control. At deeper layers, the tilled plots had signi cantly higher soil moisture content than the control plots (Fig. 1b and c). Shallow formed a soil mulch that reduced water losses from deeper layers. At the start of land soaking, soil moisture content of the topsoil in Bulacan was higher than in Nueva Ecija due to intermittent rains before land soaking (Table 2). Soil moisture content of the tilled plots in 0.2±0.3 m

6 110 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105±116 Fig. 1. Soil moisture contents in different treatments after eld drainage at the IRRI elds, 1993 and 1994 dry seasons, (a) soil moisture at 0± 0.1 m depth; (b) at 0.1±0.2 m depth, and (c) at 0.2±0.3 m depth. Error bars are standard error (nˆ3); (c) also includes daily rain (vertical bars).

7 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105± (Bulacan) and 0.2±0.4 m (Nueva Ecija) depths were slightly higher than that of the control Crack dimensions Mean crack width and depth of the control and the mulched plots at IRRI are presented in Fig. 2. Initial increase in crack depth and width on the rst 8±10 DAD corresponded to the rapid loss of moisture from the surface layer (Fig. 1). Both crack width and depth increased more rapidly in the control plots than in the mulched plots. This corresponded to the slower rate of soil drying in the mulched plots (Fig. 1) and resulted to a lower crack volume in the mulched compared to the control. In the control treatment, a slower rate of increase in crack depth and width in the 1994 experiment compared to the 1993 experiment conformed with a slower rate of decrease in moisture content in Fig. 2. Mean crack (a) widths and (b) depths in different treatments after eld drainage at the IRRI elds, 1993 dry season. Error bars are standard error (nˆ55±161); (b) also includes daily rain (vertical bars). soil layers in Ringrose-Voase and Sanidad (1996) reported similar rates of crack development in fallow rice elds. In the 1993 experiment, the mean crack depth in the control treatment reached maximum value of about 115 mm and width about 40 mm at 19 DAD. Both mean crack width and depth did not change signi cantly afterwards. At the end of the fallow period, the crack width of the mulch treatment was signi cantly lower than that in the control treatment (Fig. 2). This corresponds to wide differences in the nal moisture content of the soil surface layer (Fig. 1) in the two treatments. The nal mean crack depth in the mulch treatment was also less, but not signi cantly than that in the control treatment (Fig. 2). The formation of cracks at lower depths was in uenced by the soil moisture in the subsoil. The non-signi cant differences in soil moisture at deeper soil layers from 20 DAD between two treatments might have resulted in only slightly different crack depth. It was likely that at the day of shallow (at 18 DAD), crack depth of the shallow plots are similar to those of the control plots. In the 1993 experiment, this depth was about 110 mm (Fig. 2), i.e. about the same as the nal crack depth of the mulched plots (106 mm). Crack width in the control plot in the 1993 and 1994 experiments were of about the same size (Fig. 2). The zero shrinkage portion of the shrinkage characteristic curve of the same puddled soil began at a soil moisture content of 0.25 m 3 m 3 (Wopereis, 1993). This implies that upon further drying, no more shrinkage will take place. Soil moisture of the topsoil layer in both years, was below 0.25 m 3 m 3, implying maximum shrinkage had taken place in this layer. Crack depth was, however, greater in the 1994 experiment, reaching a mean value of about 130 mm (Fig. 2b). This corresponded to a lower soil moisture at the deeper soil layers in the 1994 experiment. The crack dimensions prior to land soaking at IRRI, Bulacan and Nueva Ecija are shown in Table 3. The average widths and depths of cracks were similar to results of Ishiguro (1992), Wopereis et al. (1994), and Tuong et al. (1996). The wider and deeper crack at the Nueva Ecija site was probably due to the higher clay content in the Nueva Ecija elds. Intermittent rains during the fallow period at IRRI and in Bulacan probably caused some swelling and closure of cracks.

8 112 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105±116 Table 3 Dimensions of crack and soil islands (soil masses distinctively separated by cracks) prior to land soaking in IRRI, Bulacan and Nueva Ecija Location and treatment Crack Soil island Width (mm) a Depth (mm) a Volume Surface area Peripheral surface (m 3 m 2 ) b (m 2 m 2 ) b,c area (m 2 m 2 ) b IRRI 1993 Mulched IRRI Bulacan Nueva Ecija a MeanS.D. of 120±139 observations at IRRI, 148 in Bulacan, and 352 in Nueva Ecija. b Per unit eld surface area. c MeanS.D. of six sampling subplots in IRRI, six in Bulacan, and 12 in Nueva Ecija Overland and subsurface ow During land soaking of the control treatment in Nueva Ecija, water moved in the crack networks and on the soil surface. Water ow in the cracks advanced faster than overland ow of irrigation water. As a result, the water table rose close to the soil surface at a distance of about 10 m ahead of the advancing surface water front. No signi cant rise of the water table was observed ahead of the surface water front in the shallow treatment (Fig. 3). The surface water front advanced faster in the shallow than in the control plots (Fig. 4). Fig. 3. Groundwater pro les with distance from surface water front in the control and shallow plots during ood irrigation for land soaking at Nueva Ecija. Bars indicate standard deviation for 25 (in the control) and 23 (shallow ) measurements.

