some distance above and below. Although the river did not reach as high a stage as the previous year, the current along the revetment and for some distance above had materially increased and was quite rapid. The result was that caving occurred at numerous places.

The first failure of the bank was noticed early in May, 1884, a little above Winchester street, and very soon afterwards, as the water fell, active caving set in along nearly the entire bank, being most active from Wolf River to the foot of Market street. Along this stretch the average caving was 85 feet, with a maximum of 140 feet just below Wolf River. From Market street to Poplar street only slight upper bank caving took place, as here the revetment was practically of double thickness and extended to near the top of the bank. At the foot of Washington street a cave 300 feet long by 30 to 50 feet back destroyed a portion of the grain elevator. Below the foot of Adams street and along the Hanauer Oil Mills was a cave 150 feet long by 25 feet wide, taking in about all of the roadway between the mill and the top of the bank. Still farther below, at the foot of Jefferson street, there was a cave in the paved levee 130 feet long by 75 feet back in the middle; this was a top bank cave not extending to deep water. Above Wolf River the bank caved actively for a distance of about a quarter of a mile.

The caving this year practically destroyed the entire existing revetment, proving that it was ineffective in protecting a caving bank when the current along it became quite rapid.

CAUSES OF FAILURE

There were probably several causes for this failure, the principal being that the mats covered only the upper portion of the subaqueous slope, leaving an unprotected belt between them and deep water from 75 to 125 feet wide, which during flood stages was scoured, undermining the mats and allowing them to slide down the slope. Another cause may have been the scour at the intervals left between the mats, for in sinking this type of narrow mats in deep water, frequently with strong currents, there could be no certainty that proper laps were always made, especially when an average of only 12 to 15 feet was allowed for them. A third cause was the incomplete upper bank revetment. This was entirely absent below Poplar street, where three large caves occurred, probably all from this cause.

In general, the experience here gained from the employment of

small, detached mats showed that they were expensive to construct and difficult to place in a manner certain to cover the entire bank and that they were of insufficient length to reach to deep water, as was essential. A change in plan was therefore necessary; either the size of the detached mats must be largely increased or con tinuous mats must be used, and as the latter method was the cheaper in construction and could be more rapidly done, it was employed in future work.

The location of all the early work in Memphis Harbor is shown on Plate XXIV.

EXPERIMENTS WITH SCREENS

(Plate XXV.) The revetment work just described was situated entirely below Wolf River, but during the progress of the work a series of experiments was carried on with another type of work above that river, where the bank was also caving.

At the beginning of the season of 1881 the main portion of the original project for the protection of the bank below Wolf River having been completed, and the results thus far having proved satisfactory in averting caving, it was decided to begin the protection work along the hitherto rapidly caving bank above Wolf River, and as the mattress work was quite expensive it was deemed expedient to experiment with some less expensive work, and to this end open dikes and floating screens of various designs were placed on the bar at the head of the caving, with the object of checking the flow of the current and thus stop the caving. No detailed description of the screen can be given, but in general they were built of poles or sawed lumber, usually 3 by 6 inches, spiked and bolted edgeways to longitudinal strips and spaced apart their own widths. One edge of these screens was anchored to a sill by hinged fastenings, and the other was supported by barrels or other buoys to keep it floating on the surface.

The first of these structures was located just above the head of the caving, or about 3,000 feet above Wolf River. It was 450 feet long, extending out from the top of the bank and nearly perpendicular to it. The foundation sill was a brush mattress 22 feet thick and 60 feet wide up and down the river. At low water about 200 feet of this sill was submerged, the balance being on the dry bank. The entire sill was ballasted with stone, and on it were placed ten floating screens. Those on the bank above the low-water line were anchored by chains to posts, the bottoms of which were

bolted to buried logs. The screens in the water were each hinged by two pieces of chain attached to cribs of poles 16 by 4 by 4 feet, which were loaded with stone and dropped in place by means of a pile-driver hoist.

A second line of screen was begun 600 feet below the first and four screens placed, when rising water compelled the closure of the season's work.

After a fall of the water and a subsequent rise, it was found that the screen placed above the low-water line had silted up and did not rise, and hence had become ineffective; the other screens, however, had responded well to the changing stages of the river and had caused some considerable deposits. It was therefore decided to replace the screen work above low water with a pile dike extending from the top of the bank to the low-water line. This dike was built of two rows of piling 8 feet each way, with tops cut off 8 feet below high-water mark to permit the passage of drift. It was further intended to brace the two rows of piles together and place waling pieces along the lower row, but owing to the limited funds and the urgency of other work this was not done, neither was anything more done toward the completion of the open dikes or screen work. The total of this work for the season was the driving of fifty-two piles.

