The In uence of Moss Colonization and Biochar Application on Evaporation Losses and Surface Crack in Shallow Carbonate–Derived Laterite During Dry–Wet Cycles


 AimsSoil water deficit in karst mountain lands is becoming an issue of concern owing to porous, fissured, and soluble nature of underlying karst bedrock. It is important to identify feasible methods to facilitate soil water preservation in karst mountainous lands. This study aims to seek the possibility of combined utilization of moss colonization and biochar application to reduce evaporation losses in carbonate-derived laterite.MethodsThe treatments of the experiments at micro-lysimeter included four moss spore amounts (0, 30, 60, and 90 g·m−2) and four biochar application levels (0, 100, 400, and 700 g·m−3). The dynamics of moss coverage, characteristics of soil surface cracks and surface temperature field were identified. An empirical evaporation model considering the interactive effects of moss colonization and biochar application was proposed and assessed.ResultsMoss colonization reduced significantly the ratio of soil desiccation cracks. Relative cumulative evaporation decreased linearly with increasing moss coverage under four biochar application levels. Biochar application reduced critical moss coverage associated with inhibition of evaporation by 33.26%-44.34%. The empirical evaporation model enabled the calculation of soil evaporation losses under moss colonization and biochar application, with the R2 values ranging from 0.94 to 0.99.Conclusions Our result showed that the artificially cultivated moss, which was induced by moss spores and biochar, decreased soil evaporation by reducing soil surface cracks, increasing soil moisture and soil surface temperature.Moss colonization and biochar application has the potential to facilitate soil moisture conservation in karst mountain lands.


Introduction 37
The karst landform, which plays a significant role in water supply and 38 consumption, covers approximately 10% of the Earth's land surface (Yang et al. 39 2017). It covers extensive parts of China (~3,440,000 km 2 ), especially in 40 southwestern China. These lands are usually characterized by a surface-underground 41 structure. This sieve-like structure enhance soil water loss, soil drought, and rocky 42 desertification. Because of the unrestrained water loss and the low water storage 43 capacity, flora in karst lands grow slowly and exposed soils are easily eroded, which 44 accelerate soil drought and pose threats to the stability of local ecosystems (Yang et  were regarded as the control. Three replicates were prepared for each treatment and, 170 thus, a total of 48 microlysimeters (10 cm diameter,4 cm soil thickness, and free 171 drainage at the bottom) were operated (Fig. 1c). The microlysimeter scale referenced 172 to previous studies on shallow soil amelioration (Table 1). It should be noted that 173 using a microlysimeters in this study not give a complete soil evaporation processes 174 occurring under field conditions. However, this microlysimeter scale of study enables 175 systematic studies to be carried out under controlled conditions that enable insights 176 into the changes in soil evaporation process under dynamic of soil surface 177 characteristic. The soil materials were firstly divided into two layers to pack into these 178 microlysimeters and biochar materials were added into the middle of the 179 microlysimeter as a layer (Meng et al. 2014). The dry bulk density of the soil-biochar 180 in the microlysimeter was controlled at 1.3 g·cm −3 . The moss spores were mixed with 181 the soil materials and evenly seeded onto the microlysimeter surface. After that, these 182 11 microlysimeters were saturated from the bottom by capillary action (self-absorption 183 method) until the top surface of soil column got wet and soaked for 72h to ensure 184 saturation. Finally, the bottom of microlysimeter was sealed with polyvinyl chloride 185 film to prevent drainage. 186 Evaporation experiments took place over six dry-wet cycles (D-W1, DW-2, D-187 W3, D-W4, D-W5 and D-W6) (Fig.1 c). During each dry-wet cycle, evaporation 188 lasted for 12 days and saturation lasted for 3 days. The air temperature and 189 atmospheric relative humidity in the greenhouse were automatically measured using a 190 temperature and humidity recorder (Testo 174H, Testo SE & Co. KGaA, Germany), 191 which recorded data every 10 minutes. Because this study was conducted in the 192 closed greenhouse, the influence of wind was negligible. The microlysimeters were 193 weighed at 12:00 h every 2 days with an electronic precision balance of 0.01 g. 194 Evaporation losses were determined by applying the differences between the masses 195 of the microlysimeter. Simultaneously, thermal radiation of the soil surface were 196 recorded using a thermal camera (Testo 865, Testo SE & Co. KGaA, Germany). Real 197 soil surface images were also recorded by a high pixel digital camera (Eos 850D, 24.1 198 million pixels, Japan) to monitor the development of moss coverage and soil surface 199 crack. The vertical distance between the camera and the soil surface was set at 0.2 m. 200 All experiments were carried out within 10 min to ensure the accuracy of the data. 201 The initial variables were the number of moss spores, the number of biochar 202 applications, and sampled soil properties. Various evaporation indicators were 203 12 determined, including cumulative evaporation (Ec, mm), cumulative evaporation of 204 bare soil (Ec0, mm), evaporation rate (E, mm·d -1 ), relative cumulative evaporation 205 (Ec/Ec0, -), and mean soil water content (SWCmean, g·g -1 ). Ec was determined by the 206 total evaporation losses during the entire evaporation time. E was calculated as the 207 ratio of the cumulative evaporation to evaporation time. Ec/Ec0 was calculated as the 208 ratio of cumulative evaporation to bare soil evaporation for each treatment. where Ec is the cumulative evaporation (mm), Em is the total evaporation loss during a 212 given duration (g), is the density of water (1.0 g·cm -3 ), and r is the radius of the 213 microlysimeter (10.0 cm). 214 where SWCmean is the mean soil water content (g·g -1 ), Sm is the initial soil mass (g), Sw 216 is the increased water mass when microlysimeter saturated (g), and is the initial 217 soil water content (0.15 g·g -1 ). 218

