周懷宇 劉海龍

物聯(lián)網(wǎng)（internet of things, IoT）技術(shù)的進(jìn)步在改變城市空間使用模式的同時，也推動著風(fēng)景園林走向在線化、定量化。在智慧城市空間新技術(shù)和功能新需求的雙向驅(qū)動下，數(shù)據(jù)空間與實(shí)體空間開始融合疊加，城市景觀也面臨著進(jìn)一步的數(shù)字化重構(gòu)和轉(zhuǎn)型[1-2]。在線化的趨勢要求風(fēng)景園林師走出傳統(tǒng)空間設(shè)計的舒適圈，主動采集結(jié)構(gòu)化信息用于設(shè)計循證、評估與 管理[3]。

筆者將中小尺度的綠色雨水設(shè)施（green stormwater infrastructure, GSI）與在線化技術(shù)結(jié)合，通過文獻(xiàn)整合、原型總結(jié)等方法，為風(fēng)景園林師提供一個初步的GSI在線監(jiān)測指南（online monitoring guidelines）。首先對風(fēng)景園林師較陌生的GSI的監(jiān)測原理、傳感設(shè)備、計算方法進(jìn)行了詳細(xì)論述，重點(diǎn)探討了3種將傳感與實(shí)體空間結(jié)合的監(jiān)測系統(tǒng)原型（monitoring system prototypes），最終闡述了一種可行的物聯(lián)網(wǎng)數(shù)據(jù)傳輸方案。

作為系列研究的最新成果，本研究基于清華校園GSI水量控制監(jiān)測的實(shí)踐成果[4]29，在原理、方法、原型方面進(jìn)行了較大程度的深化，從特定案例分析轉(zhuǎn)向更具普適性的設(shè)計原則探討。

1 GSI在線監(jiān)測

作為發(fā)揮著雨洪調(diào)節(jié)功能的城市基礎(chǔ)設(shè)施，GSI建成后有復(fù)雜的水文響應(yīng)過程和明顯的功能衰退現(xiàn)象，其效能的發(fā)揮依賴良好的后期維護(hù)、監(jiān)測與反饋[5]。由于傳統(tǒng)設(shè)計思路和技術(shù)的限制，目前GSI項(xiàng)目往往只關(guān)注空間設(shè)計和模型模擬，而忽視后期監(jiān)測，引發(fā)設(shè)計方案難核查評估、設(shè)計更新無據(jù)可依的新問題。智慧城市中的新GSI建設(shè)亟須與在線監(jiān)測結(jié)合，避免符號化的低質(zhì) 擴(kuò)張[6]。

當(dāng)前，水文監(jiān)測技術(shù)主要用于以下4個方面：1）指導(dǎo)城市水文響應(yīng)的理論研究，利用數(shù)據(jù)監(jiān)測還原多種尺度下的城市產(chǎn)匯流機(jī)制[7-8]；2）輔助高精度的水文機(jī)理模型開發(fā)，用于綜合水項(xiàng)目管理與災(zāi)害預(yù)報[9-10]；3）監(jiān)測數(shù)據(jù)用于績效評估及后期維護(hù)，識別多種綠色、灰色水設(shè)施的效能衰減[11-12]；4）在線監(jiān)測技術(shù)方案與產(chǎn)品研發(fā)[13-14]。

雖然相關(guān)技術(shù)不斷進(jìn)步，但由于學(xué)科關(guān)注點(diǎn)的差異，水文、物聯(lián)網(wǎng)學(xué)科提供的在線監(jiān)測技術(shù)和分析方法尚未與GSI空間設(shè)計結(jié)合，即缺少將技術(shù)方案整合到實(shí)際設(shè)計的相關(guān)指導(dǎo)。風(fēng)景園林學(xué)科作為理論、技術(shù)與設(shè)計的橋梁，應(yīng)作為統(tǒng)籌者，探索GSI在線監(jiān)測的落地方案。簡要列舉當(dāng)前國內(nèi)外風(fēng)景園林師主導(dǎo)設(shè)計建設(shè)的、基于IoT的GSI項(xiàng)目：美國威蘭偌瓦大學(xué)和密歇根大學(xué)開展的智能雨洪管理創(chuàng)新課題研究和校園試點(diǎn)項(xiàng)目建 設(shè)[15]；美國馬里蘭大學(xué)團(tuán)隊(duì)設(shè)計的智能綠色屋頂[16]，美國Opti公司智能GSI在線監(jiān)測項(xiàng)目[17]（圖1-1）以及清華大學(xué)勝因院雨洪管理監(jiān)測系統(tǒng)[4]30（圖1-2）。

與水文學(xué)科所開展的諸多GSI監(jiān)測實(shí) 驗(yàn)[11,18]稍有不同，風(fēng)景園林學(xué)科開展戶外GSI監(jiān)測更關(guān)注設(shè)施布局合理性驗(yàn)證、超標(biāo)徑流控制以及景觀績效評估。同時，戶外監(jiān)測面臨著諸多挑戰(zhàn)：1）長周期、多雨型的監(jiān)測數(shù)據(jù)收集需要制定在線化、系統(tǒng)化的監(jiān)測方案，保證數(shù)據(jù)完整有效；2）需要細(xì)致到指標(biāo)、場地和設(shè)備裝置的監(jiān)測設(shè)計，滿足耐用性及戶外通訊需求；3）滿足經(jīng)濟(jì)、美觀的要求，并在一定程度上達(dá)到生態(tài)教育與展示的目的。