9 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105± surface water front (Bassett et al., 1983) in the shallow plots. Shallow can, thus, help reduce time for land soaking. In Muda Irrigation scheme, Malaysia, it is credited with the bene ts of timely crop establishment (Ho et al., 1993) Water ow components during land preparation Fig. 4. Distance travelled by the advancing surface water front in the control and shallow plots, Nueva Ecija, 1995 dry season. Observations in the control plot con rmed ndings by Tuong et al. (1996) and indicated that part of water that moved in the cracks bypassed the topsoil, in ltrated into the subsoil, and recharged the groundwater. Small soil aggregates in the shallow treatment blocked the cracks, making them discontinuous and impeded water ow in the cracks and reduced recharge to the groundwater. This conformed with Wopereis et al. (1994) who reported that small soil aggregates reduced 45±60% of the bypass loss through cracks in large undisturbed soil columns. Less loss to the subsoil meant more water was available for the surface water ow, and resulted in a faster rate of advance of the Table 4 shows the water balance components for different treatments during the 1993 and 1994 experiments at the IRRI elds. The amount of irrigation water needed for land soaking ranged from 93 to 272 mm. The increase in water needed for land soaking in the control plots in 1994 compared to 1993 was caused mainly by increased water loss. This corresponded to deeper cracks and deeper water table in In the 1993 experiment, compared to the control plots, straw mulching did not reduce signi cantly the amount of irrigation water for land soaking at IRRI elds (Table 4). By reducing the evaporation loss during the fallow period, mulching reduced the amount of absorbed water needed to saturate the surface soil layer (27 mm compared to 53±56 mm in other treatments). This reduction was a very small portion of the water input and did not result in total water savings. In both years at the IRRI elds, the shallow tilled plots used signi cantly less water for land soaking Table 4 Water balance components during land soaking as affected by surface soil management treatments at the IRRI elds in 1993 and 1994 Component a Year Treatment (mm) Mulch Shallow Irrigation water ab b 93 b 130 a b 272 a Surface storage Crack storage Water absorbed in the 0±0.2 m layer Losses a 20 b 41 a b 205 a a There was no rain and evaporation was neglected during land soaking. b In the same row, means followed by a common letter are not signi cantly different at 5% level by DMRT.

10 114 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105±116 Table 5 Water balance components during land preparation as affected by treatments in farmers' elds, Bulacan, 1993 Component Prior to land soaking Land soaking stage Harrowing stage Total Total water input a * 69 b 26 a 30 a 346 a 238 b Irrigation water Rainfall Evaporation Surface storage Crack storage Absorbed in 0±0.2 m layer Losses a 30 b 19 a 22 a 235 a 140 b Duration (day) Loss rate (mm per day) 25 a 13 b 7 a 7 a 8 6 * All water components are expressed in mm of water over the area of the eld. Total water inputs are sum of irrigation water and rainfall. In the same row and land preparation stage, treatment means followed by a common letter are not signi cantly different at 5% level by DMRT. than the control plots. Losses from the shallow tilled plots (20 mm in 1993 and 99 mm in 1994) were about 50% of those from the control plots (41 mm in 1993 and 205 mm in 1994). In 1993, the amounts of irrigation water for land soaking and water loss in shallow tilled plots were signi cantly less than those in the mulched plots, though cracks in shallow tilled were at least as deep as those in the mulched treatment plots. This highlighted the role of small soil aggregates in blocking and impeding water ow through the cracks. In farmers' elds, shallow reduced the total water input for land soaking by 54±58% of the amount needed in the control plots (69 mm compared to 150 mm in Bulacan, Table 5; and 95 mm to 227 mm in Nueva Ecija, Table 6). Most of the savings in the water input for land soaking came from the reduced losses in the shallow plots compared to the control. Findings in farmers' elds, thus, con- rmed those at IRRI. Shallow surface reduced the total water input for land preparation by 31% in Bulacan (238 mm Table 6 Water balance components during land preparation as affected by treatments in farmers' elds, Nueva Ecija, 1995 Component Land soaking Stage Harrowing stage Total Total water input (227) a * (95) b (122) a (137) a (349) a (232) b Irrigation water Rainfall Evaporation Change in storage Crack storage 25 0 ± ± 25 0 Water absorbed in 0±0.2 m layer ± ± Losses 87 a 19 b 65 a 60 a 152 a 79 b Duration (day) Loss rate (mm per day) * All water components are expressed in mm of water over the area of the eld. Total water inputs (values in brackets) are sum of irrigation water and rainfall. In the same row and land preparation stage, treatment means followed by a common letter are not signi cantly different at 5% level by DMRT.