Image processing 219
Information on the development of moss coverage and soil surface crack was 220 extracted from real surface images using a Python script. The image processing was 221 conducted in the following two parts. 222 Part 1: Crack-recognition module. The soil surface cracks were discriminated 223 using a method similar to that used in Wang et al. (2017). Initially, this script picked 224 13 and cropped the edge areas and reserved the soil areas by threshold image 225 segmentation and edge recognition. Subsequently, all the source images were 226 converted into grayscale images, which switched from color differences between soil 227 and crack to pixel difference. Owing to the visible difference in grayscale between 228 cracks and soil blocks, these grayscale images were segmented into binary images by 229 an automatic threshold method, called the OTSU method, implemented in the 230 software package, Python OpenCV. If the grayscale values of some areas were higher 231 than the threshold value, these areas were judged as cracks and shifted to black. 232 Conversely, those areas below the threshold value were voted as soil materials and 233 changed to white. The optical threshold value was automatically determined. It must 234 be noted that very few isolated black spots could not be distinguished owing to the 235 microtopography of the soil surface. Eventually, threshold denoising and 236 characterization of cracks were performed (Fig. S1a). Several crack parameters were 237 determined (see Section 2.4). 238 Part 2: Moss recognition module. There were obvious color differences 239 between moss and soil, and thus color segmentation was applied to extract the 240 dynamics of moss coverage. First, the mean filtering method was used to enhance the 241 moss pixels. Later, the color mode of the source images was converted to HSV mode 242 by the cvtColor function in OpenCV. The converted pictures were handled with color 243 gamut segmentation by the THRESH_BINARY function in OpenCV. If the gray 244 values were higher than the threshold value, these areas were judged as soils and 245 14 shifted to black. These areas below the threshold value were regarded as moss and 246 changed to white. The optical threshold value was also automatically determined. 247 Because the moss surface was irregular, there were a few noise points after image 248 binarization and segmentation. Finally, mathematical morphology processing was 249 conducted to fill holes after image binarization and segmentation (Fig. S1b). 250 The temporal and spatial evolution of the surface temperature for different 251 treatments were determined from these images using the IRSoft 4.5 software. The 252 thermal images of soil surface were collected using a thermal camera (Testo 865, 253 Testo SE & Co. KGaA, Germany). The IRSoft 4.5 software was used to automatically 254 analyze a thermal image. These thermal images were, then, imported into IRSoft 4.5 255 software to calculate the average temperature of the surface temperature for different 256 treatments (Fig. S1c). 257 Soil surface cracks, moss coverage, and moss growth 258 The crack ratio (Rcr) was calculated to illustrate the crack dynamics during the 259 dry-wet cycles. The crack ratio was determined as the ratio of crack pixels to total 260 pixels in the binarization images (Tang et  and evaporation calculation. We introduced the logistic growth equation to 282 characterize the dynamics of moss coverage, which included three critical 283 parameters-moss growth rate (k, cm 2 •cm -2 •day -1 ), time for mutation of growth rate (b, 284 day), and upper limit to moss coverage (M, cm 2 •cm -2 ). There was a significant linear 285 relationship between the value of Ec/Ec0 and moss coverage Mc (see Section 3.4). 286 Based on the variation of this linear fitting relationship, the cumulative evaporation 287 16 was calculated as follows: 288 where E is the predicted evaporation (mm), Ec0 is the evaporation of bare soil (mm), 290 Mc is the moss coverage (cm 2 •cm -2 ), and a and c are empirical coefficients (-). 291 Data collected over the 88 days included evaporation for each treatment within 292 the given duration Ec, evaporation of bare soil Ec0, and moss coverage with each day 293 Mc. The data were applied to test the empirical model. 294