During the next high water drift carried away the buoys, and the screens were driven to the bottom by the force of the current and were there silted up. Consequently, they did not again rise, acting henceforth merely as a sill mat, and probably were somewhat effective in averting bank caving for some distance below, for this did not again seriously cave. On account of the difficulty in keeping the screens afloat, this experiment was not repeated.

EARLY REVETMENT WORK IN PLUM POINT REACH

A description has been given of the early revetment work in Memphis Harbor carried on by the Engineer officers stationed at that place. While this work was in progress the Mississippi River Commission was organized and began the work of improving Plum Point Reach.

The first work was begun late in the season of 1881. A method successfully used on the Missouri River for checking bank erosion and for causing deposits was tried at the Head of Bullerton Bar. This consisted of the erection of wire screens having meshes 3 feet square and hung from piles, the screens being hung with the bottom.

edge at about the low-water plane. On the Missouri River the screens in twenty-four hours are said to have become so completely filled up with roots, fibres, and other matter held in suspension as to make them almost impervious, thus checking the current and causing deposits. On this river, however, they failed utterly. No results were obtained except the washing out of the suspension piles by the next high water, and the experiment was never repeated. A small mat 500 feet long was built in connection with this work and it also was destroyed.

After the high water of 1882 had passed, revetment work was resumed. The mattress work done now differed from that above described in Memphis Harbor, in that, instead of being composed of separate small mats, it was what we may call a continuous mattress, made floating on the surface of the water directly over where it was to lie when sunk. For its construction a barge was prepared by building inclined ways across it. This barge was towed to the point that was to be the upstream end of the mattress and was there moored and fastened by cables to the bank, with its length per pendicular to the shore and the ways inclining upstream. (PlateXXVI.)

Upon these ways the construction of the mattress was then begun at the upstream end of the ways. When enough had been built to nearly cover the barge, the upstream edge or head of the mat was fastened by cables to the bank independently of the barge, and when this was done the barge was allowed to float or was pulled downstream from under the mattress, which was thus launched into the water, upon the surface of which it floated, being held in place by the above-mentioned cables and moorings. When the barge had slipped downstream far enough so as to leave only a narrow strip of mattress still resting on the ways, the construction of the mattress was resumed from this narrow strip and continued until the barge was nearly covered again, when a new launch was made, and so on. The width of the mat that could thus be made was, of course, limited to the length of the barge, but the construction of the mat could be continued to any desired length, even to several thousand feet if necessary. When the desired length had been built, a final launch was made and the barge removed, which left the mat floating on the surface and ready for sinking.

During the seasons of 1882 and 1883 two types of mat were used (Plate XXVI), one consisting of brush and poles, with the brush woven alternately over and under the poles like basket work, and

the other type with a ground work of wire net with brush filling on top. In the former the poles were laid longitudinally, spaced from 7 to 8 feet apart, and were spliced together, butt to top, with spikes and lashed together with wire, thus forming lines of continuous poles throughout the length of the mat. These poles corresponded to the warp in textile fabrics, and over and under these was woven the woof of brush, each piece of brush being woven alternately over and under the poles, care being taken to have all butts and tops on the upper side of the mat and that no contiguous butts were upon the same pole. The butts were preferred on the top side, so that they would not interfere in launching. Tops of the brush were also placed on top, because they were too yielding to hold up the stone ballast if unsupported by the pole.

The only resistance to longitudinal rupture was such as was afforded by the splices of the poles, while transversely it depended for strength simply on the friction of the brush against the poles and against each other. Later, after several mats had been ruptured in sinking, two or more wires, according to the strength of current and depth of water, were run the full length of mat, being fastened to it about every 25 feet; cross wires of the same size were also placed every 25 feet. With poles not over 3 inches thick at the butts some degree of flexibility was obtained, but at the expense of strength. With larger poles, such as were mainly used, the mat was somewhat stiff and unyielding. In launching off, some of the poles would break, as they were too stiff to assume the proper curvature. This, of course, reduced its longitudinal strength. The second type of mat consisted of a wire net base, made of No. 8 galvanized wire with meshes 3 feet square. The brush was laid on top of this net in a single layer 12 to 15 feet wide. This layer of brush was tied down by poles perpendicular to the brush and spaced 5 feet apart. Each pole had its butt end thrust over the brush and under the transverse wire of the net. The tops of the poles were then pulled down over the brush and wired to the net, thus holding down the brush. The usual care of so distributing the brush that too many butts would not lie on the same line was taken. These mats were built on mat barges specially designed for this work, the nets being woven on the barges as fast as construction of the mat progressed, and the brush was delivered upon the net by overhead carriers.

This type formed a very flexible mat, but of little strength, as it was almost impossible to so weave the net that the wires had an