因此，GSI在線監(jiān)測系統(tǒng)設(shè)計往往需要：1）理清監(jiān)測指標(biāo)與計算原理；2）掌握基本的監(jiān)測方法并合理選取傳感設(shè)備；3）基于GSI實(shí)際場地設(shè)計需求制定可落地的技術(shù)方案；4）合理利用實(shí)時通信手段完成數(shù)據(jù)傳輸和收集。

2 GSI在線監(jiān)測指標(biāo)、原理與方法

2.1 GSI在線監(jiān)測指標(biāo)

首先需要明確，GSI在線監(jiān)測指標(biāo)為可長時間采集的結(jié)構(gòu)化數(shù)據(jù)，指標(biāo)具備變化步長短或突變的特點(diǎn)。結(jié)構(gòu)化數(shù)據(jù)也稱作行數(shù)據(jù)，可用二維表結(jié)構(gòu)實(shí)現(xiàn)邏輯表達(dá)。當(dāng)前大部分成熟傳感設(shè)備都僅限于采集溫度、濕度、位置、光強(qiáng)、壓力、速率、降雨量等具有突變特性結(jié)構(gòu)化數(shù)據(jù)[19]。結(jié)構(gòu)化數(shù)據(jù)常以電信號、光信號、電磁波等形式獲取，數(shù)據(jù)體量小，分析處理速度快。表1列出了在GSI在線監(jiān)測中可結(jié)構(gòu)化實(shí)時采集的指標(biāo)，共分為氣象、水質(zhì)、水量控制（地表及地下）等方面，能夠比較全面地反映GSI雨洪管理的生態(tài)響應(yīng)。

表1 智能GSI在線監(jiān)測指標(biāo)及步長Tab. 1 Indicators and steps of GSI online monitoring

2.2 基本原理——基于水量平衡的出入流估算方法

出入流監(jiān)測是GSI水量、水質(zhì)監(jiān)測的核心，是徑流控制的核心評價指標(biāo)。相比于水文學(xué)科關(guān)注水文機(jī)理、規(guī)律、過程的定量分析[20]459，風(fēng)景園林學(xué)科開展戶外GSI監(jiān)測時較難做到全面的數(shù)據(jù)收集及嚴(yán)格的實(shí)驗(yàn)條件控制，因此需要對GSI復(fù)雜水文過程進(jìn)行簡化，本研究提供一種基于降雨—滯蓄—下滲—溢流水量平衡經(jīng)典公式[21-22]的GSI出入流簡化計算方法。

對于簡單的封閉流域型GSI，其入流量Qin等于凈空的降雨量Qr。為提高準(zhǔn)確性，可同時監(jiān)測樹冠截留后的穿透降雨量Qca，根據(jù)凈空即遮擋面積按比例計算入流（公式1）。其中i為凈空降雨的面積系數(shù)。而對于絕大多數(shù)收集徑流邊界不清晰的開放流域式GSI，可用降雨過程結(jié)束后土壤持水增量ΔQsw、進(jìn)入填料層的下滲量Qinf、溢流量Qof之和估算得到（圖2），其中土壤持水增量ΔQsw為土壤水分（體積分?jǐn)?shù)）增量ΔIsw與土壤層厚度Hs的乘積（公式2）。

在這一算法中，土壤持水增量ΔQsw是對降雨過程中土壤滲潤、滲漏和滲透等復(fù)雜情況的簡化；下滲量Qinf指的是進(jìn)入填料層被收集到徑流量，可通過獨(dú)立監(jiān)測井收集，是對降雨過程中下滲性能的簡化。嚴(yán)格意義上的出流量Qout僅指GSI超標(biāo)外排或溢流口的流量Qof，進(jìn)入GSI填料層下滲量Qinf應(yīng)視為徑流消納的一部分，不算做出流量。因?yàn)檎`差小于設(shè)備檢測本身的誤差，降雨過程中的蒸發(fā)量可忽略不計。

進(jìn)一步結(jié)合降雨過程分析相關(guān)可以計算得到的數(shù)據(jù)：1）非降雨時土壤溫濕度數(shù)據(jù)可以反映場地光照及溫度條件，輔助植物生長分析；2）降雨開始后，GSI的表層土壤（厚100～200 mm）濕度變化反映主要的入流來向，可反映地形設(shè)計合理性和施工質(zhì)量；3）GSI多層土壤（厚100～200 mm）濕度變化亦可反映水分的水平及垂直運(yùn)動趨勢；4）在有積水的條件下，GSI表層土壤水分體積分?jǐn)?shù)為最大持水量，對應(yīng)穩(wěn)定階段的下滲速率。

2.3 GSI在線監(jiān)測方法

2.3.1 氣象與樹冠截留

氣象信息獲取一般采用便攜式或小型氣象站，可集成溫度、濕度、風(fēng)向、風(fēng)速、太陽輻射、雨量、氣壓、PM2.5/PM10等多項(xiàng)要素，采用通用無線分組業(yè)務(wù)（GPRS）遠(yuǎn)距離數(shù)據(jù)通信。GSI雨量監(jiān)測常用翻斗式及光電式雨量計，要注意傳感器的安裝高度（0.7 m以上）并保證周邊各個方向10 m左右范圍內(nèi)沒有障礙物阻擋雨水進(jìn)入，確保雨量監(jiān)測的準(zhǔn)確 （圖3-1）。蒸發(fā)量傳感器體量較大，需要一定的遮蔽，避免雨水進(jìn)入（圖3-2）。具體安裝及設(shè)計要求需要根據(jù)監(jiān)測場地的具體情況制定，并參考行業(yè)標(biāo)準(zhǔn)《GB/T 21978.5—2014降水量觀測儀器》。