Data analysis 295
All statistical analyses were performed using Python version 3.6. Partial 296 correlation and ANOVA analyses were conducted using a significance level of 0.05 297 and 0.01, respectively. The partial correlation coefficient measured the degree of 298 association between the two variables, with the effects of a set of controlling variables 299 removed. Means were compared using the least significant difference determined by 300 one-way analysis of variance. Regression analysis was conducted to determine the 301 relationships between each variable, and the determination coefficient (R 2 ) and root 302 mean square error (RMSE) were used to evaluate the performance of the applied 303 regression equations. We used structural equation modeling (SEM) to separate the 304 effects of moss colonization and biochar application on evaporation of 305 carbonate-derived laterite. The SEM was conducted using AMOS 2.1 software. 306

Results 307
Dynamics of moss coverage 308 The moss coverage versus time under different biochar application (B0, B1, B2, 309 and B3) and moss spores (M1, M2, and M3) during six dry-wet cycles was indicated 310 in Fig. 2. The soil surface of the moss colonization treatments showed similar 311 changing patterns in the moss growth during each dry-wet cycle. In particular, moss 312 coverage considerably increased, but dramatically decreased after approximately 3 313 days of evaporation. Another notable finding is that the mean growth rate of moss in 314 the first three dry-wet cycles (D-W1, D-W2, and D-W3) was higher than in the last 315 two dry-wet cycles (D-W4 and D-W6) and, therefore, it appears that the dynamics of 316 moss coverage were greatly affected by evaporation times and dry-wet cycles. 317 The logistic growth equation could estimate the dynamics of moss coverage, but 318 there were some restrictions in explaining the wilt of moss under drought conditions 319 (Table 3). The determination coefficients R 2 (dimensionless) ranged from 0.60 to 0.74 320 and the RMSE (dimensionless) values were between 0.02 and 0.16. The M values 321 (upper limit to moss coverage) and k values (moss growth rate) increased with the 322 initial amount of moss spores, but they were not affected by the biochar application 323 for all treatments. 324 The initial amount of moss spores and soil water content significantly increased 325 moss coverage with partial correlation coefficients (pr) of 0.64 and 0.48, respectively 326 (Table 4). However, moss coverage significantly reduced with the increase of biochar 327 18 application (pr = -0.11, P < 0.01) and atmospheric relative humidity (pr = -0.33, P < 328 0.01). These results suggested that the initial amount of moss spores and soil water 329 content had the largest impact on the growth of moss, whereas the biochar application 330 had the lowest influence. These trends were also confirmed by the functional 331 relationship among initial amount of moss spores, soil water content, air temperature, 332 and atmospheric relative humidity (Mc = 0.08 + 0.001Mb + 0.12SWCmean -0.04Tair -333 0.01RHmean, R 2 = 0.56), where Mb is the initial amount of moss spores, SWCmean is the 334 mean soil water content during the drying process, Tair is the air temperature, and 335 RHmean is the mean atmospheric relative humidity. 336