11 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105± versus 346 mm, Table 5); and by 34% in Nueva Ecija (232 mm versus 349 mm, Table 6) of the amount needed for the control plots. Much of the differences in the amount of water-use between the two treatments occurred during the land soaking phase. High loss rate sustained in the control plots during this phase indicated that water loss through cracks continued until the rst harrowing was carried out. Harrowing reduced the loss rate of the control plots considerably and made them equivalent to those of the shallow plots. The ndings supported the observation that cracks did not completely close upon rewetting (Moorman and van Breemen, 1978; Ishiguro, 1992; Wopereis et al., 1994; Tuong et al., 1996). Harrowing broke soils in the control plots into aggregates, which sealed the cracks and produced the puddling effects which reduced soil permeability (De Datta, 1981). Where dry shallow can not be carried out, shortening the duration between land soaking and the rst harrowing may be an important measure to reduce water loss during land preparation. In the above computations, it was assumed that the same soil was taken into consideration before and after land soaking. Since the whole puddled layer (control treatment) and the tilled layer (shallow treatment) were included in the water balance, changes in their thickness during land soaking did not cause any error in the computation. Soaking decreased the bulk density of the 0.1 m stratum below the tilled layer by about 10% (e.g. from 1.15 to 1.04 at IRRI, data not shown). Assuming an isometric swelling in the stratum, the corresponding change in thickness of the stratum would be about 3%. Thus, error due to neglecting the effect of soil swelling in the computation was negligible. 4. Conclusion Straw mulching helped conserve moisture in the soil pro le, reduced crack development during the fallow period but did not reduce the bypass loss during land preparation. Shallow formed small soil aggregates, which blocked and impeded water ow in the cracks and reduced the amount of water that recharged the groundwater via the bottom of the cracks and crack faces. Water was, therefore, retained better in the topsoil. Shallow surface could reduce about 31±34% of the water input for land preparation, equivalent to a saving of 108±117 mm of water depth and shortened time required for land preparation. Water savings during land preparation may increase the service area of an irrigation system. In rainfed areas, shallow surface may also lead to earlier crop establishment and, thus, reduce the risk of late-season drought. This kind of does not necessarily require high-powered tractors. Further more, tractors/rototillers are becoming more accessible to small farmers for custom hiring, offering better opportunities for incorporating shallow surface practice in the rice production system. References Bassett, D.L., Fangmeier, D.D., Stelkoff, T., Hydraulics of surface irrigation. In: Jensen, M.E. (Ed.), Design and Operation of Farm Irrigation Systems. Am. Soc. Agric. Eng., St. Joseph, MI, pp. 447±498. De Datta, S.K., Principles and Practices of Rice Production. Wiley, New York, 618 pp. Ho, N.K., Chang, C.M., Murat, M., Ismail, M.Z., MADA's experiences in direct seeding. In: Paper Presented at the Workshop on Water and Direct Seeding for Rice. Muda Agricultural Development Authority, Ampang Jajar, Alor Setar, Malaysia, 14±16 June Hundal, S.S., Tomar, V.S., Soil-water management in rainfed rice-based cropping systems. In: Soil Physics and Rice. International Rice Research Institute, Los Banos, Laguna, Philippines, pp. 337±349. IRRI, Water: A Looming Crisis. International Rice Research Institute, Los Banos, Philippines, 90 pp. Ishiguro, M., Effects of shrinkage and swelling of soils on water management in paddy elds. In: Murty, V.V.N., Koga, K. (Eds.), Soil and Water Engineering for Paddy Field Management. Irrigation Engineering and Management Program, Asian Institute of Technology, Bangkok, Thailand, pp. 258±267. Moorman F.R., van Breemen, N., Rice, Soil, Water and Land. International Rice Research Institute, Los Banos, Laguna, Philippines. Raymundo, M.E., Mamaril, C.P., De Datta, S.K., Environment, Classi cation and Agronomic Potentials of some Wetland Soils in the Philippines. Philippine Council for Agriculture, Forestry and Natural Resources Research and Development and International Rice Research Institute, Los Banos, Laguna, Philippines, Book Series No. 85/1989, 174 pp. Ringrose-Voase, A.J., Sanidad, W.B., A method for measuring the development of surface cracks in soils: application to crack development after lowland rice. Geoderma 71, 245±261. Soil Survey Staff, Keys to soil taxonomy. Soil Manage. Support Serv. Technol. Monogr., 5th Edition, Vol. 19. Pocahontas Press, Blacksburg, VA.

12 116 R.J. Cabangon, T.P. Tuong / Soil & Tillage Research 56 (2000) 105±116 Tuong, T.P., Cabangon, R.J., Wopereis, M.C.S., Quantifying ow processes during land soaking of cracked rice soils. Soil Sci. Soc. Am. J. 60, 872±879. Wickham, T., Sen, L.N., Water Management for lowland rice: water requirements and yield response. In: Soils and Rice. International Rice Research Institute, Los Banos, Laguna, Philippines, pp. 649±669. Wopereis, M.C.S., Quantifying the impact of soil and climate variability on rainfed rice production. Ph.D. Thesis, Wageningen University, ISBN Wopereis, M.C.S., Bouma, J., Kropff, M.J., Sanidad, W.B., Reducing bypass ow through a dry, cracked and previously puddled rice soil. Soil Till. Res. 29, 1±11.