樹冠截留影響著GSI所在場地的一系列水文過程，可通過監(jiān)測定量研究樹冠層對降雨的再分配作用。樹冠穿透降雨的監(jiān)測方法與雨量監(jiān)測方法相同，但翻斗式雨量計不適宜用于長期監(jiān)測樹冠截留（雨量計置于樹冠下），落葉等雜物易填滿雨量桶帶來巨大的誤差，宜采用光電式雨量計，如紅外線式雨量計（圖3-3）。

2.3.2 持水、下滲與溢流

土壤持水量常用土壤濕度傳感器監(jiān)測，只需將設(shè)備埋在不同深度的土壤層即可記錄土壤水分體積分?jǐn)?shù)的變化。土壤下滲量需要設(shè)計專門的監(jiān)測井，其原理是通過預(yù)埋管將GSI小部分下滲收集，進(jìn)而估算整體的下滲量。收集的下滲徑流或溢流指標(biāo)需要用到流量計法和液位計法2種方法：流量計法直接反映通過流量，液位計法則是將徑流暫存形成液位變化，需要搭配排空裝置（圖4-1）。

1）流量計法。GSI流量監(jiān)測常用到接觸式、管道流量計為主要渦輪流量計（turbine flowmeter）和渦街流量計（vortex flowmeter），適用于有進(jìn)出管道的GSI，優(yōu)點(diǎn)是節(jié)省空間，缺點(diǎn)是有一定的誤差且測流要求滿管。超聲波明渠流量計配合量水堰槽（三角堰、矩形堰、巴歇爾槽）非接觸式測量，不影響流速，測量結(jié)果更為準(zhǔn)確，通過液位—流量換算公式把液位高度轉(zhuǎn)換為量水槽內(nèi)的流量。流量計法要注意保持滿管的狀態(tài)和順暢的排水，尺度較小的綠色基礎(chǔ)設(shè)施宜選用渦輪、渦街計流法（圖4-2）。用地富余的GSI應(yīng)選用明渠計，保證流量槽入流平穩(wěn)，出流順暢，嚴(yán)格防止倒灌（圖4-3）。

2）液位法。該方法主要利用監(jiān)測井形成穩(wěn)定的液面，并定期排空積水。超聲波液位計精準(zhǔn)度高、誤差小于1 cm，為GSI首選的液位監(jiān)測設(shè)備。需要強(qiáng)調(diào)的是，由于發(fā)射的超聲波幅度一定，傳感器附近的反射波與發(fā)射波容易重疊，形成一定的測量盲區(qū)（圖4-4），因此須嚴(yán)格確保監(jiān)測物體與傳感器之間的最小距離，一般為300～500 mm。投入式液位計誤差較大，一般大于50 mm，不可用于計量流量和下滲量，僅僅適用于蓄水池水位監(jiān)測[23]。

2.3.3 水質(zhì)監(jiān)測

常見的現(xiàn)場取樣、低溫保存、國標(biāo)測定的水質(zhì)分析結(jié)果較為準(zhǔn)確，但也有運(yùn)輸中水質(zhì)變質(zhì)、無法時序性分析的缺點(diǎn)。在線水質(zhì)探頭一般采用電導(dǎo)率及分光光度法，水質(zhì)變化的絕對數(shù)值較不準(zhǔn)確，但能夠清晰地反映水質(zhì)指標(biāo)的相對時序變化特征。在線水質(zhì)監(jiān)測主要采用浮板等漂浮節(jié)點(diǎn)方法，一般可搭載酸堿值、總?cè)芙夤腆w（TDS）、電導(dǎo)率、溶解氧（DO）、濁度（NTU）等多種傳感器[9]。

監(jiān)測點(diǎn)布局時，除了考慮對照組的設(shè)置，還應(yīng)根據(jù)水體特點(diǎn)，選取重點(diǎn)監(jiān)測的指標(biāo)（圖5）：常見的再生水補(bǔ)給河道或曝氣濕地應(yīng)注重溶解氧的監(jiān)測；有城市雨污合流制溢流排口的水面應(yīng)注重總氮（TN）、總磷（TP）的在線監(jiān)測；有綜合植物凈化功能的濕地或人可接觸到的水面應(yīng)重點(diǎn)在線監(jiān)測化學(xué)需氧量（COD）和生物需氧量（BOD）；發(fā)揮雨水收集和突出游憩功能的公園水面須重點(diǎn)關(guān)注固體懸浮物總量（TSS）。另外，由于傳感器長期浸泡在水中，水中的污染物、微生物、藻類等會逐漸附著在傳感器表面，影響傳感器的監(jiān)測精度，因此定期維護(hù)十分重要[24-25]。

3 依據(jù)原型的GSI監(jiān)測指南

在具體實(shí)踐中，僅掌握基本的GSI監(jiān)測原理是不夠的，傳感層與空間設(shè)計的結(jié)合才是開展GSI在線監(jiān)測的難點(diǎn)。設(shè)計師需要根據(jù)場地的特定條件采用不同的設(shè)備布局方案，平衡空間營造與數(shù)據(jù)監(jiān)測。雖然設(shè)計師不需要懂得全部的傳感參數(shù)，但仍須掌握基本的平面及豎向布局原則，做好空間預(yù)留。鑒于相關(guān)技術(shù)設(shè)備與實(shí)際空間設(shè)計結(jié)合的指導(dǎo)較為缺乏，嘗試依據(jù)GSI流域開放/封閉、下滲/不可下滲及有無穩(wěn)定水體的特性，歸納出傳感層與空間設(shè)計相結(jié)合的3類設(shè)計原型，供風(fēng)景園林師參考。