Characteristics of soil surface cracks 337
The ratio of soil surface crack fluctuated significantly during dry-wet cycles. For 338 example, the mean ratio of soil surface crack at the end of the D-W6 was 42.5% 339 higher than that at the end of the D-W1 (Fig. 3). In each dry-wet cycle, there were 340 significant differences in the mean soil surface crack ratio between different initial 341 amounts of moss spores (P < 0.05), especially for the M3 treatment. In contrast, there 342 were no significant differences in the mean soil surface crack ratio between different 343 biochar applications (P > 0.05). The further fitting analysis showed that the mean ratio 344 of soil surface crack declined linearly with an increase in the number of moss spores 345 for the B0, B1, B2, and B3 treatments (R 2 = 0.76, 0.99, 0.25, and 0.59, respectively) 346 ( Fig. 4). The mean soil surface crack ratio increased linearly with an increase in the 347 amount of biochar application for the M1 and M3 treatments (R 2 = 0.25 and 0.96, 348 19 respectively). Nevertheless, the mean ratio of soil surface crack was not influenced by 349 the biochar application for the M0 and M2 treatments (R 2 = 0.02 and 0.08, 350 respectively). It seems that the moss colonization significantly impeded the 351 development of soil surface cracks in most cases, but the effects of biochar 352 application on the development of soil surface cracks could not reach a clear 353 conclusion. 354 The above fitting results were slightly contradicted the results of the partial 355 correlation analysis. The partial correlation analysis indicated that the biochar 356 application and initial amounts of moss spores did not have significant effect on the 357 development of soil surface cracks (Table 4). This was mainly because the fitting 358 analysis did not considered the effects of moss coverage on the formation of soil 359 surface cracks (see Section 3.1). On the other hand, the ratio of soil surface crack 360 significantly decreased with an increase in the mean atmospheric relative humidity (pr 361 = -0.18, P < 0.01), but it significantly increased with an increase in the mean air 362 temperature (pr = 0.34, P < 0.01). 363 Relationship between growth of moss, soil water content and crack development 364 The ratio of soil surface crack increased significantly with the decreasing soil 365 water content (Fig. S2). The soil water content at the occurrence of cracking was 366 defined as critical water content (θc). The θc in the first dry-wet cycle (D-W1) does 367 not synchronize with that in the sixth dry-wet cycle (D-W6). For D-W1, θc varied in 368 the range of 39%-43%. On the other hand, θc ranged from 58% to 64% for all 369 20 treatments in D-W6. The most interesting discovery was that the moss growth could 370 impede the formation of soil surface cracks for most dry-wet cycles, except for the 371 first and fourth dry-wet cycles (D-W1 and D-W4) (Fig. 5). The main reason for the 372 exceptions was that the moss coverage in D-W1 was too low to affect the formation 373 of soil surface cracks, whereas the formation of soil surface cracks in D-W4 was 374 inactive owing to the relatively low evaporation losses. 375

Soil evaporation process 376
Evaporation processes had gone through a constant rate stage and a falling rate 377 stage in most dry-wet cycles (D-W1, D-W2, D-W3 and D-W5) ( Table S1). The 378 evaporation remained at a constant rate stage in D-W4 and D-W6 due to the 379 decreasing atmospheric energy (Fig. S3). Despite the existence of moss colonization 380 and biochar, the cumulative evaporation showed fitting relationships with evaporation 381 time, and R 2 ranged from 0.77 to 0.97 (Table S1). For all dry-wet cycles, the lowest 382 soil evaporation was observed in treatment with moss colonization and biochar 383 application. For instance, the value of λ1 for B1M1 treatment was the lowest in D-W1, 384 D-W2 and D-W3, which were 5.87, 4.27 and 4.75, respectively. The value of λ2 for 385 B1M3 treatment was the lowest in D-W4 and D-W6, which were 0.66 and 0.88, 386 respectively. These results showed that the moss colonization and the biochar 387 application could reduce evaporation losses. 388 The relative cumulative evaporation Ec/Ec0 (the ratio of cumulative evaporation 389 Ec to cumulative evaporation of bare soil Ec0) was calculated to eliminate the 390 21 influence of the atmospheric condition. The dynamics of relative cumulative 391 evaporation for the M1, M2, and M3 treatments largely depended on the changes in 392 moss coverage (Fig. S4). For all treatments of moss colonization and biochar 393 application, the Ec/Ec0 was significantly lower in D-W6 than D-W1, and the moss 394 coverage was significantly higher in early D-W6 than the initial moment of D-W1 395 (Fig. 6). In the other words, the moss colonization and biochar application over six 396 dry-wet cycles decreasing evaporation losses by 4.9%-28.3%. However, the effects of 397 moss colonization and biochar application on the relative cumulative evaporation 398 were hard to distinguish due to moss coverage was fluctuated in six dry-wet cycles. 399 The further fitting analysis indicated that moss colonization and biochar 400 application could decrease greatly evaporation losses. The fitting analyses showed 401 that the relative cumulative evaporation linearly decreased with an increase in moss 402 coverage for the B0, B1, B2, and B3 treatments (R 2 = 0.34, 0.44, 0.46, and 0.49, 403 respectively) (Fig. 7). The relative cumulative evaporation for different biochar 404 applications (B0, B1, B2, and B3) decreased with increasing moss coverage by 405 27.07%, 40.69%, 37.73%, and 31.37%, respectively. The relative cumulative 406 evaporation was 1 as the moss coverage reached a critical value. In other words, the 407 moss enhanced the evaporation losses when the moss coverage was below the critical 408 value, and reduced the evaporation losses when the moss developed to the critical 409 moss coverage. The critical moss coverage was 4.51%, 2.51%, 3.01%, and 0% for B0, 410 B1, B2 and B3 treatments, respectively. The critical moss coverage decreased with the 411 22 increasing biochar dosage even further to 0 percent. Partial correlation analyses 412 indicated that the mean evaporation rate significantly increased with air temperature, 413 the initial moss spore amount, and soil water content (pr = 0.42, 0.11, and 0.12, 414 respectively) (Table 4) In response to this, we drew a mind map of the effects of moss colonization and 502 biochar application on the evaporation losses of carbonate-derived laterite, and 503 constructed a SEM based on the mind map and partial correlation analyses (Fig. 9). 504 Moss colonization and biochar application played an important role in evaporation 505 processes in three primary ways. First, the increase of initial amount of moss spores 506 promotes the growth of moss (pr = 0.64, P < 0.01), but biochar application reduced 507 slightly moss coverage (pr = -0.11, P < 0.01). Second, moss covers promoted the 508 development of soil surface cracks (pr = -0.13, P < 0.05), and the increasing soil 509 surface cracks enhanced evaporation losses (pr = 0.32, P < 0.01). Moreover, the moss 510 decreased evaporation losses by reserving soil water content (pr = 0.47, P < 0.05), and 511 the increasing soil water underlying the moss layer decreased evaporation losses (pr = 512 -0.21, P < 0.05). Finally, the moss increased soil surface temperature (pr = 0.11, P < 513 0.01), and the increasing soil surface temperature enhanced evaporation loss (pr = 514 0.28, P < 0.01). In addition, our SEM explained 71% of variation in soil evaporation 515 27 under moss colonization and biochar application. Moss spores (standardized 516 coefficients = 0.48, P < 0.001) and soil water content (standardized coefficients = 0.6, 517 P < 0.001) had positive effects on moss growth. Atmospheric humidity (standardized 518 coefficients = -0.12, P < 0.001) and atmospheric temperature (standardized 519 coefficients = -0.37, P < 0.001) had negative effects on moss growth. Moss 520 (standardized coefficients = -0.98, P < 0.001) and biochar (standardized coefficients = 521 -0.17, P < 0.001) had negative effects on soil evaporation. 522