3.1 原型1：無穩(wěn)定水體、不可下滲GSI

該類型傳感層設(shè)計面向封閉流域的粗放式的綠色屋頂、屋頂花園、高架橋綠化以及開放式流域的車庫頂板花園、極少數(shù)的有防滲處理的保證季節(jié)性水面的干塘和濕地等。該原型GSI土壤介質(zhì)深度?。ㄐ∮?.5 m），植物以灌木、地被為主，無大型喬木，可重點(diǎn)監(jiān)測其雨洪滯蓄能力、濕熱調(diào)節(jié)功能、土壤持水性能[20,26]；而水質(zhì)控制更適合選擇典型降雨進(jìn)行人工取樣、實(shí)驗(yàn)室分析。監(jiān)測系統(tǒng)在實(shí)際空間中的布局方式參考圖5。

1）封閉流域條件下入流明確，一般為頂空降雨或單一入口，空間設(shè)計時注意盡可能減少出流口，保證入流口高于出流口，以便于流量計安裝。

2）開放流域條件下入流較難確定，須重點(diǎn)監(jiān)測土壤持水量、溢流量及積水深度變化。不可下滲條件下，水量平衡計算公式應(yīng)做相應(yīng)的調(diào)整，GSI的入流量Qin為土壤持水增量ΔQsw及溢流量Qof之和（公式3）。

3.2 原型2：無穩(wěn)定水體、可下滲GSI

在常見的可下滲、無穩(wěn)定水體的GSI中，封閉流域的有小型的雨水花園，開放流域的一般為植草溝、旱溪、下凹綠地、雨水濕地、蓄洪濕地等。該原型中，土壤介質(zhì)復(fù)雜，包含表層石子、土壤層、大直徑的填料層等，而下滲速率則主要取決于土壤層。圖6展示了該原型下在線監(jiān)測系統(tǒng)設(shè)計以及傳感器在實(shí)際空間中的布局方式。

1）下滲量監(jiān)測需預(yù)埋收集管，收集管表層用土工布封口，均勻布滿設(shè)施土壤層下方。此時，下滲收集管與監(jiān)測井形成連通器，井中液位可反映GSI單位面積的下滲量。監(jiān)測井積水可預(yù)先安裝水泵，降雨后及時排空。

2）溢流量監(jiān)測可根據(jù)設(shè)施尺度，小型GSI選用渦輪/渦街流量計法，空間富余的GSI考慮使用明渠流量計法。

3）這類設(shè)施周圍往往存在完整的喬木—灌木—地被的植物群落，喬木樹冠對場地降雨過程影響較大，建議按照樹種進(jìn)行重點(diǎn)監(jiān)測。

3.3 原型3：濱水或有穩(wěn)定水體的GSI

與非濱水GSI相比，有穩(wěn)定水體/濱水的城市河道綠地、濕地公園、濕塘等設(shè)施，其水質(zhì)是反映徑流污染以及設(shè)施污染控制效果的最重要指標(biāo)，較為穩(wěn)定的水面為水質(zhì)在線監(jiān)測創(chuàng)造了條件。圖7展示了該原型在線監(jiān)測系統(tǒng)設(shè)計以及傳感器的布局方式。

1）水質(zhì)在線監(jiān)測常利用小型浮板或無人船開展，為保證監(jiān)測數(shù)據(jù)的準(zhǔn)確性，建議科研人員每個月進(jìn)行取樣及實(shí)驗(yàn)室測定加以驗(yàn)證。

2）靜態(tài)的水面可以利用液位計計量水量變化（枯水、常水、豐水），流動性強(qiáng)的河道更適合用流量計觀察水量變化。

3）濱水綠地一般具有完整的喬木、灌木和地被結(jié)構(gòu)，大型闊葉喬木的樹冠截留對場地降雨過程影響較大，需要重點(diǎn)監(jiān)測。另外，可選取濱水綠地地塊，構(gòu)建封閉的小流域，用矩形堰方法監(jiān)測濱水綠地產(chǎn)流量，計算濱水綠地綜合徑流系數(shù)。

4 物聯(lián)網(wǎng)傳輸方案概述

在線監(jiān)測系統(tǒng)搭建過程中涉及大量通信協(xié)議、設(shè)備及數(shù)據(jù)庫編程知識，風(fēng)景園林師對于這些技術(shù)細(xì)節(jié)不是很熟悉。雖然現(xiàn)在相關(guān)的服務(wù)可以直接定制，但風(fēng)景園林師仍須對物聯(lián)網(wǎng)傳輸方案有一定的了解，以保證跨專業(yè)合作的高效性（圖8）。

1）通信協(xié)議。在選擇傳感器時須明確設(shè)備串口形式，建議選擇RS-485端口，注意設(shè)備寄存器地址的重編以及空間位置的對應(yīng)，避免數(shù)據(jù)混淆或傳輸沖突。

2）傳輸模塊與網(wǎng)絡(luò)。根據(jù)監(jiān)測場地的尺度，合理選擇數(shù)據(jù)傳輸設(shè)備及數(shù)據(jù)傳輸網(wǎng)絡(luò)。成熟的數(shù)據(jù)采集、傳輸模塊有ZigBee模塊、數(shù)據(jù)傳輸單元（DTU）和遠(yuǎn)程終端單元（RTU），成熟的傳輸網(wǎng)絡(luò)有蜂窩網(wǎng)絡(luò)及低功耗廣域網(wǎng)絡(luò)（LPWAN）[27]。

3）云服務(wù)器。根據(jù)監(jiān)測指標(biāo)，對數(shù)據(jù)傳輸模塊完成采集腳本編寫及相關(guān)設(shè)置工作，并根據(jù)數(shù)據(jù)容量選擇合適算力的云服務(wù)器。

4）數(shù)據(jù)庫。在云服務(wù)器上搭載分布式數(shù)據(jù)庫，根據(jù)指標(biāo)類型設(shè)計合理的表單結(jié)構(gòu)，做好數(shù)據(jù)庫數(shù)據(jù)與真實(shí)空間的對位。

5）監(jiān)聽與解碼。傳輸模塊通過協(xié)議以字節(jié)流的形式將數(shù)據(jù)發(fā)送到服務(wù)器端口，工程師需要利用Python等編程語言在服務(wù)器端編寫監(jiān)聽程序。另外，這些十六進(jìn)制的字節(jié)流需要被解碼為十進(jìn)制數(shù)據(jù)后方可存儲到數(shù)據(jù)庫的指定位置中。

5 總結(jié)

通過對GSI監(jiān)測原理、原型及傳輸方案的論述中可以看出，在線監(jiān)測技術(shù)方案對于GSI實(shí)體空間干預(yù)（space intervention）較小，但能引發(fā)風(fēng)景園林師在設(shè)計思維和設(shè)計模式方面的重要轉(zhuǎn)變。

1）設(shè)計思維方面，在線監(jiān)測系統(tǒng)以較小的空間干預(yù)，實(shí)現(xiàn)了虛擬空間與實(shí)體空間的疊加。智慧城市建設(shè)中，傳統(tǒng)的空間設(shè)計手法不再是風(fēng)景園林師介入環(huán)境、帶來積極變化的唯一途徑。具備數(shù)據(jù)生成和自我反饋的未來景觀空間將為風(fēng)景園林師提供新機(jī)遇，也為城市感知、空間改善提供更多維度的數(shù)據(jù)資料。本研究中，GSI在線監(jiān)測拓寬了傳統(tǒng)雨洪管理的內(nèi)容，提高了設(shè)計標(biāo)準(zhǔn)，使得績效評估、韌性數(shù)據(jù)庫、解說教育、動態(tài)模型等多種應(yīng)用場景的實(shí)現(xiàn)成為可能[28]。

2）而在設(shè)計模式上，多學(xué)科的合作機(jī)制將改變風(fēng)景園林師的工作方式。GSI在線監(jiān)測這一場景下，由風(fēng)景園林師確定監(jiān)測目標(biāo)，統(tǒng)籌監(jiān)測平臺搭建；水文工程師提供技術(shù)框架，實(shí)驗(yàn)方案；電子工程師及設(shè)備廠商則主要提供傳感器及數(shù)據(jù)傳輸方案。

作為系列研究的最新進(jìn)展，與先前所完成的監(jiān)測實(shí)踐成果相比，本研究成果的推進(jìn)深化主要體現(xiàn)在：對監(jiān)測指標(biāo)、原理進(jìn)行了總結(jié)凝練，技術(shù)要點(diǎn)進(jìn)一步完善，補(bǔ)充了大量技術(shù)細(xì)節(jié)與計算方法；對監(jiān)測系統(tǒng)空間原型的總結(jié)及深入制圖，成果更具備推廣性和指導(dǎo)價值；進(jìn)一步探討在線監(jiān)測對設(shè)計思維及設(shè)計流程的影響。

最后，由于GSI系統(tǒng)的復(fù)雜性，筆者僅為風(fēng)景園林師提供了初步的開展GSI在線監(jiān)測指南，研究成果具有一定的局限性，主要體現(xiàn)在：1）提供的監(jiān)測方法只適用于1 hm2以內(nèi)的小尺度GSI在線監(jiān)測，不適用更大尺度的復(fù)雜GSI系統(tǒng)；2）提供的系統(tǒng)主要關(guān)注GSI的水文過程監(jiān)測，未涉及人流使用監(jiān)測和植物群落監(jiān)測； 3）僅提供了傳感器的選型、安裝要求及選點(diǎn)建議，未提及監(jiān)測點(diǎn)位分布密度問題，有待進(jìn)一步完善；4）提供的數(shù)據(jù)傳輸方案僅為多種可行傳輸方案中的一種。隨著5G及NB-IoT技術(shù)的更新，未來監(jiān)測數(shù)據(jù)傳輸、分析將有新的形式。

圖表來源：

圖1-1引自參考文獻(xiàn)[6]，其余均由作者繪制。

（編輯/王亞鶯）

Designing Online Monitoring Systems for Green Stormwater Infrastructure

In recent years, techniques of the internet of things(IoT) have been reforming the usage pattern of urban spaces and promoting online applications in landscape projects. Driven by these new technologies in smart cities, the data domain and the physical space have begun to merge and overlap with each other. Correspondingly, the urban landscape is now confronting the challenge of digital transformation and reconstruction[1-2]. To adapt efficiently to this trend, landscape architects have to step out of their comfort zone(traditional methods of space design) and actively collect information for evidence-based design, evaluation, and management[3].