Implications for and limitations to field situations 523
The prospect of applying moss colonization to hold water in karst lands is 524 promising. Moss colonization could help to create supplementary approach during the 525 high-cost afforestation in karst lands. The results of this study showed that growing 526 moss significantly increased soil water holding capacity by reducing evaporation 527 losses. In addition, as confirmed by the different slopes of the linear fitting equations, 528 and the role of moss in reducing soil evaporation could be slightly enhanced by 529 biochar application ( Fig. 6 and 7). One of the most important findings in this study is 530 that moss colonization could inhibit significantly surface crack development of 531 carbonate-derived laterite (Fig. 3-5). This means that soil surfaces without moss 532 colonization behave in completely different ways from those that retain them. 533 Our results shows that moss growth is not merely linear and irreversible, but that 534 the moss can degrade after drying and recover after wetting. The relationship between 535 moss growth and soil evaporation, therefore, fluctuates (Fig. 2). In Chen et al.'s (2018) 536 28 study, however, the model hypothesis was that moss coverage was always increasing 537 and did not respond to natural disturbances (that is, dry-wet cycles). Therefore, the 538 response of moss coverage to water stress should be considered in the eco-hydrology 539 model to improve accuracy. 540 If moss colonization was used as a supplementary approach to afforestation in 541 karst lands, the dynamic and ever-shifting relationships between moss growth and soil 542 evaporation under dry-wet cycles should be emphasized. First, this study showed that 543 moss growth was highly vulnerable to water deficits, especially in their early stages 544 (Fig. 3). Therefore, the dampening effects of moss colonization on soil evaporation 545 performed poorly with low coverage but worked well with high coverage (Fig. 7). A 546 few previous studies have discussed the effects of biocrust with a high moss coverage 547 (60% or 75%) on soil water movement (Berdugo et al., 2014;Zhao et al., 2014). 548 However, moss coverage in this study was only up to 25%, due to cultivation time. It 549 is also important that these findings be extended to other moss species. Hypnum 550 Hedw., which was used in this study, might be unique, or it might represent a broader 551 pattern among other moss species. Previous experiments and this study had ignored 552 the moss transpiration due to the limitation of measurement method (Table 1)

Conclusions 559
The effects of moss colonization and biochar application on the evaporation 560 processes of carbonate-derived laterite were studied using four moss spore amounts (0, 561 30, 60, and 90 g·m -2 ) and four biochar application levels (0, 100, 400, and 700 g·m -3 ) 562 during six dry-wet cycles. The following results regarding the effects of moss 563 colonization and biochar application were obtained: 1) The increase of initial amount 564 of moss spores promotes the growth of moss (pr = 0.64), but biochar application 565 reduced slightly moss coverage (pr = -0.11); 2) Moss colonization decreased clearly 566 soil surface cracks (pr = -0.13), whereas biochar application had no distinguishable 567 influences on soil surface cracks (P > 0.05); 3) The relative cumulative evaporation 568 for the different biochar applications decreased with increasing moss coverage by 569 27.07%, 40.69%, 37.73%, and 31.37%, respectively. Biochar application reduced 570 critical moss coverage associated with inhibition of evaporation by 33.26%-44.34%; 4) 571 The mean surface temperature increased with an increase in the moss coverage (pr = 572 0.11); 5) A simplified empirical evaporation model could accurately estimate the 573 evaporation losses that were affected by the moss colonization and biochar 574 application, and the R 2 values ranged from 0.94 to 0.99. 575 These results suggested that moss colonization and biochar application played an 576 important role in evaporation processes in three primary ways. The artificially 577 30 cultivated moss, which was induced by moss spores and biochar, could decrease soil 578 evaporation by reducing soil surface cracks, increasing soil moisture and soil surface 579 temperature. It was concluded that the combination of moss colonization and small 580 amount biochar could be used as a supplementary approach to facilitate soil water 581 preservation in karst lands. However, there are still many questions to be answered in 582 future studies, including the role of moss colonization in the whole karst hydrology 583 (that is, infiltration, interception, runoff, and soil erosion).

Conflict of interest 592
The authors declare that they have no known competing financial interests or 593 personal relationships that could have appeared to influence the work reported in this 594 paper. 595   Moss spore (g m -2 ) Biochar application (g m -3 )

Fitting equation
Step: Step: = 0 ( + )  daily mean evaporation rate (mm day -1 ); Significant at *P < 0.05, **P < 0.01.     Moss coverage (%) Fig. 9 Schematic or mind map about the effects of initial moss spores amounts, moss coverage and biochar application on soil evaporation 768 processes of carbonate-derived laterite during dry-wet cycles. Numbers adjacent to arrows are standardized path coefficients of the relationship. 769 R 2 = the proportion of variance explained. P values are as follows: * < 0.05; ** < 0.01; *** < 0.001. Blue and red lines are positive and negative 770 relationships, respectively. 771