This research integrates small-scale green stormwater infrastructure(GSI) with online monitoring systems, and provides a preliminary monitoring guidance for landscape designers. Firstly, we discuss the monitoring principles, indicators, sensors, and calculations, with which landscape architects are relatively unfamiliar. Secondly, we propose three monitoring prototypes(typical monitoring systems) of GSI that integrate sensing layers with real space. Finally, we provide an available data transmission paradigm based on multiple IoT equipment.

從理論上講，盒馬鮮生并不能完全算新零售業(yè)態(tài)，但它確實(shí)把原本“死氣沉沉”的萊蒙都會帶活了。無論從效益還是效率上看，盒馬鮮生都將零售業(yè)的水平帶上了新高度。它不但能在極短時間內(nèi)就實(shí)現(xiàn)盈利，還能將坪效做到普通超市的3倍以上。雖然現(xiàn)在討論新零售也太是否可以拯救傳統(tǒng)零售還為時過早，但把新零售做好本身就意味著效率與效益的雙豐收，萊蒙都會的商業(yè)轉(zhuǎn)型之路也值得一些正處在水深火熱中的傳統(tǒng)賣場學(xué)習(xí)。

We conducted this research based on a previous monitoring project at the Tsinghua campus[4]29. In this paper, we have demonstrated new outcomes about the monitoring principles, formulations, and design prototypes from a specific case study to general approaches.

1 GSI Online Monitoring

As an urban stormwater regulator, the GSI shows complicated hydrological responses to different rainfall events and a gradual functional decline after the construction. Designers need monitoring tools and feedbacks to conduct post-maintenance and performance evaluation[5]. However, due to a lack of feasible monitoring methods, the current design projects of GSI focus more on space construction and simulation rather than post analysis and evaluation. Unavailable feedbacks always lead to difficulties in maintenance and shallow understandings of what we have designed. Therefore, new urban GSI in smart cities must be integrated with online monitoring systems to avoid symbolic and debased expansions[6].

Nowadays, state-of-the-art tools of hydrological monitoring can be utilized to the followings: 1) Guide theoretical research of urban hydrological responses and reveal the mechanisms of runoff production in multi-scale urban areas[7-8]. 2) Assist the development of high-precision hydrological simulation models for comprehensive water management and disaster forecasting[9-10]. 3) Evaluate the performance of various green and gray infrastructure and uncover their functional attenuation[11-12]. 4) Develop new data transmission methods and monitoring equipment[13-14].

Although related technologies are being continually improved, monitoring methods in hydrological and IoT disciplines have not been applied to design smart GSI. There is a lack of relevant guidelines of how to integrate IoT solutions with the spatial design. To bridge this gap, landscape architects should act as a coordinator to explore the approaches of GSI online monitoring. We list several typical projects of IoT-based GSI lead by landscape architects here: 1) Intelligent stormwater management innovations and campus pilot projects led by the University of Michigan and Villanova University[15]. 2) An intelligent green roof designed by a team from the University of Maryland[16]. 3) Water quality monitoring projects conducted by an American company named Opti[17](Fig. 1-1). 4) Stormwater monitoring system at Shengyinyuan, Tsinghua University[4]30(Fig. 1-2).

Different from hydrological experiments of GSI[11,18], landscape architects carry out outdoor monitoring to verify the rationality of the layout, ascertain runoff control, and evaluate landscape performance. Outdoor monitoring conditions bring the following new challenges: 1) A stable and effective long-term data system is needed to ensure an efficient data collection. 2) Architects have to utilize available sensors and arrange them appropriately in a real site. 3) Designers are required to balance the practicability and visual beauty and achieve the purpose of ecological education and demonstration.

2 Indicators, Principles and Methods

2.1 Indicators

First of all, the monitoring indexes should be structured data with a short step. Structured data has the advantage of being easily entered, stored, queried, and analyzed, which can be presented logically with a two-dimensional table. Most available sensors obtain structured and mutational indicators such as temperature, humidity, locations, light intensity, pressure, velocity, and precipitation[19]. These signals transmit in the form of electrical or electromagnetic waves with a small data volume. Tab. 1 lists the structured data that can be collected for a GSI monitoring system. We divided them into four categories: meteorology, water quality, surface runoff reduction, and underground runoff control. These indexes can comprehensively reflect the whole hydrological process of GSI.

2.2 Calculating Principles

Quantity and quality indicators of the inflow/outflow are essential for designers to ascertain the runoff reduced by GSI. Unlike experiments of hydrological mechanisms[20], it is difficult for architects to conduct comprehensive monitoring with strict experimental conditions in a complicated outdoor site. Therefore, designers can simplify the hydrological processes appropriately and obtain core indexes. To calculate the inflow/outflow, we provide a simplified method based on the classic water balance equation. In this equation, crucial indicators of precipitation, soil-water storage, infiltration, and overflow can be acquired via sensors[21-22].

For a closed-basin GSI, inflowQinequals to the rainfallQr. To further improve the accuracy, we can consider the canopy interception and monitor the throughfallQca. Then, the inflow can be calculated according to the area ratio of the headroom and the canopy cover(Equation 1).iis the area coefficient of the headroom rainfall with no interception. For open GSI that collects runoff from unclear watersheds, we can sum up the following indicators to estimate the inflow: the increment of soil-water capacityΔQsw, the amount of infiltrationQinf, and the amount of overflowQof(Fig. 2). In Equation 2, ΔQswis the product of the soil moisture incrementΔIswand the thicknessHsof the soil layer.

In Equation 2, ΔQswis a simplification of the complex process in the soil filter during a rainfall event. Qinfrefers to the runoff that penetrates the filter layer into the drainage layer, which can be collected by pipes and a monitoring well.Qinfcan be regarded as runoff reduction instead of a part of outflow. And the outflow merely means the overflow to the municipal pipe networks. Besides, the evaporation during rainfall events is negligible because it is often smaller than the error of the devices.

Based on the principles and indicators described above, relevant data can also be calculated. 1) The soil temperature and humidity can reflect the sunlight and temperature conditions, assisting analyses of plant growth. 2) As the rainfall starts, the soil moisture increases and uncovers the main inflow direction in the meanwhile. Designers can then verify the rationality of the elevation design. 3) Soil moisture fluctuations can also reveal the horizontal and vertical movement of the runoff in the soil. 4) If the runoff accumulates in the ponding area(captured by level sensors),Qswat that time reveals the maximum water capacity of the GSI, which corresponds to the stable infiltration rate as well.

2.3 Monitoring Methods and Sensors

2.3.1 Meteorology and Canopy Interception

Designers can obtain meteorological information using portable weather stations, which can integrate multiple sensors(temperature, humidity, wind direction, solar radiation, rainfall, air pressure, or PM2.5/PM10) and achieve data transmission via general packet radio services(GPRS). Bucket and infrared rain gauges are standard tools for monitoring precipitation. Engineers should guarantee that rain gauge is installed at an appropriate height(above 0.7 m), and there are no surrounding shelters within 10 m(Fig. 3-1). In contrast, the evaporation sensor does need a shelter to prevent the rainwater from entering(Fig. 3-2). To conclude, the arrangement of a weather station should adapt to the specific conditions of the site, and designers can refer to theChinese Industry Standard GB/T21978.5—2014for more details.

Canopy interception affects the hydrological processes of the site by intercepting and redistributing the rainfall. The monitoring method of the throughfall is the same as the rainfall. However, the tipping-bucket gauge is not suitable for long-term monitoring of the throughfall since the bucket is placed right under the canopy, which can be filled up by fallen leaves. Photoelectric sensors such as infrared rain gauges(Fig. 3-3) are better choices.2.3.2 Soil Moisture, Infiltration and Overflow

Soil moisture sensors can monitor soil-water content(volume fraction). Designers can assemble many of them in different soil depths to record the changes. Measuring soil infiltration requires collecting pipes and monitoring wells. We can collect a small part of the penetration with this vessel system to estimate the overall amount(Fig. 4-1). Flowmeters and level meters can be utilized to probe both the infiltration and the overflow.

1) Flowmeters. Flowmeters measure the flow rate of the runoff directly through pipes or channels. Standard ultrasonic sensors are the turbine and the vortex flowmeters. These spacesaving sensors are installed at the pipes, and the measurement requires full-pipe flows(Fig. 4-2). Open channel flowmeters, another type of probes, measure the flow using multiple flumes(triangular, rectangular, and Parshall). The probe does not need to touch the water by converting the level to flow volume with specific formulations. GSI with a large overflow should be equipped with open channel meters. Both methods above demand smooth drainage, and backflow should be strictly prevented(Fig. 4-3).

2) Level sensors. When using level sensors, we can establish a monitoring well as a container so that the runoff can be stored temporarily and form a liquid level. Additionally, an empty pump is needed to drain the water after a rainfall event(Fig. 4-4). Ultrasonic level gauges are highly recommended for GSI monitoring, with errors less than 1 cm. Moreover, a minimum distance of 300-500 mm between the objects and the sensor is requisite. That is because the reflected wave and the transmitted wave near the sensor can easily overlap with each other, forming a monitoring blind zone. Another tip to emphasize, we cannot apply capacitance level gauges(sensors placed in a water container) to measure the level since the error is over 50 mm[23].

2.3.3 Water Quality

Standard analysis of water quality includes sampling, cryopreservation, and experimental determination, which is relatively accurate but timeconsuming. Online sensors probe conductivity and spectrophotometry to estimate different indicators of water quality. Although the absolute values show less accuracy, online sensors can disclose the sequential variation of different indicators. Sensors of water quality are always assembled with floating platforms[9], and key monitoring indicators should be selected according to the characteristics of the water body.

1) We suggest monitoring dissolved oxygen(DO) in the lakes, ponds or rivers with a supply of reclaimed water. 2) Purification wetlands that store urban sewage should be equipped with sensors of total nitrogen(TN) and total phosphorus(TP). 3) In urban water parks for children, sensors of chemical oxygen demand(COD) and biological oxygen demand(BOD) are recommended. 4) Scenic lakes and rivers that collect urban stormwaters demand sensors of total suspended solids(TSS). Since the sensors are immersed in the water for a long time, pollutants, microorganisms, and algae can easily adhere to the probes, decreasing the accuracy. Therefore, regular maintenance is essential when using these sensors[24-25].

3 Three Monitoring Prototypes of GSI

In real monitoring practice, principles and formulations are far from enough. The combination of sensors and real GSI is the most challenging job for designers. Although architects do not need to understand all the details, they should balance the space design and the monitoring system. Therefore, prototypes of sensing assembly under specific conditions are much needed. Given the lack of such guidelines, we attempt to propose three archetypes of GSI online monitoring, considering the following conditions: open/closed watershed, permeable/impermeable, stable/unstable water bodies.

3.1 Impermeable GSI with Unstable Water Bodies

Typical GSI in this category includes extensive green roofs, roof gardens, and viaduct green roofs in closed watersheds and underground garage roofs, ponds, and wetlands with anti-seepage treatment in open watersheds. The depth of soil filter is always less than 1.5 m, and plants are mainly shrubs and ground covers. In this prototype, designers should focus more on monitoring fluctuations of the heat, soil moisture, and overflow[20,26]. Manual sampling and laboratory analysis are recommended to evaluate the performance of water quality control. Fig. 5 illustrates the assembly of sensors in this prototype.

1) Closed watershed. Precipitation is the only inflow source. Architects should remove unnecessary outflow ports to ensure a precise and simple routine of the runoff(reduce the number of sensors).

2) Open watershed. The inflow is difficult to determine under this condition. Designers can sum up the increment of soil-water capacityΔQswand the overflowQof(Equation 3).

3.2 Permeable GSI with Unstable Water Bodies

Typical GSI in this category includes small rain gardens, bio-retention ponds in closed watersheds and grass ditches, rainwater wetlands in open watersheds. The infiltration capacity of permeable GSI is the core indicator to evaluate runoff reduction. Fig. 6 illustrates the layout of sensors in this prototype to give more details.

1) Infiltration. Underground pipes collect part of the infiltration and form a communicating vessel with the monitoring well. The water level of this vessel reflects the infiltration per unit area. Pumps in the monitoring well serve to vacuum up the water after a rainfall event.

2) Overflow. Excessive runoff can overflow to a pipe(turbine/vortex flowmeters), or a channel(open channel flowmeters). Designers should ascertain available locations of sensors and choose appropriate flowmeters.

3) Throughfall. Canopy trees affect the rainfall process. Therefore, we recommend the designers to monitor the throughfall under different species of trees.

3.3 GSI with a Stable Water Body

For GSI with stable water bodies, such as riverfront greenbelts, wetland parks, and wet ponds, water quality is an essential indicator reflecting the effects of pollution control. Fig. 7 demonstrates the layout of sensors in this prototype to give more details.

1) Water quality. Online water quality monitoring is conducted using small floating platforms or unmanned surface vehicles(USV). To improve the accuracy, we suggest researchers to test some water samples in the laboratory every month.

2) Water capacity. Level meters are applied to measure the volume fluctuations of small lakes or ponds. For rivers, it is more applicable to observe the changes with a flowmeter.

3) Runoff coefficient. Plant communities in the waterfront greenbelt has a significant impact on the hydrological process. Therefore, monitoring of the canopy interception is highly recommended. Besides, designers can select small closed watersheds to measure the overall runoff coefficient of the waterfront greenbelt with open channel flow meters(rectangular flumes).

4 IoT Data Transmission

Details of communicating protocols, sensors, and database programming involved in the data transmission are quite intricate for most landscape architects. Although professional enterprises can provide related services, landscape architects are strongly advocated to have a basic understanding of the IoT data transmission to ensure an efficient interdisciplinary cooperation(Fig. 8).

1) Protocols. Select sensors with RS-485 ports(a common serial-port based protocol). Pay attention to reprogram the registered address of the device. Guarantee the correspondence between the spatial location and the digital address to avoid conflicts.

2) Transmission platforms and wireless networks. Accessible data collection and transmission modules include ZigBee module, data transfer unit(DTU), and remote terminal unit(RTU). Available wireless networks are cellular network and low power wide area networks(LPWANs)[27].

3) Cloud storage. Designers can purchase a third-party cloud server with appropriate computing power according to the data capacity. Scripts for data acceptance and related settings are also required.

4) Database. A distributed database with precise data alignments should be established on the cloud server.

5) Listeners and decoders. The cloud server receives byte streams of the data from the data transfer units. Therefore, a listener is needed. After registration, a decoder transfers these hexadecimal data to decimal and stores them in the database.

5 Conclusions

In this paper, we discuss the indicators, formulations, prototypes, and data transmission methods of GSI online monitoring. Although IoT solutions conduct inconspicuous interventions to the real space, the online systems have changed the way we think and design.

1) Design thinking. In a smart city, IoT systems are always imperceptible while realizing the superposition of the virtual domain and the physical space. Traditional space design is no longer the only path for landscape architects to bring about positive changes in our cities. Future landscapes will show the capabilities of data generation and self-reaction, providing more data for urban perception and spatial improvement. Therefore, GSI online monitoring described in the research can broaden the scope of stormwater management and induce improvements in design standards, performance evaluation, design database, education, and dynamic urban models[28].

2) Design modes. A multi-disciplinary cooperation mode needs to be introduced to the landscape discipline. In the scenario of GSI online monitoring, landscape architects determine the monitoring targets and coordinate the construction of the monitoring platform. Hydrological and electronics engineers provide technical support of sensing assembly and data transmission.

Compared to previous monitoring practice in Shengyinyuan, our team has made the following progress in this paper: Explicit monitoring principles are demonstrated, and a large number of technical details and formulations are supplemented. We summarize three monitoring prototypes and provide detailed guidelines for landscape architects. We further explore the impact of online monitoring on design thinking and design modes.

Due to the complexity of the urban GSI system, there are several limitations in the preliminary guidelines we have provided: 1) The prototypes demonstrated in this article aim at small-scale urban GSI within 1 hectare instead of large-scale GSI systems. 2) The monitoring system proposed focuses more on the hydrological process rather than plant communities and human activities. 3) We expound core installation requirements of the sensors in this paper. However, we do not involve the distribution density of monitoring points, which need to be further improved. 4) The data transmission scheme described is only one of many feasible plans. In the future, there will be new forms of data transmission, including 5G and NBIoT.

Sources of Figures and Table:

Fig. 1-1 ? reference[6], others? the authors.